AMBIENT-6G D5.1: Use cases, KPIs and KVIs

Towards standardized 6G connectivity for ambient-powered energy neutral IoT devices Deliverable D5.1 Use cases, KPIs and KVIs AMBIENT-6G project has received funding from the Smart Networks and Services Joint Undertaking (SNS JU) under the European Union’s Horizon Europe research and innovation programme under Grant Agreement No 101192113. Date of delivery: 30th Aug, 2025 Version: 1.1 Project reference: 101192113 Call: HORIZON-JU-SNS-2024 Start date of the project: 1st Jan, 2025 Duration: 36 months D5.1 - Use cases, KPIs and KVIs Document properties Document Number: D5.1 Document Title: Use cases, KPIs and KVIs Editor(s): Rubén Altuna Pérez (TEL) Authors: Dragan Subotic (IMEC), Ritesh Singh (IMEC), Samer Nasser (IMEC), Henrique Duarte Moura (IMEC), Andrey Belogaev (IMEC), Maby Johns (IMEC), Riku Jäntti (AAU), Osmel Martínez Rosabal (OUL), Efstathios Katranaras (SEQ), Miltiadis Filippou (WIN), Aimilia Bantouna (WIN), Sokratis Barmpounakis (WIN), Vasiliki Lamprousi (WIN), Panagiotis Demestichas (WIN), Bert Cox (KUL), Daniel Pöhl (NXP), Jarne Van Mulders (KUL), Guus Leenders (QKS), Benjamin J. B. Deutschmann (TUG), Lukas D’Angelo (TUG), Rubén Altuna Pérez (TEL), Bikramjit Singh (LMF), Hamza Khan (LMF). Contractual Date of Delivery: 31st Aug, 2025 Dissemination level: PU Status: Final Version Revision: 1.1 Filename: AMBIENT-6G_5.1_v1.1 Revision History Revision Date Issued by Description v0.1 17/03/2025 TEL Baseline draft for the deliverable. v0.2 24/03/2025 TEL Merge contributions to Use Cases. v0.3 12/05/2025 TEL Inputs in Introduction, Technical Description, Use Case SotA and Conclusions. v0.4 26/05/2025 TEL Internal review. v0.5 24/06/2025 TEL External review. v1.0 01/08/2025 TEL Final version. v1.1 11/11/2025 TEL Reviewed final version (authors list updated). Page I D5.1 - Use cases, KPIs and KVIs Abstract This deliverable describes relevant use cases in different vertical areas where energy-neutral device operation can bring particular benefits in terms of ease of use and maintenance, cost, and ecological footprint. Keywords AMBIENT-6G, Ambient-IoT, Energy-Neutral Devices, Use Cases, Key Performance Indicators, Key Value Indicators. Disclaimer Funded by the European Union. The views and opinions expressed are however those of the author(s) only and do not necessarily reflect the views of AMBIENT-6G Consortium nor those of the European Union or Horizon Europe SNS JU. Neither the European Union nor the granting authority can be held responsible for them. Internal reviewers Henrique Duarte Moura (IMEC) Priyesh Pappinisseri Puluckul (IMEC) Rubén Altuna Pérez (TEL) External reviewers Andrey Belogaev (IMEC) Lieven De Strycker (KUL) Gilles Callebaut (KUL) Page II D5.1 - Use cases, KPIs and KVIs Executive Summary This document constitutes the issue of Deliverable D5.1: ‘Use cases, KPIs and KVIs’, within the framework of the project titled "AMBIENT-6G – Towards standardized 6G connectivity for ambiently-powered energy-neutral IoT devices" (Project Acronym: AMBIENT-6G; Grant Agreement No: 101192113). This document presents a comprehensive analysis of the potential of Ambient Internet of Things (A-IoT) in enabling innovative and sustainable use cases, aiming at supporting future developments and standardization efforts within the European context. It is structured into five chapters, each addressing a key aspect of the topic. •Chapter 1introduces the motivation behind the work, outlines the main objectives, and presents the structure of the document. •Chapter 2provides a technical overview of A-IoT and the core technologies it builds upon, offering a conceptual foundation for the rest of the document. •Chapter 3reviews the current state of the art in defining use cases for A-IoT, focusing on international standardization bodies and relevant European projects, both completed and ongoing. •Chapter 4presents the proposed A-IoT use cases in detail. Each use case includes an analysis of functional requirements, key performance indicators (KPIs), and key value indicators (KVIs), covering environmental, social, economic, and innovation-related impacts. •Chapter 5summarizes the main conclusions drawn from the work described in the document. Page III D5.1 - Use cases, KPIs and KVIs Contents 1 Introduction 1 1.1 Motivation..................................... 1 1.2 ScopeandObjectives............................... 1 1.3 Structure of the Document . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2 Ambient Internet of Things: Technical Description 3 2.1 Holistic Energy Neutral Device architecture . . . . . . . . . . . . . . . . . . . 3 2.1.1 Encompassing Performance vs. Cost vs. Material Usage Trade-off . . . 4 2.1.2 Circuits for Energy Storage, Harvesting, and Management . . . . . . . 5 2.1.3 Enabling Energy Neutral Device Connectivity . . . . . . . . . . . . . . 6 2.1.4 Secure Low-power Protocol Design . . . . . . . . . . . . . . . . . . . 6 2.1.5 Infrastructure Enablers for Wireless Power Transfer . . . . . . . . . . . 7 2.2 Cloud-Edge Device Orchestration, Offloading and On-device Machine Learning 8 3 Use Cases: State of the Art 11 3.1 International Standardization Bodies . . . . . . . . . . . . . . . . . . . . . . . 11 3.2 European Research and Innovation Projects . . . . . . . . . . . . . . . . . . . 13 4 Use Cases: AMBIENT-6G Vision 16 4.1 Interpretation of the Use Case Analysis . . . . . . . . . . . . . . . . . . . . . 16 4.2 ElectronicShelfLabel............................... 22 4.2.1 Description ................................ 22 4.2.2 Functional Requirements . . . . . . . . . . . . . . . . . . . . . . . . . 22 4.2.3 Key Performance Indicators . . . . . . . . . . . . . . . . . . . . . . . 25 4.2.4 Key Value Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 4.3 SensorsinSmartHomes ............................. 28 4.3.1 Description ................................ 28 4.3.2 Functional Requirements . . . . . . . . . . . . . . . . . . . . . . . . . 29 4.3.3 Key Performance Indicators . . . . . . . . . . . . . . . . . . . . . . . 32 4.3.4 Key Value Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 4.4 Smart Bridge Health Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . 35 4.4.1 Description ................................ 35 4.4.2 Functional Requirements . . . . . . . . . . . . . . . . . . . . . . . . . 36 4.4.3 Key Performance Indicators . . . . . . . . . . . . . . . . . . . . . . . 39 4.4.4 Key Value Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 4.5 Personal Belongings Finding . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 Page IV D5.1 - Use cases, KPIs and KVIs 4.5.1 Description ................................ 41 4.5.2 Functional Requirements . . . . . . . . . . . . . . . . . . . . . . . . . 42 4.5.3 Key Performance Indicators . . . . . . . . . . . . . . . . . . . . . . . 44 4.5.4 Key Value Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 4.6 In-body or Wearable Medical Sensors . . . . . . . . . . . . . . . . . . . . . . . 47 4.6.1 Description ................................ 47 4.6.2 Functional Requirements . . . . . . . . . . . . . . . . . . . . . . . . . 48 4.6.3 Key Performance Indicators . . . . . . . . . . . . . . . . . . . . . . . 51 4.6.4 Key Value Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 4.7 SmartAgriculture................................. 53 4.7.1 Description ................................ 53 4.7.2 Functional Requirements . . . . . . . . . . . . . . . . . . . . . . . . . 54 4.7.3 Key Performance Indicators . . . . . . . . . . . . . . . . . . . . . . . 56 4.7.4 Key Value Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 4.8 Asset, Product, Tool, and Item Tracking . . . . . . . . . . . . . . . . . . . . . 59 4.8.1 Description ................................ 59 4.8.2 Functional Requirements . . . . . . . . . . . . . . . . . . . . . . . . . 59 4.8.3 Key Performance Indicators . . . . . . . . . . . . . . . . . . . . . . . 63 4.8.4 Key Value Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 4.9 MuseumGuide .................................. 65 4.9.1 Description ................................ 65 4.9.2 Functional Requirements . . . . . . . . . . . . . . . . . . . . . . . . . 66 4.9.3 Key Performance Indicators . . . . . . . . . . . . . . . . . . . . . . . 69 4.9.4 Key Value Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 4.10End-to-endLogistics ............................... 72 4.10.1 Description ................................ 72 4.10.2 Functional Requirements . . . . . . . . . . . . . . . . . . . . . . . . . 73 4.10.3 Key Performance Indicators . . . . . . . . . . . . . . . . . . . . . . . 75 4.10.4 Key Value Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 4.11 Industrial Predictive Maintenance . . . . . . . . . . . . . . . . . . . . . . . . . 77 4.11.1 Description ................................ 77 4.11.2 Functional Requirements . . . . . . . . . . . . . . . . . . . . . . . . . 78 4.11.3 Key Performance Indicators . . . . . . . . . . . . . . . . . . . . . . . 80 4.11.4 Key Value Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 4.12 Cooperative Mobile Robots . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 4.12.1 Description ................................ 82 4.12.2 Functional Requirements . . . . . . . . . . . . . . . . . . . . . . . . . 83 4.12.3 Key Performance Indicators . . . . . . . . . . . . . . . . . . . . . . . 86 4.12.4 Key Value Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 5 Conclusions 89 Page V D5.1 - Use cases, KPIs and KVIs Glossary 3GPP 3rd Generation Partnership Project. 4G fourth-generation. 5G fifth-generation. 6G sixth-generation. ADC analog-to-digital converter. AI artificial intelligence. A-IoT Ambient Internet of Things. AOA angle-of-arrival. AP access point. BLE Bluetooth low energy. BS base station. CDMA code division-multiple access. CJT coherent joint transmission. CN core network. CO continuous operation. COTS commercial off-the-shelf. CRC cyclic redundancy check. CSI channel state information. D-MIMO distributed MIMO. DB duty-cycled barebone. DLI direct link interference. DP duty-cycled persistent. DPPM dynamic power path management. DTLS Datagram Transport Layer Security. Page VI D5.1 - Use cases, KPIs and KVIs EMF Electromagnetic Field. eNB Evolved Node B. END energy-neutral device. EoL end of life. ESL electronic shelf label. ET Energy Transmitter. ETSI European Telecommunications Standards Institute. FCC Federal Communications Commission. FEC forward error correction. FFT fast Fourier transform. GDPR General Data Protection Regulation. GNSS global navigation satellite system. GPS Global Positioning System. HF high frequency. IC integrated circuit. IEEE Institute of Electrical and Electronics Engineers. IMU inertial measurement unit. IoT Internet of Things. ISAC Integrated Sensing and Communication. ISM industrial, scientific and medical. KPI key performance indicator. KVI key value indicator. LCA life cycle assessment. LCD liquid crystal display. LED Light Emitting Diode. LoRa long range. LoRaWAN long-range wide-area network. LoS line-of-sight. LPWAN low-power wide-area network. LTE Long Term Evolution. Page VII D5.1 - Use cases, KPIs and KVIs M&O management and orchestration. MAC Medium Access Control. MCU microcontroller unit. MEMS micro-electromechanical system. MIMO multiple-input multiple-output. ML machine learning. NB-IoT narrowband IoT. NFC near-field communication. NOMA non-orthogonal multiple access. OBU on-board unit. OFDM orthogonal frequency-division multiplexing. OLoS obstructed-line-of-sight. OT Operational Technology. OTA over-the-air. OTA-TinyML Over-the-air TinyML. PHY physical. PMIC power management integrated circuit. POS point-of-sale. QoE Quality-of-Experience. QoS quality-of-service. QR quick response. RAM random-access memory. RBAC Role-Based Access Control. RF radio frequency. RFID radio frequency identification. RIS reflective intelligent surface. ROI return of investment. RSS received signal strength. RSSI received signal strength indicator. RTI reusable transport item. Page VIII D5.1 - Use cases, KPIs and KVIs technologies continue to evolve, they will play a pivotal role in the widespread deployment of sustainable and autonomous A-IoT systems. 2.1.3 Enabling Energy Neutral Device Connectivity Ambient-powered ENDs suffer from intermittently turning off due to insufficient power for communication and operation of the radio and microcontroller unit (MCU). The inconsistency of the energy harvesting is the major challenge, as ENDs cannot rely on a stable harvesting rate to predict if they have enough power to perform a communication cycle. The most challenging is the support of active ENDs that can generate the signal on their own instead of communicating using back-scattering. Feasibility of different connectivity protocols have been studied in literature: long-range wide-area network (LoRaWAN) [9]–[11], Bluetooth low energy (BLE) [12]– [14]narrowband IoT (NB-IoT): [15]–[17]. In 6G, active ENDs are supposed to function similar to legacy IoT devices with a significantly lower complexity [18]. Legacy IoT devices in cellular networks operate according to NB-IoT standard. The major part of the energy consumption of NB-IoT devices is spent on channel access. The default channel access mechanism for these devices after returning from the sleep state is random channel access, when a random preamble from a predefined set is sent to request a grant. However, to enable low-latency critical IoT applications, two other channel access mechanisms has been introduced: grant-free and fast uplink grant, also known as configured grant. When grant-free access is used, devices are allowed to transmit data in randomly selected resources without negotiation with the base station. Therefore, the latency is significantly reduced, but under an expense of a higher chance of unsuccessful transmissions due to collisions. Fast uplink grant allows the base station to provide grants to the associated devices in advance, before data packets appear in their queues. The efficiency of this mechanism significantly depends on how accurately the base station predicts the activity of devices. For ENDs, the selection of the channel access mechanism is especially important, as different mechanisms require different energy consumption at different network conditions. Studies [16], [17] compare the performance of these mechanisms in terms of outage probability and energy consumption for ambient powered ENDs. Both studies reveal a high potential for the fast uplink grant scheme, especially for predictable traffic patterns. 2.1.4 Secure Low-power Protocol Design ENDs have an extremely low energy budget and thus are typically resource constrained, making them less suited to implementing traditional security mechanisms [19]. Due to limited processing power, memory, and energy availability, adding complex cryptographic protocols has been challenging. Since communication typically accounts for the largest share of power consumption in IoT devices [20], security solutions that increase communication overhead or require extensive computation are typically viewed as impractical or too costly for such nodes. This perspective especially applies to security frameworks that involve frequent or continuous connections with other devices or cloud infrastructures. Protocols like Transport Layer Security (TLS) and Datagram Transport Layer Security (DTLS), while providing strong security guarantees in traditional IoT or Internet systems, tend to impose significant computational and communication overhead that is unusable for ultra-low-power systems [21]. The handshakes, Page 6 of 94 D5.1 - Use cases, KPIs and KVIs key exchanges and multiple message exchanges involved can quickly drain the limited energy budgets of ENDs. To overcome these limitations, research has been shifted toward lightweight cryptographic protocols and efficient key management schemes designed specifically for constrained devices. In the context of low-power wide-area networks (LPWANs) — such as LoRaWAN,NB-IoT, and Sigfox — security is typically integrated at the network and application layers with a strong emphasis on minimizing energy consumption [22]. For example, LoRaWAN employs symmetric AES-128 encryption for both network authentication and payload confidentiality, relying on predistributed keys and a lightweight joint procedure. NB-IoT, as part of the Long Term Evolution (LTE)/fifth-generation (5G) family, inherits the robust authentication and encryption features of the cellular stack while tailoring its signaling and power behavior for IoT scenarios. Sigfox, despite its ultra-narrowband and highly constrained message format, supports optional payload encryption and sequence-based replay protection mechanisms. These protocols exemplify how security can be designed to coexist with strict energy budgets — limiting handshake complexity, minimizing message exchanges, and relying on pre-shared or efficiently negotiated keys. Although such mechanisms may not match the flexibility or end-to-end guarantees of traditional Internetgrade protocols like TLS, they represent a balance between security and energy efficiency for ENDs. In addition to protocol-level approaches, there is growing interest in physical (PHY) layer security techniques, especially for backscatter and energy-harvesting devices. PHY security leverages the inherent characteristics of the wireless medium - such as channel randomness or hardwarespecific imperfections (e.g., radio frequency fingerprints) - to enable lightweight authentication and key generation without relying on heavy cryptographic computations. These techniques are particularly attractive for ultra-low-power and intermittently powered devices, as they can reduce or even eliminate the need for energy-expensive message exchanges [23]. Although current implementations face challenges related to environmental variability and device stability, recent research has shown encouraging progress in improving the reliability and robustness of PHY-layer security mechanisms [24], [25]. 2.1.5 Infrastructure Enablers for Wireless Power Transfer RF-WPT is a promising solution for powering large-scale ENDs ecosystems, overcoming the limitations of traditional battery-powered ENDs in terms of maintenance, scalability, and environmental impact. It stands among WPT technologies for its ability to support simultaneous charging of multiple ENDs, enable mobility, and operate in non-line-of-sight conditions—albeit with some efficiency trade-offs. Unlike ambient energy harvesting systems, the performance of WPT-enabled networks is dictated by the end-to-end conversion efficiency—which encompasses the Energy Transmitter (ET), the wireless channel, and the RF energy harvesting circuits— safety regulations, and coverage. This is because the ETs provide a controllable contactless power delivery to sustain the long-term operation of ENDs. Conveniently deploying multiple spatially distributed ETs is critical to eliminate blind spots in the network and distribute the energy according to the application requirements. Robotic WPT implementations with moving or flying ETs [26] bring significant advantages over traditional static ETs due to their ability to dynamically shorten the charging distance. Moreover, robotic WPT potentially reduces deployment costs as fewer ETs may be required to accomplish the same Page 7 of 94 D5.1 - Use cases, KPIs and KVIs task. More importantly, robotic WPT can cope more efficiently with temporal service requirements. Distributed ET infrastructures inherently have physically large apertures, which benefit WPT through (i) better regulatory compliance, (ii) higher WPT efficiency, (iii) inherent interference mitigation and channel hardening [27]. Coherent joint transmission (CJT) is an emerging paradigm for distributed multiple-input multiple-output (MIMO) architectures, promising jointly phase-coherent downlink beamforming with spatially distributed ETs, but necessitates establishing tight temporal synchronization and phase calibration of ETs with distributed clocks [28], [29]. Energy beamforming plays a crucial role in enhancing end-to-end conversion efficiency without requiring an increase in the ETs’ transmit power. Efforts to develop cost-effective beamforming solutions primarily focus on two key domains: system architecture and signal processing optimization. The former research direction focuses on direct modifications to the architecture of the ETs, whereas the latter focuses on minimizing the overhead of the signal processing algorithm. Popular techniques in the architectural domain include the use of low-resolution analog-todigital converters (ADCs) and simplifications to the analog front-end via antenna selection architectures, parasitic arrays, and even implementations without RF chains. Motivated by the idea of hybrid beamforming, recent efforts have focused on developing novel architectures that can efficiently handle numerous antennas with a limited number of RF chains. This includes lens antenna arrays [30], dynamic metasurface antennas [31], and reflective intelligent surface (RIS)-equipped ETs [32]. Other efforts, such as the development of movable antennas, have been directed to give the ETs sufficient flexibility to explore the channel in the quest for the best configuration that increases the harvested power and with a limited number of antennas. Another promising research direction that allows the use of energy-efficient low-resolution ADCs is direct link interference (DLI) mitigation in bistatic or multistatic ET infrastructures [33], [34]. From the signal processing perspective, the main challenge of energy beamforming lies in balancing computational efficiency with performance gains, necessitating innovative approaches to mitigate complexity. Notably, the benefits of accurate channel state information (CSI)-based strategies quickly vanish, and may even reverse, as the number of ENDs increases due to the energy-demanding training process. That is why alternative beamforming strategies have been proposed to rely on statistical CSI [35], received energy feedback [36], and the positions of the ENDs [37], which are easier to acquire and vary slowly. 2.2 Cloud-Edge Device Orchestration, Offloading and Ondevice Machine Learning Cloud-edge device orchestration and offloading are critical for low-power A-IoT devices due to their inherent resource constraints, including limited computational power, memory, and energy supply. The traditional cloud-centric AI paradigm, which offloads all data processing to centralized servers, introduces significant latency, bandwidth limitations, and privacy concerns, making it unsuitable for real-time and sensitive A-IoT applications. However, by distributing intelligence closer to the device, at the edge, these limitations can be mitigated. The primary importance of cloud-edge orchestration and offloading for A-IoT devices lies in enabling on-device and edge intelligence, which significantly improves response times by eliminating Page 8 of 94 D5.1 - Use cases, KPIs and KVIs the need for constant data transmission to the cloud. These transmissions are costly for ENDs Local processing allows devices to analyse sensory inputs and execute actions instantaneously, fostering greater autonomy and efficiency. Furthermore, processing data at the device enhances privacy by reducing the exposure of sensitive information across networks. It also enables personalization, allowing devices to adapt continuously to user behaviours and preferences without requiring constant connectivity. Crucially, by reducing frequent data transmissions, device intelligence contributes to substantial energy efficiency, a vital factor for battery-operated IoT devices and essential for ENDs. However, the local processing comes at a cost of higher complexity and energy/resource consumption. Therefore, it is necessary to analyse whether these adverse effects are compensated by the benefits of local processing. On-device ML for ENDs is a rapidly advancing field that enables intelligent data processing directly on ultra-low-power IoT nodes. These devices operate under severe constraints in terms of energy, memory, and compute capacity. To address these limitations, researchers have developed highly optimized ML models using techniques, such as quantization, pruning, and knowledge distillation, allowing inference to run on microcontrollers with as little as 32–256 KB of random-access memory (RAM). Two foundational tools in this space are TensorFlow Lite for Microcontrollers, which enables the deployment of compact ML models on bare-metal systems without an operating system [38] and CMSIS-NN, a library of highly optimised neural network kernels for ARM Cortex-M processors that significantly boosts inference efficiency on embedded hardware [39]. These frameworks are often used together to deliver real-time ML capabilities on devices with minimal energy budgets. These capabilities are being applied across a wide range of use cases. In cooperative mobile robotics, ENDs equipped with on-device ML can perform local obstacle detection or gesture recognition, enabling real-time interaction without relying on continuous connectivity. In smart agriculture, devices can classify soil moisture levels or detect crop diseases using lightweight classifiers. Wearable medical sensors use on-device ML to detect anomalies in heart rate or motion patterns, while in smart logistics, ENDs monitor package conditions and predict equipment failures. Even in cultural heritage settings, such as museums, ambient powered tags with embedded ML could personalize visitor experiences by recognizing user behaviour or preferences. These applications benefit from the ability to process data locally, which reduces latency, preserves privacy, and minimizes energy consumption [40]. Recent research highlights the growing interest in enabling not just inference but also on-device learning. Techniques for continual learning and federated learning are being adapted to operate under the constraints of ENDs, allowing devices to personalize models without transmitting raw data. For example, the authors in [41] provide a comprehensive survey of on-device ML from an algorithmic and learning theory perspective, emphasizing the importance of resourceconstrained learning. Meanwhile, newer work explores the feasibility of on-device training and adaptation in real-world systems [42]. As 6G and A-IoT ecosystems evolve, on-device ML is expected to become a foundational capability, enabling scalable, autonomous, and sustainable intelligence across sectors ranging from industrial automation and smart homes to healthcare and environmental monitoring. On the other hand, there are some advantages to processing data at the edge. A primary benefit is that computationally intensive tasks can be shifted from local devices to more powerful edge servers or cloud infrastructure. Offloading from the device effectively overcomes inherent hardware limitations of the IoT devices, allowing them to participate in complex analytical tasks Page 9 of 94 D5.1 - Use cases, KPIs and KVIs and machine learning inferences that would otherwise be beyond their native processing capabilities, ultimately expanding their functional scope and enabling advanced functionalities like personalized services and intelligent automation. Thus, there is a balance that must be achieved between local processing and offloading the task to the edge/cloud. Several techniques are employed to facilitate cloud-edge orchestration and offloading for low-power devices. Device Edge Offloading is a key strategy where computational tasks are offloaded from resource-constrained devices to more powerful edge servers. This allows complex machine learning inferences to be remotely executed in proximity to the devices, optimizing latency and network bandwidth, particularly for ENDs. Techniques such as model compression, federated learning, and resource-efficient energy-aware algorithms are crucial for overcoming hardware limitations and enabling Tiny Machine Learning (TinyML) on constrained devices. Energy-aware task scheduling is also vital, optimizing energy consumption by prioritizing essential tasks, dynamically allocating processes, and leveraging techniques like duty cycling to ensure reliable operation despite fluctuating power availability from ambient sources. Split learning and federated learning can be integrated to enhance performance and learning, addressing both personalization and generalization in edge AI environments. Over-the-air TinyML (OTATinyML) [43] is another method that enables remote updates, configuration, and execution of TinyML models, enhancing scalability and flexibility. Despite the numerous advantages, several challenges hinder the full realization of cloud-edge device orchestration and offloading for low-power devices. Foremost among these are the inherent limitations of hardware resources, particularly for ENDs, and the energy availability. Training complex models on such devices remains impractical, necessitating advanced techniques like model compression and federated learning. Updating models dynamically on IoT devices is also a significant challenge, often requiring physical access or complex communication protocols. The intermittent power availability in battery-less environments can lead to disrupted inference and decreased model accuracy, despite the use of techniques like duty cycling and model quantization. Furthermore, deep learning models often demand substantial memory and computational resources, creating trade-offs between inference complexity and device longevity. While edge processing reduces cloud dependencies, maintaining model updates and adaptability can be challenging if connectivity is not consistently stable. Page 10 of 94 D5.1 - Use cases, KPIs and KVIs Chapter 3 Use Cases: State of the Art The purpose of this chapter is to provide an overview of the state of the art on the use cases for A-IoT. As A-IoT continues to gain importance, identifying and analysing its potential applications has become a priority for both research and standardisation communities. This chapter presents the main efforts to define A-IoT use cases from two primary sources: international SDOs and publicly funded research and innovation projects. First, this chapter explores how organizations like 3rd Generation Partnership Project (3GPP) and the Institute of Electrical and Electronics Engineers (IEEE) have contributed to shaping the A-IoT landscape by identifying key application and technical requirements. Second, it reviews relevant use cases proposed in European research projects, highlighting real-world deployments. 3.1 International Standardization Bodies In recent years, standardization of A-IoT has gained momentum with several organizations recognizing its potential. 3GPP has taken a leading role in trying to integrate A-IoT into the 5G ecosystem. In addition to 3GPP, entities such as IEEE have also started working on A-IoT standardisation. 3GPP has recognized A-IoT as an IoT technology with the potential to enable large-scale connectivity with net zero energy consumption. The concept is based on the integration of ENDs into cellular and non-cellular networks, enabling ubiquitous connectivity in a number of applications. In Release 18, 3GPP initiated preliminary studies on A-IoT as part of its efforts to enhance 5G. Release 18 laid the groundwork by exploring the potential applications of energy-harvesting IoT devices. The study item Technical Report (TR) 38.848 “Study on Ambient IoT (Internet of Things) in RAN” [44] addresses the feasibility of meeting design targets for relevant use cases of A-IoT. The study considered various aspects (including device types, communication methods, and network requirements) to support the operation of A-IoT devices in 5G. It also identified several use cases for A-IoT, emphasizing their applicability in scenarios where battery replacement is challenging. More specifically, two main use case groups were proposed. The first group, or Group A, is based on the deployment environment, and consists of three subcategories: Indoor, Outdoor and Indoor/Outdoor. The second group, or group B, is based on the functionality/application of the use case, and it is formed by four subcategories: Inventory, Page 11 of 94 D5.1 - Use cases, KPIs and KVIs Sensing, Positioning and Command. By combining both groups, eight different categories are generated, which were called representative Use Cases (rUC) in TR 38.848 [44]. Table 3.1 defines these eight rUCs. Table 3.1: Use cases defined in TR 38.848 [44]. rUC Description Indoor Inventory Discover what goods are present in a specific area. A-IoT devices attached to these goods can report an identifier and additional info upon request from the network. Outdoor Inventory Indoor Sensing A-IoT devices are associated with a sensor. A-IoT devices send data obtained by the sensor triggered by different conditions (periodically, triggered by the network, when the A-IoT device is completely powered...). Outdoor Sensing Indoor Positioning Determine the location of certain goods. A-IoT devices attached to these goods report information like identification or location. Outdoor Positioning Indoor Command A-IoT device is associated with an actuator. Commands to control the actuator are usually sent by the network. Outdoor Command In Release 19, 3GPP continued to develop the vision introduced in Release 18. While the term “Ambient IoT” is not always explicitly used, the use cases being discussed align closely with its principles: massive-scale deployment of low-cost, low-power devices with intermittent or passive connectivity to mobile networks. These use cases are being explored within several 3GPP working groups, most notably SA1 (Services) and SA2 (System Architecture). In TR 22.840 "Study on Ambient power-enabled Internet of Things" [45], 3GPP presented a feasibility study on the potential support of A-IoT within 3GPP systems. This document represented the first formal effort within 3GPP to analyse the applicability of A-IoT to existing and future mobile network. The objective of TR 22.840 was to evaluate use cases that reflect realistic applications of A-IoT across multiple vertical domains, and to identify the technical requirements and limitations in current 3GPP specifications. The report focused on exploratory analysis to support future normative work. In Technical Specification (TS) 22.369 "Service requirements for ambient power-enabled IoT" [46], which used the outcome from TR 22.840, the functional requirements corresponding to the rUCs in TR 38.848 were defined. These requirements cover aspects such as communication, positioning and location services, device and service management, information collection, network capability exposure, charging models, and security. Moreover, TS 22.369 also provides some general KPI definitions for some of the rUC. In addition to the standardization efforts within 3GPP, the IEEE is also focusing on A-IoT. The IEEE 802.11bp Ambient Power (AMP) Task Group started a technical investigation to support A-IoT devices over wireless local area networks (WLANs). In March 2023, the group published a technical report titled “Support of Ambient IoT Devices in WLAN” (IEEE 802.11-23/0436r0) [47], which explores the requirements, challenges, and possible enhancements to the IEEE 802.11 standard to include A-IoT. The report identifies use cases that show how A-IoT devices can be integrated into WLAN environments. Each use case imposes different requirements in terms of energy efficiency, data collection frequency, network access procedures, and device identification mechanisms. Table 3.2 shows a summary of these use cases. Page 12 of 94 D5.1 - Use cases, KPIs and KVIs Table 3.2: Use cases defined in IEEE 802.11-23/0436r0 “Support of Ambient IoT Devices in WLAN”[47]. Use Case Description Smart Manufacturing Real-time tracking of assets, workers, and products using low-cost and batteryfree devices for dense deployment. Data Center Environmental and asset monitoring through wireless sensors. Shares similar requirements with smart manufacturing: battery-free, long-life, and lowmaintenance. Logistics/Warehouse Accurate inventory checks and sorting using battery-free tags. Needs high identification accuracy, small size, and support for moving items. Smart Home Compact low-power sensors for safety, comfort, and locating personal items within home environments. Smart Agriculture Monitoring environmental conditions and assets in large outdoor areas. Requires wide coverage, battery-free operation, and the ability to handle thousands of devices. Indoor Positioning Navigation and tracking in large indoor spaces. Needs high-density, low-cost tags with 1–3 m horizontal and 1–2 m vertical accuracy. Smart Power Grid Monitoring equipment and environmental conditions in substations and transmission lines with battery-free, low-maintenance sensors with long-range coverage. Fresh Food Supply Chain Tracking of food transport conditions using ultra-low-cost, battery-free devices with long intervals between data transmissions. 3.2 European Research and Innovation Projects Beyond standardization activities in 3GPP and IEEE, several European research and innovation projects have contributed to the definition of use cases aligned with A-IoT. Although not all of these projects explicitly use the term "Ambient IoT", many of them address its key enablers. Table 3.3 provides an overview of some of the European projects that present use cases with strong compatibility with A-IoT. The ADROIT6G project focuses on enabling distributed intelligence for 6G AI and cloud-native networks. Some of its use cases defined in [48], [49] can also be applied to A-IoT: •Assisting First Responders: ultra-low-power wearables can be used to transmit medical and environmental data without battery replacement, •Rail Automation supported by Non-Terrestrial Networks: A-IoT nodes can harvest ambient energy and send reports via satellite links, minimizing maintenance costs, •Collaborative Construction Robots: self-powered sensors can continuously report status data to nearby robots in the construction environment, giving robots the necessary context to coordinate movements, avoid obstacles, and adjust tasks as site conditions change. The Hexa-X project, which is a flagship project for 6G vision, defines five main use case families: Sustainable Development, Immersive Telepresence for Enhanced Interactions, Local Trust Zones for Human and Machine, Massive Twinning, and Robots and Cobots [50]. A-IoT can complement four of them: Page 13 of 94 D5.1 - Use cases, KPIs and KVIs •Sustainable Development: self-powered sensors and actuators can be deployed in the 6G ecosystem, enhancing energy efficiency and reducing maintenance costs, •Local Trust Zones for Human and Machine: ENDs can provide short-range data collection and distributed intelligence at the edge, thus enabling low-latency decision making, •Massive Twinning: energy-harvesting sensor arrays can enable digital-twin platforms with continuous status updates on infrastructure or industrial assets, ensuring that virtual replicas remain up-to-date without power-hungry or complex logistics, •Robots and Cobots: ENDs can be embedded in workspaces to guide and coordinate and command collaborative robots. In the Hexa-X-II project, which continues the work performed in Hexa-X, many of the use case families mirror those of Hexa-X, and four—Collaborative Robots, Physical Awareness, Digital Twins, and Trusted Environments—remain compatible with A-IoT [51]. The IntellIoT project develops integrated, distributed, human-centred and trustworthy IoT frameworks applicable to a number of scenarios [52], [53]. Three use cases in which A-IoT can be potentially applied are proposed in this project: •Agriculture: A-IoT devices can harvest energy from sunlight, thermal gradients, or RF sources to provide environmental data, reducing maintenance costs and supporting sustainable farming, •Healthcare: A-IoT wearables can be used to monitor medical data without frequent battery replacements, enabling long-term remote care and minimizing electronic waste. •Manufacturing: A-IoT devices embedded on equipment can eliminate battery maintenance of sensors that monitor machine performance, enabling predictive maintenance and process optimization while integrating into existing Industrial IoT networks. In the REINDEER project, four application scenarios are introduced: Adaptative Robotized Logistics, Human-Machine Interaction in Care Environments, Immersive Entertainment for Crowds, and Smart Homes [54]. A-IoT can be applied to three of its four use cases. •Adaptive Robotized Logistics: END can be used in warehouse infrastructure and autonomous vehicles, enabling continuous tracking of goods and route optimization without battery replacement, •Human-Machine Interaction in Care Environments: sensors powered by body heat or ambient light can monitor medical and environmental conditions, supporting assisted living without frequent maintenance, •Smart Homes: A-IoT sensors can be used to monitor or modify the environmental conditions of the house, as well as to monitor its structural conditions to enable predictive maintenance. In the SUPERIOT project, three main application scenarios are analysed: Smart Tags and Labels, Large-scale Sensing and Actuation, and Enhanced IoT Communications in Demanding Environments [55]. Each of these three application scenarios are divided into more concise use cases. A-IoT technology can be applied to the three main application scenarios of SUPERIOT as follows: Page 14 of 94 D5.1 - Use cases, KPIs and KVIs •Smart Tags and Labels: A-IoT tags can be attached to market products, healthcare equipment in hospitals and medical centres, industrial equipment in factories and production lines, or even pets for tracking and monitoring. This enables sustainable and low-cost tracking and location in multiple scenarios. •Large-scale Sensing and Actuation: ENDs can be used to monitor and actuate over large like smart buildings and facilities, construction environments, smart cities and rural locations. This is possible thanks to the low maintenance requirements and long autonomy of ENDs. •Enhanced IoT Communications in Demanding Environments: A-IoT is difficult to implement in this application scenario due to the particularly high level of security and privacy required. However, ENDs can be used as sensors in on-body/in-body applications or in remote zones to deploy efficient, low-maintenance and long-lasting networks. Table 3.3: European research and innovation projects. Project Name Focus Relevance to A-IoT.Reference ADROIT6G AI and cloud-native design for 6G. Involves intelligent, context-aware connectivity for massive low-cost devices. [48], [49] Hexa-X Flagship for 6G vision and architecture. Introduces concepts like ambient connectivity, massive IoT, and 6G location services. [50] Hexa-X-II Continuation of Hexa-X with system-level pilots. Considers extreme energy efficiency and ultra-dense IoT devices deployment in 6G. [51] IntellIoT Integrated, distributed, humancentred and trustworthy IoT framework in healthcare, farming and industry. Energy-aware IoT network management, application function allocation, edge computing, AI/ML... [52], [53] REINDEER Cell-free, distributed beamforming and distributed intelligent processing. Enables energy-efficient, programmable wireless environments for ubiquitous sensing and tracking, as well as interaction with ENDs. [54] SUPERIOT Sustainable, felixble and adaptable IoT system based on optical and radio communications. Develops IoT nodes based on energy harvesting to enhance energy efficiency and environmental sustainability. [55] Page 15 of 94 D5.1 - Use cases, KPIs and KVIs 4.2 Electronic Shelf Label Table 4.3: Use Case Families for Electronic Shelf Label use case. Use Case Families Hard-to-reach devices Massive-scale sensor networks Extreme-lifetime applications Table 4.4: END classes for Electronic Shelf Label use case. END Classes UP ENDs DB ENDs DP ENDs CO ENDs 4.2.1 Description The use of electronic shelf label (ESL) tags is becoming more widespread on retail shelves that house large numbers of products. The ESL tag provides information about the price of a product, along with other potential parameters, such as the expiration date, sales offers, bar codes, quick response (QR) codes, etc. ESL tags are usually in front of a product or a group of products on the shelf, providing a convenient way to manage or update product information or set dynamic pricing over-the-air (OTA), using a mobile handheld reader (connected to a radio unit e.g., Macro basestation), or static on-site (Micro basestation), or remote reader (connected via an intermediate node) accompanied by a user-friendly application interface. This approach saves considerable time and labour costs compared to commonly used static labels (e.g., paper labels), while also enhancing the shopping experience [45]. A typical ESL tag is composed of several components that are required for communication and information display, including the RF module, MCU, battery, and display unit. To manage the information on the ESL tag, communication occurs via a wireless link between the tag and the reader connected to a server application. Some examples of wireless technologies include RF, infrared, bluetooth [56], and visible light [57], where each technology offers a different level of reliability and range. Among the different available technologies, radio frequency identification (RFID)-based communication offers a significant advantage of low infrastructure costs. However, this comes at the expense of poorer quality-of-service (QoS) compared to other communication technologies, which have extensive protocol stacks, dedicated spectra, and centralized control to support OTA communication. 4.2.2 Functional Requirements Communication ESL tags must support unique identification within a local service area. The tag identifier can include third-party components in addition to the device unique identifier (UID) assigned by the core network (CN). The third party can ensure the uniqueness of tag UIDs within their Page 22 of 94 D5.1 - Use cases, KPIs and KVIs Item I1 Item I2 CN/ESL network ESL tags Intermediate node Mobile handheld reader Remote reader Radio unit On-site reader Figure 4.1: Illustration of Electronic Shelf Label use case. local service area, such as warehouses, shopping malls, grocery stores, etc, and thus relax the constraints on ensuring uniqueness on a global scale. The abundance of tags within a small service area mandates that communication protocols should include features for robust resource selection, such as efficient random access and effective collision resolution techniques to support bulk operations, e.g., bulk write operations. Additionally, the wireless channels experience temporary shadow fades due to obstacles, i.e., moving carts, humans, robots near shelves, aisles, etc. Therefore, it is essential to support multi-reader operation to circumvent temporary coverage holes and ensure high availability. Positioning/Location Given that in this use case all ESL-based ENDs have static locations, updates to their spatial positioning during runtime are not required. A third party ensures the mapping of ESL tags and their spatial context in a database, which is maintained as part of the application’s functionality. Management Regarding management, the system must be equipped with features and mechanisms to ensure: •permanent disabling, decommissioning, and/or off-boarding of ENDs, •temporary enabling and disabling, •activation-on-demand for newly deployed ENDs, •software/firmware updates via centralized, automated provisioning and configuration, •group operation, •security management (at the CN and application levels, at a minimum), •fault detection, Page 23 of 94 D5.1 - Use cases, KPIs and KVIs •multi-reader operation, and •ENDs ownership management, particularly if the device UID includes third-party components. Network Security The network must be equipped with robust security protocols designed for low-power IoT systems to ensure the integrity and confidentiality of the information handled by ENDs. The system should support strong authentication procedures for ESL by network nodes, as well as additional authentication procedures for network nodes at the tag/device side, if the ENDs’ capabilities allow. Energy The ESL-based ENDs would typically be placed in indoor environments, or at the very least, in areas where natural ambient energy sources are limited. Therefore, to operate effectively, ENDs must support energy harvesting from RF, visible light, or both. To ensure reliable operation and longer transmission duration, ENDs must incorporate a small energy storage unit (e.g., capacitor, micro-battery, or supercapacitor) to maintain functionality during temporary unavailability of ambient energy sources. For ESL tags intended for outdoor use, the storage unit must be robust enough to withstand corrosive environmental conditions (such as moisture, unmanned vehicle (UV) radiation, dust, etc.) over extended periods. Especially in applications involving dynamic pricing, the ENDs should support features like rapid charging under low input power conditions (ranging from µW to mW) to facilitate relatively strenuous transmission operations. User-Safety and Robustness The use case is predominantly an indoor scenario, and therefore, variations in humidity and temperature should not be drastic. Consequently, the ENDs can be designed to operate under more relaxed conditions, with a relatively narrower range for temperature and humidity. On the other hand, the use case requires frequent handling of products by customers (e.g., picking up, touching, etc.), which may expose the ENDs to frequent shocks and vibrations. From this perspective, the ENDs should be designed with shock and vibration-absorbing materials or be robust enough to withstand such conditions. Additionally, they should be compact enough to maintain user comfort and object functionality, but not so compact that they cause difficulty for customers when reading information from the ESL-equipped liquid crystal display (LCD). The most important consideration is that END tags should be manufactured using non-toxic materials. AI/ML For extremely passive ENDs, employing AI/ML techniques on the devices would be challenging due to uncertainty in energy availability, limited energy storage, and low processing capabilities. However, for relatively capable ENDs, such techniques can be deployed to understand customer engagement, behavioural choices, and product demand, especially if the ESL tag is supplemented Page 24 of 94 D5.1 - Use cases, KPIs and KVIs with additional hardware, sensors and infrastructure, such as energy systems to support large processing cost, or sensor like infrared sensor to detect customer proximity or interest. Alternatively, the AI/ML techniques can be deployed on the reader side (base station, or intermediate user equipment (UE)), or in the cloud. These techniques can enhance customer engagement analysis, helping to boost business output. Additionally, they can assist with channel and radio conditions estimation and error corrections, thus improving the communication link. 4.2.3 Key Performance Indicators Table 4.5: KPIs for Electronic Shelf Label use case. Device KPIs Network KPIs KPI Value KPI Value Sensing accuracy N.A. End-to-end latency <1 s Device autonomy >20 years Age of information N.A. Avg. power consumption <10 µW Packet loss rate <1 % Peak power consumption <100 µW Message size <100 Bytes Communication range <50 m Service area dimension <20000 m2 Max simultaneous devices <30000 devices Transfer interval 0.3 - 2 hours Device speed Static User-experienced data rate 0.8 kbps Application KPIs AI/ML KPIs KPI Value KPI Value Communication service availability > 99.9 % Training complexity N.A. Positioning/Location service availability Not required Inferencing accuracy N.A. Positioning/Location service accuracy N.A. Inferencing latency N.A. Command service availability Not required AI/ML capabilities Not required 4.2.4 Key Value Indicators Environmental KVIs Impact Score: 3 Traditional labelling methods typically rely on RFID-based technology or manual labelling, both of which have environmental impacts. RFID integration requires precise implementation, and often includes batteries which need to be replaced periodically. This increases costs and requires additional human effort. Moreover, the communication links are not always optimal, leading to higher energy consumption (due to increased collisions or retransmissions), as well as the need for denser deployment of readers, which increases infrastructure costs, material costs, and puts additional strain on the resources. Page 25 of 94 D5.1 - Use cases, KPIs and KVIs At the same time, ESLs contribute to sustainability by reducing paper waste (typically used in paper tags), as well as minimizing food and product waste through better management, tracking, and updates. On the other hand, manual labelling requires a significant human presence, leading to delays and being highly prone to errors. This results in waste of human effort, which could be minimized through automation with better-equipped technology, such as ESL. Social KVIs Impact Score: 3 ESLs have a modest social impact by enhancing both the customer and shopping experience and store efficiency. Their use in businesses increases transparency by providing regular, fast, and accurate pricing updates. They also save time by reducing human errors and infrequent customer queries associated with paper tags and can ensure faster checkouts if ESLs integrated with the point-of-sale (POS) system. Furthermore, with dynamic pricing strategies, ESLs can help accelerate the sale of products with near-expiry dates through targeted discounts and sales, thus reducing wastage. Economic KVIs Impact Score: 4 ESLs provides a significant economic impact for retailers, store owners, or warehouse managers by reducing costs and improving revenue. Although there may be substantial initial costs for infrastructure deployment, these can be outweighed by the gains, especially with large-scale utilization in supermarkets or hypermarkets, mega marts, malls, and similar venues. ESLs offer a highly automated solution that reduces pricing errors and incorrect pricing, significantly minimizing financial losses. They also lower costs associated with labour for manual label changes or order fulfilment when integrated with the POS system, as well as material costs for paper, ink, and printing used in traditional paper tags. Furthermore, ESLs ensures efficient inventory management by enabling real-time tracking and reducing waste. Additionally, they facilitate dynamic pricing, which can potentially increase revenue through targeted promotions and price optimization. Innovation KVIs Impact Score: 4 ESL tags powered by ambient sources running on cellular technologies can offer a wide range of diverse and rich features. Some of these features can be remotely managed if not directly incorporated into the ESL tags due to their small form factor. This allows businesses to optimize operations, reduce costs, and minimize waste. The data collected from tags can be combined with other sensors, such as temperature or infrared sensors, depending on the use case, to enable improved data analysis and prediction. For example, a temperature sensor could allow ESL tags to dynamically adjust expiration times/dates or modify the prices of temperature-sensitive food items. In summary, the innovation will be driven by the following indicators, which will propel the adoption of ambient-powered ESL tags in the retail world: •battery-less (reduces cost, easy adoption and offers more flexibility), •utilization cellular technologies’ features, centralized control, reliable communication link, rich data collection, analysis, and prediction features. Page 26 of 94 D5.1 - Use cases, KPIs and KVIs Environmental Social Economic Innovation 12345 Figure 4.2: Impact Scores for Electronic Shelf Label. Page 27 of 94 D5.1 - Use cases, KPIs and KVIs 4.3 Sensors in Smart Homes Table 4.6: Use Case Families for Sensors in Smart Homes use case. Use Case Families Hard-to-reach devices Massive-scale sensor networks Extreme-lifetime applications Table 4.7: END classes for Sensors in Smart Homes use case. END Classes UP ENDs DB ENDs DP ENDs CO ENDs 4.3.1 Description In the last decade, smart homes have seen an increase in popularity due to their improved comfort and quality of life, and the way they can transform our daily lives. In such smart homes, everyday devices and appliances (such as thermostats, lights, refrigerators, ovens, etc.) get connected to a central monitoring and control unit, often connected to the Internet for remote access. This reshapes the way we engage with our home environment, potentially enhancing energy efficiency, sustainability, home security and health outcomes. However, getting everyday objects smart by connecting them to the internet makes them vulnerable to hacking (security and privacy), has a certain level of complexity, and costs money. This use case addresses another hidden challenge inside smart homes. Smart objects require extra hardware for (i) communicating with the central controller in case of wireless connection, and for (ii) processing the incoming or outgoing information to or from the device, increasing the power consumption at each unit. Research shows that in smart lighting [58], [59] the standby power consumption of the electronics for smart Light Emitting Diode (LED) bulbs’ control and radio communication lies around 0.4 W, resulting in an additional 3.5 kWh per year. The active power of a LED bulb lies around 10 W for a 800 Lm lamp. This means that every 20 hours, a smart lamp uses as much as a generic LED lamp in active mode during 1 hour. RF backscatter communication in indoor environments in combination with ultra-low power designs can counteract this unseen energy consumption. In this use case, we explore how we can set up an indoor backscattering network for monitoring of ambient parameters (light, temperature, air quality). These devices operate without an integrated power supply by leveraging backscatter communication, wherein they modulate and reflect incident RF signals emitted by an external RF source to transmit data. The information that will be added to the channel consists of small data packets with a low update rate. An example of the smart home with backscattering integration is depicted in Figure 4.3. It consists of several ENDs, capturing, processing and transmitting the received sensor data and a central controller that receives, demodulates and processes the backscattered sensor data. Page 28 of 94 D5.1 - Use cases, KPIs and KVIs Figure 4.3: Illustration of Sensors in Smart Homes use case. Typical examples are temperature, humidity, air quality, water and power consumption and presence detection sensors. 4.3.2 Functional Requirements Communication The system shall enable ultra-low-power, low-throughput communication between semi-passive backscatter ENDs and UEs by leveraging a custom backscatter modulation scheme. The protocol must support sufficient data rates to facilitate the periodic transmission of sensor readings to a central controller, while maintaining minimal active power draw on the tag side. Communication should be optimized for short-range, asymmetric links, with emphasis on tag simplicity and reader-side processing complexity. The custom backscatter communication protocol shall incorporate lightweight but effective synchronization primitives and error detection mechanisms, such as preamble-based timing alignment and cyclic redundancy check (CRC), to ensure data integrity. To account for the challenges inherent to reflective and interference-prone RF environments, the protocol should support adaptive symbol timing, collision avoidance through time or frequency domain separation, frequency hopping, code division-multiple access (CDMA), to enable scalable, collision-resistant operation and robust envelope detection strategies. Additional redundancy or forward error correction (FEC) may be integrated as needed based on the application’s reliability requirements. The communication stack should be designed for seamless integration with commercial off-theshelf (COTS) mobile and IoT devices by employing standard-compliant physical and MAC layer abstractions where applicable. While leveraging custom backscatter modulation techniques, the protocol should maintain compatibility with widely adopted wireless frameworks (e.g., BLE,WiFi, or sub-GHz industrial, scientific and medical (ISM) bands) through proxy gateways or protocol translation layers. This ensures scalable deployment and interoperability within heterogeneous IoT ecosystems, while facilitating straightforward user access via smartphones or other existing infrastructure. The system must support concurrent interactions from multiple backscatter tags or user devices within a shared RF domain without incurring packet collisions or identity ambiguities. Tag identification must be robust against signal interference and closely spaced transmissions, ensuring Page 29 of 94 D5.1 - Use cases, KPIs and KVIs accurate differentiation and tracking of multiple active entities in dynamic environments. Positioning/Location The system architecture shall not actively determine or update the spatial positioning of energyneutral devices during runtime. Instead, device commissioning, including the assignment of spatial context, shall be conducted manually during initial system setup. To minimize complexity and conserve energy, spatial granularity should be limited to the room or zone level rather than requiring precise localization. This approach aligns with the constraints of energy-harvesting devices and supports scalable deployment in infrastructure-limited environments. Management Energy neutral devices shall embed a UID within their transmitted payloads, ensuring traceable association between sensor data or device state and the originating hardware instance. This UID is critical for the aforementioned system commissioning, enabling deterministic mapping of data streams to specific devices during deployment and operation. While spatial metadata, such as room-level location, should be defined during initial setup, the system shall support controlled updates to this metadata to accommodate reconfiguration or relocation. The system should support configurable mechanisms to temporarily or permanently disable noncritical ambient IoT sensors, thereby optimizing energy usage and minimizing unnecessary RF activity. This functionality should be controllable via centralized policies or local triggers, allowing for dynamic adaptation to environmental conditions, user preferences, or operational priorities. To ensure operational integrity, devices must undergo periodic health checks to verify active status, functional correctness, and communication reliability. Fault conditions (e.g., transmission failures, sensor anomalies, or energy depletion) shall be logged and flagged for maintenance. Furthermore, performance metrics, including signal strength, data latency, and activity frequency, should be continuously monitored and analysed to inform diagnostics, support predictive maintenance, and ensure long-term system robustness. Collected Information and Network Exposure Devices shall transmit user-associated telemetry, such as environmental readings (e.g., temperature) or device state information, to the central controller for purposes including system analytics, behavioural inference, or personalization. While the transmitted data may appear non-sensitive in isolation, appropriate safeguards must be implemented to preserve user privacy. All data must be anonymized or pseudonymized in accordance with established privacy standards, ensuring that individuals cannot be directly identified through the dataset without authorized correlation. Back-end systems must provide mechanisms for secure data access, with full support for userinitiated data erasure and auditing of stored records. Furthermore, the system architecture shall incorporate robust cybersecurity measures, including authentication, encryption, access control, and intrusion detection, to safeguard against unauthorized access, data breaches, and malicious tampering. Page 30 of 94 D5.1 - Use cases, KPIs and KVIs Energy This use case depends on device classes requiring limited data buffering or short-duration computational tasks (e.g., Class 3 devices). The integration of small-scale energy storage elements, such as supercapacitors or high-efficiency capacitors may be employed. These storage components enable transient high-power operations, such as time-critical response coordination or data framing, while preserving the system’s overall ultra-low power footprint. In short-term deployments, small battery-assisted tags may be provisionally utilized. However, this strategy is generally discouraged due to environmental concerns associated with battery disposal and the regulatory implications of hazardous material classification. In some scenarios, ambient light harvesting can enable long-term, batteryless operation. User-Safety and Robustness RF emissions from the system must adhere to all applicable regional regulatory standards (e.g., Federal Communications Commission (FCC),European Telecommunications Standards Institute (ETSI)), ensuring compliance with spectral power density limits and interference thresholds. Devices shall be designed to operate in a non-intrusive manner, exhibiting minimal electromagnetic or physical impact on their immediate environment. In the event of system failure, communication loss, or degraded performance, critical fallback functionalities, such as manual control interfaces (e.g., light switch activation), must remain accessible to the user. Furthermore, the system shall support autonomous fault recovery mechanisms, enabling self-reconfiguration or re-synchronisation without user intervention to restore nominal operation. This is possible with low-power watchdog timers, monitoring firmware execution and trigger system resets or safe state fallback if abnormal behaviour is detected. Periodic diagnostics or health beacons (where energy budget permits) can provide minimal status updates to indicate connectivity, sensor state, or charge status of local energy storage. AI/ML Advanced signal processing at the central processing unit (access point (AP)) side, such as adaptive filtering and interference suppression, can enhance tag detection in challenging RF backscatter environments, particularly in multipath-rich or spectrally congested conditions. While such techniques are valuable, many smart home applications currently rely on simpler processing approaches, which will be benchmarked against AI/ML-based methods. A-IoT models can be leveraged at the system level to analyse user interaction patterns, environmental sensor data, and actuation behaviour. These models enable energy optimization, climate control adaptation, and support automated device commissioning, reducing manual setup efforts. Additionally, ML models play a key role in predictive maintenance by detecting anomalies such as irregular RSS or delayed responses. A notable application is the early detection of faults in household battery systems or power electronics, helping prevent hazards such as overheating or fire, while maintaining a lightweight footprint on individual devices. However, in many practical smart home use cases, such complexity may not be necessary, basic signal processing techniques are often sufficient to achieve reliable performance, particularly in static, low-interference environments. As part of the system evaluation, AI/ML models should Page 31 of 94 D5.1 - Use cases, KPIs and KVIs the bridge or embedded into the structure, energy harvesting from ambient sources might not be viable as a stand-alone solution. This is why, the gateways can act the times of energy transmitters charging the ENDs using RF waves. Notably, the ENDs deployment and their energy requirements will affect the optimal deployment of such gateways. User-Safety and Robustness ENDs must operate under harsh conditions, including large temperature variations, high humidity, corrosion, and mechanical vibrations. Water/dust-proof enclosures, e.g., IP-graded, will be required to maintain the ENDs operational. In addition, physical-security measures including tamper-resistant housings, lockable mounts, and low-complexity intrusion detection mechanisms must be in place to prevent unauthorized access or vandalisms in ENDs within the reach of citizens. Finally, to ensure rapid recovery from malfunctions, ENDs should support manual or autonomous recovery mechanisms depending on the installation conditions. Manual options include hot-swappable modules for field technicians, while autonomous recovery can encompass built-in watchdogs, self-rebooting firmware, and self-healing software routines. AI/ML The amount of measurement data generated by the ENDs on the bridge can grow significantly over time. Notably, AI/ML-aided techniques can help reduce the volume of generated data by dynamically adjusting the thresholds that trigger sensing operations. Additionally, AI/ML methods can assist engineers responsible for bridge maintenance by detecting anomalies in the collected data, thereby helping to predict potential risks, schedule timely maintenance operations, and implement traffic restrictions or load management measures when necessary. AI/ML techniques can also be used for communication tasks such as filtering, interference suppression, joint channel estimation and signal detection, and well as for boosting energy efficiency, e.g., via AI/ML-based duty-cycling. Notably, the application of AI/ML in this context must be optimized to operate within the limited processing power, memory, and energy constraints of ENDs, with computationally intensive workloads delegated to the gateways. Page 38 of 94 D5.1 - Use cases, KPIs and KVIs 4.4.3 Key Performance Indicators Table 4.11: KPIs for Bridge Health Monitoring use case. Device KPIs Network KPIs KPI Value KPI Value Sensing accuracy Depends on magnitude End-to-end latency < 1 s Device autonomy > 20 years Age of information < 10 s Avg. power consumption < 1 mW Packet loss rate <1 % Peak power consumption < 10 mW Message size <1 kbit Communication range < 100 m Service area dimension Bridge size Max simultaneous devices < 60 devices Transfer interval < 10 s Device speed Static User-experienced data rate 100 kbps Application KPIs AI/ML KPIs KPI Value KPI Value Communication service availability > 99.9 % Training complexity 200 MFLOPS Positioning/Location service availability Not required Inferencing accuracy > 95 % Positioning/Location service accuracy N.A. Inferencing latency < 0.1 s Command service availability Not required AI/ML capabilities Yes 4.4.4 Key Value Indicators Environmental KVIs Impact Score: 3 While indirect, the environmental benefits of smart bridge health monitoring are still notable. By reducing the number and frequency of inspection trips and heavy equipment use, the system contributes to lower emissions from logistics operations. It also supports resource efficiency, as fewer reactive maintenance activities mean reduced consumption of materials and energy. Importantly, the system promotes long-term sustainability by preserving the structural integrity of the bridge, reducing the need for major renovations or replacements and thus lowering the environmental footprint over the bridge’s life cycle. Energy-saving and energy-harvesting technologies minimize (and even eliminate) the dependence on batteries. This not only prolongs the lifespan of the monitoring system but also reduces the environmental pollution associated with improper waste handling. Social KVIs Impact Score: 4 A smart bridge health monitoring system has a high social impact, primarily by enhancing public safety. Real-time monitoring enables early detection of structural issues, significantly reducing the risk of catastrophic failure and protecting human lives. Additionally, the presence of such a system fosters community trust in infrastructure, as users will feel safer using Page 39 of 94 D5.1 - Use cases, KPIs and KVIs bridges that are continuously monitored. The system also contributes to greater accessibility and reduced disruption by minimizing the frequency and duration of closures needed for manual inspections, thus ensuring smoother traffic flow and access to essential services. Economic KVIs Impact Score: 5 Economically, while upfront costs are higher than for unmonitored bridges due to the ENDs, gateways, and the infrastructure to support the A-IoTs network in general, the application offers substantial value by minimizing the bridge’s operational costs. Through predictive maintenance, it enables timely repairs that are less costly than emergency interventions, while also extending the lifespan of the structure. The system enhances operational efficiency by reducing the need for labour-intensive inspections and allowing maintenance crews to focus on targeted areas. Furthermore, by detecting and addressing vulnerabilities early, the system helps mitigate financial risk related to liability from potential structural failures or service disruptions. Minimizing maintenance operations on the ENDs has a significant economic impact in smart bridge health monitoring system. Considering the a large number of ENDs might be deployed on the bridge, frequent maintenance becomes costly, time-consuming, or may require traffic disruptions. Beyond direct maintenance savings, the value of smart bridge health monitoring lies in ensuring continuous service, public safety, and logistical reliability, all critical to regional economic productivity. Innovation KVIs Impact Score: 2 Incorporating context-awareness to energy harvesting and wake-up technologies brings significant benefits to scale bridge health monitoring operations. Notably, the innovative use of A-IoT systems enables the implementation of low-complexity sensors by exploiting ambient energy sources not only to charge the ENDs, but also to infer physical magnitudes directly from the characteristics of the harvested energy. Similarly, energy harvesting circuits can serve as event-driven wake-up triggers, allowing the ENDs to remain inactive until the physical magnitude to be measured changes beyond predefined thresholds. Environmental Social Economic Innovation 12345 Figure 4.6: Impact Scores for Smart Bridge Health Monitoring. Page 40 of 94 D5.1 - Use cases, KPIs and KVIs 4.5 Personal Belongings Finding Table 4.12: Use Case Families for Personal Belongings Finding use case. Use Case Families Hard-to-reach devices Massive-scale sensor networks Extreme-lifetime applications Table 4.13: END classes for Personal Belongings Finding use case. END Classes UP ENDs DB ENDs DP ENDs CO ENDs 4.5.1 Description In present days, people have with them and use all kinds of personal belongings in their daily lives. These personal items range from electronic devices like headphones or smartwatches, to less sophisticated items like keys, ID cards or wallets. In most cases, these personal items have a high economical or personal value, which means that losing them could represent a problem. However, the loss of personal belongings in the domestic environment affects people of all ages in all contexts on a regular basis. Searching for these lost items can result in time loss, interruptions in daily activities, and high levels of stress and anxiety. This affects productivity and emotional well-being, and can also impact social and professional life. The A-IoT technology can provide a sustainable solution to this problem. This use case uses A-IoT technology to enhance the everyday experience of locating misplaced personal items within a domestic environment. The system integrates END tags, which are small, low-cost, and energy-efficient devices that can be designed to be attached to personal items. These END tags enable interaction with the surrounding infrastructure without requiring constant user intervention. When needed, the END tag gets energy from a RF signal provided by the UE of the personal item’s owner (e.g. a smartphone). Additionally, other ambient energy sources can be used in order to ensure robust operation in multiple scenarios. When a personal item is lost, the owner initiates the finding procedure with their UE. The UE sends a RF signal to energize the END tag attached to the lost item and starts the positioning procedure. When energized, the END tag modulates the received RF signal relying on backscattering modulation, thus being able to respond to the UE’s positioning request. Though END tags do not have active localization capabilities like Global Positioning System (GPS) due to their simplicity and power constraints, their location can be inferred through passive techniques such as signal triangulation, proximity estimation, or RF fingerprinting. By reflecting RF signals to multiple receivers like smartphones Page 41 of 94 D5.1 - Use cases, KPIs and KVIs Figure 4.7: Illustration of Personal Belongings Finding use case. or other IoT nodes, the system can estimate the END tag’s position based on signal strength, ToF or AOA data. This enables users to detect the presence of a lost item and to receive approximate location guidance, enhancing the effectiveness of the search process. With respect to other low power technologies (Ultra-Wideband (UWB),BLE,RFID...), A-IoT uses simpler and less expensive devices. Furthermore, the fact that A-IoT does not require batteries to work also represents an advantage in terms of both complexity and environmental concerns. Figure 4.7 depicts a possible scenario of the Personal Belongings Finding use case. 4.5.2 Functional Requirements Communication The system must support ultra-low power communications while ensuring sufficient data rates and packet sizes to enable the transmission of identifiers from END tags to UEs and IoT nodes. To guarantee reliable operation even in challenging environments, END tags located in hardto-reach areas or regions with limited coverage must be capable of establishing communication with both UEs and IoT nodes. Robustness must be achieved with simple and efficient error detection and synchronization mechanisms, avoiding any increase in the complexity or power consumption of the END tags. END tags must rely on backscattering modulation to encode information onto external RF carriers, enabling the transmission of data to UEs and IoT nodes. Furthermore, communication should be initiated by UEs, thereby minimizing unnecessary data exchanges and enabling efficient and low-power operation. Positioning/Location The system must operate in all domestic environments, including bedrooms, living rooms, garages, and gardens, where signal conditions and physical layouts may vary significantly. Moreover, it should also extend its functionality outside the home, enabling users to locate items in semi-public or external places if the user has access to nearby network nodes. To ensure usability in diverse contexts, the system must offer sufficient positioning granularity to determine the location of objects in small rooms, where spatial resolution is essential for item finding. Page 42 of 94 D5.1 - Use cases, KPIs and KVIs Management Each END tag must have a UID that allows it to be distinguished from others. The system must support the registration and association of each END tag with a specific physical object. To maintain flexibility, the system must also allow for dynamic metadata updates, enabling users to modify the object associated with a given END tag. Additionally, the system must incorporate mechanisms to periodically monitor the operational status of the tags by evaluating health indicators like backscatter signal quality and response time. Any detected faults should be logged, and users must have access to the fault history to facilitate intervention or replacement of malfunctioning ENDs tags. Collected Information and Network Exposure Since END tags in this use case typically remain unpowered, the amount of information they can collect is limited. Nevertheless, certain data—such as the number of activations or the most recent locations where a tag was detected—can be recorded and transmitted on demand when required by the user. To ensure the integrity and confidentiality of this information as well as privacy of user’s data and location, the system must implement low-power yet robust security protocols, including data access control and encryption mechanisms, to protect communications against potential interceptions or data leaks. Energy The primary powering method for END tags in this use case must be WPT, as lost items may end up in locations where no ambient energy sources are available. Nonetheless, complementary energy harvesting techniques like vibration, thermal gradients, photovoltaic cells, or RF energy harvesting can be considered to support END tags operation. In scenarios where the physical size of the END tag is not a limiting factor and does not interfere with the normal use of the object to which it is attached, the use of supercapacitors or other low-energy storage elements can be considered to enhance END tag functionality. However, due to environmental considerations, such storage components should only be employed when necessary. User-Safety and Robustness END tags must be designed to work under different conditions, including shocks, humidity, vibrations, and wide temperature ranges. This ensures robust operation in different scenarios. At the same time, they must be compact enough not to interfere with the normal use of the personal objects to which they are attached, preserving user comfort and object functionality. Additionally, user safety must be a priority; END tags should be manufactured using non-toxic materials and must incorporate both hardware and software-level safety mechanisms to prevent malfunctions that could potentially lead to user harm. AI/ML Due to their low complexity and limited capabilities, it is difficult to use AI/ML techniques (including those aimed at low-power applications) in END tags. However, AI/ML can be deployed in the UE or in the cloud. These AI/ML functionalities can be used to track typical user trajectories or frequently visited locations, which can help in searching for lost objects. AI/ML Page 43 of 94 D5.1 - Use cases, KPIs and KVIs models can also be deployed in the UE to implement advanced signal processing techniques. AI/ML can help in processes like channel estimation, equalization, decoding or demodulation to reduce error rate and improve communication from END tags to UEs. 4.5.3 Key Performance Indicators Table 4.14: KPIs for Personal Belongings Finding use case. Device KPIs Network KPIs KPI Value KPI Value Sensing accuracy N.A. End-to-end latency < 1 s Device autonomy > 20 years Age of information < 10 s Avg. power consumption < 1 µW Packet loss rate <1 % Peak power consumption < 100 µW Message size <1 kbit Communication range < 10 m Service area dimension < 200 m2 Max simultaneous devices < 10 devices Transfer interval On user request Device speed Static User-experienced data rate 1 kbps Application KPIs AI/ML KPIs KPI Value KPI Value Communication service availability > 99.9 % Training complexity N.A. Positioning/Location service availability > 99.9 % Inferencing accuracy N.A. Positioning/Location service accuracy < 1 m Inferencing latency N.A. Command service availability Not required AI/ML capabilities Not required 4.5.4 Key Value Indicators Environmental KVIs Impact Score: 4 The "Personal Belongings Finding" use case has a high environmental impact. Having a method of locating lost objects reduces the need to manufacture replacement units. This reduction in the replacements means less extraction of raw materials and a consequent reduction in energy consumption associated with the design, manufacturing and transportation stages. Recovering lost items also prevents premature disposal, prolonging the life of the goods and reducing the amount of waste and e-waste. This effect translates into a reduction in the materials destined for landfills, reducing the release of harmful substances that result from the degradation of plastic and electronic components. Moreover, the use of A-IoT for this use case compared to other technologies has advantages from an environmental point of view. END tags have reduced electronic complexity and use low-environmental-impact materials. Their manufacturing phase consumes fewer resources and Page 44 of 94 D5.1 - Use cases, KPIs and KVIs produces lower CO2equivalent emissions, especially in the absence of batteries [60]. During use, A-IoT generates no operational emissions from recharging and maintenance, as no batteries are used in END tags. Finally, at the end of life, the absence of batteries and heavy components in A-IoT facilitates waste management and reduces e-waste generation. Social KVIs Impact Score: 4 Recovering lost objects through remote location reduces the anxiety and frustration associated with lost belongings. By reducing the time spent searching for lost items, the stress caused is mitigated. Moreover, the certainty of being able to locate personal items in case of loss increases the sense of control and reduces perceived vulnerability. This effect contributes to a better social environment in which individuals feel protected against accidental losses. Usage of END tags also contributes to the autonomy of users with limited mobility or cognitive impairment by facilitating access to their belongings without requiring assistance. This increased self-sufficiency improves self-esteem and reduces their dependence. Economic KVIs Impact Score: 4 From the owner’s point of view, recovering lost items directly reduces the impact associated with purchasing replacement units. For example, this savings is particularly significant for high-value items, where avoiding a single replacement can offset the investment in remote location tools. For objects susceptible to duplication (keys, access cards, credit cards...), locating the original eliminates the need to pay for copying services. From the manufacturer’s point of view, by eliminating expensive and power-hungry active components, A-IoT enables volume production of END tags, so the economic cost of each unit becomes very low. Innovation KVIs Impact Score: 3 The "Personal Belongings Finding" use case does not offer significant innovative value because existing devices already perform this function. However, it is the use of A-IoT technology that brings innovative value to this use case. The integration of backscatter communication, WPT and ambient energy harvesting allows END tags to operate without batteries or external power supplies, opening the door to devices with extremely long lifespans and no maintenance. Moreover, the simplicity of END tags allows for their integration in everyday objects (clothing, accessories, keyrings, packaging...) without modifying the aesthetics or ergonomics of the product. This increases adoption and enables new products like disposable textile labels or sensors that can be inserted into single-use packaging. Page 45 of 94 D5.1 - Use cases, KPIs and KVIs Environmental Social Economic Innovation 12345 Figure 4.8: Impact Scores for Personal Belongings Finding. Page 46 of 94 D5.1 - Use cases, KPIs and KVIs 4.6 In-body or Wearable Medical Sensors Table 4.15: Use Case Families for In-body or Wearable Medical Sensors use case. Use Case Families Hard-to-reach devices Massive-scale sensor networks Extreme-lifetime applications Table 4.16: END classes for In-body or Wearable Medical Sensors use case. END Classes UP ENDs DB ENDs DP ENDs CO ENDs 4.6.1 Description In-body and wearable sensors are revolutionizing the way human activities and physiological states are monitored, expanding far beyond traditional medical applications to include wellness, sports performance, elderly care, and lifestyle optimization. These sensors function by detecting and measuring physical (e.g., motion, pressure), biochemical (e.g., glucose, hydration), or environmental (e.g., temperature, humidity) parameters through various sensing elements. Data is then processed locally or transmitted wirelessly for interpretation and action. Wearable sensors are typically worn externally integrated into textiles, wristbands, or skin adhesives while in-body sensors are either implanted or ingested. Advances in flexible electronics, smart materials, miniaturized circuits, and wireless communication have enabled these sensors to become increasingly unobtrusive, durable, and user-friendly. For example, fitness trackers monitor cardiovascular metrics, smart fabrics detect hydration levels, smart diapers alert moisture presence [61], and implantable sensors provide real-time data for multiple medical applications (e.g. glucometers, pulse oximeters, rehabilitation feedback...). An illustrative example involves a hybrid wearable platform combining a sole-integrated Triboelectric Nanogenerator (TENG) with a microneedle-based electrochemical sensor. The TENG, embedded in footwear, serves a dual purpose: harvesting mechanical energy from walking and acting as a gait sensor [62]. This harvested energy directly powers a low-energy microneedle sensor that continuously monitors biomarkers in interstitial fluid, creating a battery-less biochemical sensing system. Additionally, integrating lightweight AI frameworks such as TinyML enables on-device processing of gait data, facilitating real-time activity monitoring [63]. An overview of the entire system is shown in Figure 4.9. Page 47 of 94 D5.1 - Use cases, KPIs and KVIs Figure 4.11: Illustration of Smart Agriculture use case. 4.7.2 Functional Requirements Communication ENDs must establish autonomous uplink connections and perform uplink transmission without external wake-up commands. Supported uplink transmissions can be periodic or event triggered, based on changes in environmental parameters (e.g., moisture threshold exceeded). Furthermore, ENDs must rely on ambient backscattering communication or ultra-low-power active transmission for uplink data transmission. They must support transmission of small payloads (e.g., 50–200 bits) representing sensed metrics like humidity, temperature, or CO2level. ENDs must also be able to interface with nearby reader nodes (e.g., a pico-cell or mobile UE) placed in greenhouses or open fields. Coexistence with thousands of similar devices operating in the same farm or greenhouse, with minimal interference and efficient channel access mechanisms, including error detection/correction, is also a very important requirement. Positioning/Location Systems must provide absolute position of ENDs in outdoor scenarios such as for underground soil sensing in open fields to support spatially targeted actions (e.g., precision irrigation, localized fertilization). They must also support relative positioning between ENDs in proximity for collaborative sensing or distributed decision making (e.g., pest zone detection by neighboring sensors). In indoor agriculture settings (e.g., smart greenhouses), systems must provide roomor rack-level positioning of ENDs, enabling microclimate monitoring and spatial decision support. Devices must support multi-source positioning where feasible to improve accuracy, using combinations of technologies and signals, such as cellular and global navigation satellite system (GNSS) (more applicable to outdoor fields), BLE,Wi-Fi, and received signal strength indicator (RSSI) fingerprinting (more useful to indoor greenhouses), RFID (for close-proximity sensing), and use of local references for mesh triangulation (e.g., other ENDs, static readers, base stations). Management Regarding management, systems must provide mechanisms for ENDs software/firmware updates, health checks to identify faults or malfunctions, and to temporarily or permanently disable ENDs, including functionality for remote deactivation (e.g., at end-of-season or for security breaches). Activation-on-demand must also be supported, allowing newly deployed or dormant ENDs to be initialized remotely by a controller or reader node. Page 54 of 94 D5.1 - Use cases, KPIs and KVIs Collected Information and Network Exposure Systems must support access control policies for different roles (e.g., farmer, agronomist, vendor) regarding read/write permissions to specific END groups or data streams. On the other hand, ENDs must provide real-time or near real-time access to collected environmental data (e.g., temperature, soil moisture) when such data collection is meaningful to the farming task progression, as well as when energy availability and communication conditions permit. ENDs must also expose lightweight metadata (e.g., device ID, location, timestamp, sensor type) along with sensor readings to facilitate automated interpretation and integration into farm management systems. Systems must implement Role-Based Access Control (RBAC) to ensure only authorized entities (e.g., farmer, agronomist, equipment vendor) can access END data, issue commands, or reconfigure behavior. They must also limit unnecessary exposure of metadata (e.g., device type, location, sensing role) to prevent profiling or inference attacks, especially when ENDs are deployed in sensitive agricultural zones or with proprietary crops. Security measures like data encryption at source and decryption at destination shall be used when such metadata is required, e.g., for data preparation, model feeding, training or inference purposes. In addition, systems must support policy-based retention and deletion of agricultural data, ensuring historical data is retained only as needed and deleted upon user request or END revocation. Energy ENDs must support multi-source energy harvesting (e.g., combining solar and RF) to improve energy availability and reliability in diverse agricultural environments. The devices must also include a small energy storage component (e.g., capacitor, micro-battery, or supercapacitor) to enable operation during temporary unavailability of ambient energy sources (e.g., night-time or cloudy conditions). The energy storage system must support safe, efficient, and rapid charging under low input power conditions (µW to mW scale). Furthermore, energy harvesting and storage systems must be robust to agricultural conditions, including moisture, dust, temperature variation, and exposure to direct sunlight or fertilizer chemicals. User-Safety and Robustness ENDs must operate reliably within a broad temperature range (e.g., –20°C to +60°C) and high humidity (e.g., > 90 %), which are typical in open-field agriculture and greenhouses. In addition, internal components must be selected to prevent failure due to condensation or thermal cycling. Furthermore, ENDs must be enclosed in weather-resistant housings to withstand harsh outdoor agricultural conditions such as rain, direct sunlight, wind, humidity, frost, and dust. Systems must also include physical protection against corrosion, especially for ENDs exposed to irrigation water and soil, or to withstand disinfectants in greenhouses after crop harvest. In addition, ENDs must tolerate mechanical shocks and vibrations, including those caused by passing farm machinery, animals, or wind-blown debris, and must be UV-resistant for long-term outdoor exposure and chemically resistant to pesticides, fertilizers, and cleaning agents used in agricultural operations. Materials used in ENDs must be non-toxic and agriculturally safe, ensuring no harmful leachates enter the soil or water supply, especially for crops intended for human consumption. ENDs designed for deployment in direct contact with soil or crops must comply with environmental safety and biodegradability guidelines. Finally, manual or autonomous Page 55 of 94 D5.1 - Use cases, KPIs and KVIs fault recovery mechanisms must be in place in case of END malfunction. AI/ML With further advancements in energy harvesting circuits and AI-powered edge computing (e.g., TinyML), ENDs are expected to evolve into more intelligent agents capable of not just sensing but also responding to environmental changes, which are typical in outdoors agriculture environments, autonomously. AI/ML models can be applied to improve agricultural tasks (crop contamination detection, optimization of water consumption in crops watering, etc.) as well as to improve communication-related tasks (filtering, interference suppression, channel estimation, etc.). AI models must be robust to noisy sensor input and long-term drift, especially in harsh agricultural environments with variable weather, soil, and sensor aging effects. Systems should enable incremental or online learning models, allowing ENDs to adjust to seasonal changes (e.g., adapting light-harvesting prediction during cloudy months) without complete retraining. 4.7.3 Key Performance Indicators Table 4.20: KPIs for Smart Agriculture use case. Device KPIs Network KPIs KPI Value KPI Value Sensing accuracy > 95 % End-to-end latency < 100 ms Device autonomy > 10 years Age of information < 1 min Avg. power consumption < 100 µW Packet loss rate <1 % Peak power consumption < 1 mW Message size < 1000 bit Communication range < 100 m; < 500 m 3 Service area dimension < 70000 m2 Max simultaneous devices < 70000 devices Transfer interval 60 min Device speed Static User-experienced data rate 1 kbps Application KPIs AI/ML KPIs KPI Value KPI Value Communication service availability > 99.9 % Training complexity > 50 MFLOPS Positioning/Location service availability > 95 % Inferencing accuracy > 90 % Positioning/Location service accuracy < 1 m Inferencing latency < 1 s Command service availability > 99 % AI/ML capabilities Yes 3< 100 m in indoor scenarios; < 500 m in outdoor scenarios. Page 56 of 94 D5.1 - Use cases, KPIs and KVIs 4.7.4 Key Value Indicators Environmental KVIs Impact Score: 5 The Smart Agriculture use case contributes significantly to environmental sustainability by enabling precise, low-resource farming practices. ENDs reduce the need for frequent human or machinery presence in the field, minimizing fuel use and soil compaction. These devices promote resource-efficient irrigation, fertilization, and pesticide application by delivering granular, real-time environmental data. This reduces runoff and contamination of nearby ecosystems. The ENDs ambient energy harvesting nature eliminates battery-related waste and maintenance trips. Their minimal material complexity and long operational lifespan reduce lifecycle emissions and support sustainable device deployment in large-scale, remote agricultural environments. Social KVIs Impact Score: 4 Smart agriculture enhances food security and quality by enabling better crop management, yield optimization, and resilience to environmental variability. The use of ENDs enables data-driven farming even in rural and underserved areas where conventional IoT infrastructure may be unavailable or unreliable. Farmers benefit from improved working conditions, reduced manual monitoring, and fewer harmful exposures (e.g., to pesticides). Furthermore, the autonomy and simplicity of ENDs empower small-holder farmers, supporting inclusive technology adoption regardless of digital literacy or capital investment. Over time, these systems contribute to more equitable and sustainable food systems. Economic KVIs Impact Score: 4 END-based smart agriculture systems offer cost savings through reduced input waste (e.g., water, fertilizer), minimized crop loss from stress or disease, and lowered labor demands for monitoring. Their ultra-low-cost and maintenance-free operation enables massivescale deployments even in resource-constrained settings, providing a strong return on investment over time. By enabling early anomaly detection and condition-based actions (e.g., irrigation on demand), they reduce expensive reactive interventions. ENDs also open new economic opportunities in the agri-tech market, including sensor manufacturing, data analytics, and AIpowered farm management platforms. Innovation KVIs Impact Score: 4 Smart agriculture with Ambient IoT introduces a high degree of innovation by combining batteryless sensing, energy harvesting, and AI-powered edge analytics for massive scale and hard-to-reach deployments. Innovations such as energy-aware sensing, TinyML-based local anomaly detection, and adaptive data prioritization enable ENDs not just to sense, but to interpret and act on agriculture-specific conditions autonomously. These advances mark a shift from conventional data collection to intelligent, sustainable, and self-managed agricultural systems. Page 57 of 94 D5.1 - Use cases, KPIs and KVIs Environmental Social Economic Innovation 12345 Figure 4.12: Impact Factors for Smart Agriculture. Page 58 of 94 D5.1 - Use cases, KPIs and KVIs 4.8 Asset, Product, Tool, and Item Tracking Table 4.21: Use Case Families for Asset, Product, Tool, and Item Tracking use case. Use Case Families Hard-to-reach devices Massive-scale sensor networks Extreme-lifetime applications Table 4.22: END classes for Asset, Product, Tool, and Item Tracking use case. END Classes UP ENDs DB ENDs DP ENDs CO ENDs 4.8.1 Description This use case focuses on providing real-time tracking of critical assets, products, tools, and individual items within confined, high-precision operational environments such as hospitals, laboratories, advanced manufacturing facilities, and pharmaceutical clean rooms. Its objective is to enable advanced process monitoring, granular quality control, robust theft prevention, and optimized operational efficiency. Differentiating from generic end-to-end logistics, this tracking provides continuous, precise monitoring with significantly higher update rates and lower latency. It offers robust and accurate positioning, achieving sub-meter level precision even in challenging and dense indoor environments, while handling larger volumes of contextual data per item (for example, environmental conditions, usage and position history, integrity status). These capabilities enable critical applications such as process monitoring and digital twinning for workflow validation and efficiency; quality control and compliance through automated verification of item placement, position history and environmental conditions; theft control and loss prevention via real-time geofencing and anomality detection; and operational optimization by providing insights into asset utilization and resource allocation. In order to allow for tracking of objects, tagging is performed, similarly to current RFID scenarios. Tags have to rely on energy harvesting or ultra-low power consumption (ENDs), which is essential for the pervasive, long-term, and maintenance-free deployment required by this kind of tracking across countless items. Integration approaches include external tagging; packagingintegrated tagging (for smart consumables/kits); design-integrated tagging (for instrument/tool intelligence); and embedded item tagging (for component/product-level digital identity). For all the objectives of this use case, precise position and continuous tracking of ENDs is crucial, requiring advanced signal processing and inference algorithms for map construction and location. A schematic overview of the use case is given in Figure 4.13. 4.8.2 Functional Requirements Page 59 of 94 D5.1 - Use cases, KPIs and KVIs Cloud Datacenter AP Edge Server END END END AP WPT and Backscatter Process Monitoring Theft Protection Service Area Contextual Data END AI/ML AI/ML Quality Control Process Efficiency Increase Figure 4.13: Illustration of Asset, Product, Tool, and Item Tracking use case. Communication Passive backscatter communication is the most desirable form of wireless communications for this use case due to its support in ultra-low power ENDs [33], [66]. Active communication may be possible in future applications if receive-side power budgets are increased from the microwattlevel to the milliwatt level [27]. Nevertheless, active communication can be cost-prohibitive and energy-intensive, particularly in large or complex indoor environments. The two exemplary solutions in [33], [66] address these limitations by employing passive communication through ambient backscatter, with signal detection carried out either at the user equipment or at the network side (e.g., the APs), depending on the deployment scenario. To enable joint backscatter communication and channel estimation (to allow for robust positioning/localization), ENDs must employ ambient backscatter communication or transmit uplink signals towards the infrastructure. For backscatter communication, the infrastructure may comprise separate transmitting and receiving APs to avoid full-duplex operation of APs. The infrastructure must detect weak and potentially heavily interfered backscatter signals from the END, where interference primarily arises from the direct transmission paths between the APs. The infrastructure must ensure high dynamic range at the receiving APs to enable reliable detection of low-power backscattered signals. The infrastructure may transmit wideband excitation signals to allow for wideband channel estimation, which is essential for ToF-based localization. The infrastructure may provide sufficient computational resources at the APs for detecting backscatter signals from the END, cancelling interference, and demodulating backscatter symbols, alongside standard processing tasks required for the legacy signals such as orthogonal frequency-division multiplexing (OFDM). In multi-AP deployments, the infrastructure may maintain synchronization of time, frequency, and phase across APs to support coherent signal processing and enhance positioning accuracy and robustness. The infrastructure and the ENDs must support multiple access protocols for massive deployments of ENDs. Positioning/Location To enable positioning, mapping, or sensing in a broader sense, the wireless infrastructure must be capable of acquiring channel observations, i.e., noisy CSI. Several “channel parameters” can be estimated from noisy CSI and contain END-position information. Page 60 of 94 D5.1 - Use cases, KPIs and KVIs RSS-based positioning. The system in [66] is based on the use of ENDs deployed at known positions throughout the building. Each END tag periodically transmits a unique identifier by modulating and reflecting an incident RF signal, rather than actively generating its own transmissions. Two complementary configurations are considered. In the downlink-based configuration, ENDs modulate ambient cellular pilot signals (e.g., from a fourth-generation (4G) or 5G base station (BS)), and these modulations are detected by the UE through analysis of perturbations in the CSI [67]. In the uplink-based configuration, ENDs reflect the uplink signals emitted by the UE, and modulated backscatter is detected at the BS or a nearby AP [68]. In both configurations, the observed signal variations are used to determine the END with the strongest signal, whose known location serves as a proxy for the UE’s position. Angle-Delay-based positioning. The systems investigated in [69] and [70] belong to the classes of a Radio Stripes and a RadioWeaves infrastructure, respectively. We have shown in [71, Fig. 2] that in distributed MIMO (D-MIMO) infrastructures with distributed APs, angular-domain-based positioning supersedes delay domain-based positioning typically for bandwidths below 100 MHz in strong line-of-sight (LoS) channels. We found that the path overlap in strong multipath channels negatively impacts the performance in such a regime [69, Fig. 4], but large bandwidth can help resolve the multipath components. Based on measured channels, we have shown that simultaneous localization and mapping (SLAM) can be used to both infer the END location and a geometric map of the propagation environment, where centimeter-level positioning accuracy has been demonstrated [70]. Delay-Doppler-based positioning. Doppler-based positioning may supersede delay-based positioning in D-MIMOs infrastructures [72, p. 20]. While the accuracy of Doppler-based velocity estimation is independent of the device motion, the positioning accuracy depends on the device velocity. Nevertheless, in combination with the delay-domain, cost-effective yet accurate wireless positioning, even with a few APs is possible in realistic scenarios with harsh channel conditions such as partial obstructed-line-of-sight (OLoS) and strong multipath propagation [73]. Close to centimeter-level positioning accuracy was demonstrated on the measurements from [72]. Doppler-based or carrier-phase-based positioning [74] may be particularly suitable for localizing ENDs in backscatter communication because these channels inherently avoid the carrier-phase calibration problem. Management The infrastructure must provide mechanisms for the lifecycle management of ENDs. This includes discovery, identification, and logical onboarding of new ENDs into the system. Conversely, it must also support effective mechanisms for disabling and offboarding ENDs when they are no longer required or become inoperable. Furthermore, to ensure consistent and scalable deployment, the systems must facilitate centralized, automated provisioning and configuration of APs. Lastly, the infrastructure needs to offer interfaces for automated firmware updates of APs. These updates are crucial for implementing improved features, such as advanced localization and estimation algorithms, and for addressing critical security vulnerabilities. Collected Information and Network Exposure APs within the network are required to provide interfaces for the exchange of channel estimates or intermediate positioning/localization results. This data exchange can occur either via backhaul Page 61 of 94 D5.1 - Use cases, KPIs and KVIs links or through over-the-air interfaces. While systems may collect these channel estimates or intermediate results, the subsequent processing of this collected information can be performed either at the network edge or on centralized systems. After processing, the systems must provide the positioning/localization results to authorized entities, ensuring data access is controlled and secure. Energy RF WPT technology is the preferred method to power ENDs in this use case, requiring an infrastructure capable of providing efficient WPT as a service on a massive scale. Antenna arrays can provide both the transmission efficiencies required to power ENDs and the regulatory compliance to ensure exposure-safe operation [27]. User-Safety and Robustness Robust operation even in harsh channel conditions such as low-signal-to-noise ratio (SNR) and under high END mobility can, for instance, be accomplished by jointly inferring the END location and a geometric map of its propagation environment, as has been shown in [70]. AI/ML AI and ML are crucial for enhancing the capabilities of this use case. They enable predictive maintenance by analysing usage patterns and sensor data from tools and assets, which helps anticipate potential failures, optimize maintenance schedules, and prevent downtime. These technologies also facilitate anomaly detection, identifying unusual movements, environmental deviations, or departures from standard operating procedures. This triggers immediate alerts for issues like theft attempts, quality excursions, or process errors. Furthermore, AI and ML contribute to workflow optimization by analysing historical tracking data to pinpoint bottlenecks, streamline item flow, and enhance overall operational efficiency in complex environments. Their role extends to enhanced localization and mapping, where neural networks and data-driven approaches are expected to improve upon model-based methods, such as in hybrid factor-graph-based systems, increasing accuracy and robustness in situations where traditional models are intractable or complex. Finally, AI and ML are instrumental in automated inventory and audit, streamlining these processes by autonomously reconciling physical items with their digital twins, thereby reducing manual effort and errors. Page 62 of 94 D5.1 - Use cases, KPIs and KVIs 4.8.3 Key Performance Indicators Table 4.23: KPIs for Asset, Product, Tool, and Item Tracking. Device KPIs Network KPIs KPI Value KPI Value Sensing accuracy > 95 % End-to-end latency < 1 s Device autonomy > 10 years Age of information < 1 s Avg. power consumption < 10 µW; < 1 mW 4Packet loss rate <1 % Peak power consumption < 100 µW; < 5 mW 5Message size < 1 kbit Communication range < 100 m Service area dimension < 10000 m2 Max simultaneous devices < 100 devices Transfer interval 1 s Device speed < 5 m·s-1 User-experienced data rate 10 kbps Application KPIs AI/ML KPIs KPI Value KPI Value Communication service availability > 99 % Training complexity 6< 2.5 GFLOPS Positioning/Location service availability > 99 % Inferencing accuracy > 95 % Positioning/Location service accuracy < 1 m Inferencing latency < 100 ms Command service availability N.A. AI/ML capabilities Yes 4.8.4 Key Value Indicators Environmental KVIs Impact Score: 3 This use case significantly contributes to environmental sustainability by enabling precise resource management and waste reduction. Through granular tracking and real-time insights, it facilitates a reduction in spoilage, particularly for perishable goods or sensitive materials in laboratories and pharmaceutical settings, thereby minimizing waste generation. The optimized utilization of assets and resources, driven by data-informed decisions, reduces overall consumption. Furthermore, the implementation of predictive maintenance, enabled by continuous monitoring of critical tools and equipment, extends asset lifecycles and reduces the need for premature replacements, leading to a decrease in manufacturing and disposal-related environmental burdens. Social KVIs Impact Score: 3 The social impact of this ambient IoT use case is substantial, primarily through the enhancement of safety and the improvement of working conditions. For instance, 4< 10 µW for passive ENDs; < 1 mW for active ENDs 5< 100 µW for passive ENDs; < 5 mW for active ENDs 6Considering fingerprinting-based positioning [75]. Page 63 of 94 D5.1 - Use cases, KPIs and KVIs Social KVIs Impact Score: 3 By offering an unobtrusive, context-aware guide, the system can enhance visitor engagement and accessibility—especially for users who prefer personalized, on-demand content without wearing or handling extra devices. The touchless interaction model also benefits hygiene and comfort in public spaces. Conversely, reliance on visitors’ own smartphones or dedicated RF readers may exclude those without compatible devices or with limited digital literacy, potentially widening the digital-access divide. Ensuring universal access—via loaner readers or multilingual, low-tech fallback options—will mitigate these social risks, but reduce the economical benefit. Economic KVIs Impact Score: 3 Most contemporary museums rely on battery-powered handheld guides or badges, which incur significant ongoing costs for battery purchase, charging infrastructure, labour to swap or charge units, and eventual battery disposal. By contrast, a fully passive backscatter-based guide eliminates all battery-related expenses over its operational life. The ultra-long tag lifetime (>20 years) means virtually zero maintenance labour and no recurring consumables, delivering substantial total cost of ownership savings. However, initial outlays for RF emitters, reader integration, and backed software may be higher than for simple battery-powered handsets. Achieving return of investment (ROI) depends on visitor volume and exhibit scale; smaller or seasonal installations might need phased deployment or shared infrastructure models to amortize capital investment effectively. Innovation KVIs Impact Score: 4 Deploying fully passive RF backscatter for context-aware museum guides represents a novel fusion of A-IoT and cultural-heritage engagement. The approach sidesteps lighting constraints and unlocks long-term, maintenance-free installations in sensitive exhibition environments, advancing the state of the art in low-power, location-based services. Nonetheless, the technology remains emergent: integration with unmodified COTS devices (e.g., smartphones), AI-enhanced tag detection, and scalable management tools are still maturing. Continued innovation in backscatter modulation schemes, AI/ML detection algorithms, and standardized protocols will determine how broadly this concept can be adopted. Page 70 of 94 D5.1 - Use cases, KPIs and KVIs Environmental Social Economic Innovation 12345 Figure 4.16: Impact Factors for Museum Guide. Page 71 of 94 D5.1 - Use cases, KPIs and KVIs 4.10 End-to-end Logistics Table 4.27: Use Case Families for End-to-end Logistics use case. Use Case Families Hard-to-reach devices Massive-scale sensor networks Extreme-lifetime applications Table 4.28: END classes for End-to-end Logistics use case. END Classes UP ENDs DB ENDs DP ENDs CO ENDs 4.10.1 Description Efficient logistics management is a cornerstone of modern supply chains, ensuring that goods are delivered accurately, safely, and on time. End-to-end logistics encompasses most of the lifecycle of a product—from its origin at the manufacturing site to its final delivery to the consumer. This process involves multiple stages, including production, warehousing, transportation, and distribution, each of which can benefit from intelligent monitoring and automation. The emergence of the A-IoT has introduced a new paradigm in logistics. By embedding intelligence into everyday objects and environments, A-IoT enables seamless data collection, communication, and decision-making across the supply chain. RFID technology is a foundational element in this context. Passive RFID tags, attached to products or packaging during manufacturing, store unique identifiers and essential metadata. As goods move through the supply chain, RFID readers—strategically placed at checkpoints such as warehouses, loading docks, and retail outlets—capture data from these tags. This enables real-time tracking of product location and status. Although passive tags do not contain their own power source, they can be energized via WPT from nearby readers or mobile devices, ensuring continuous operation without the need for batteries. Beyond traditional RFID,A-IoT expands the range of capabilities. For example, temperature-sensitive goods can be monitored using semi-passive tags over a longer period of time with internal supercapacitors to expand the operation. In addition, on-board units (OBUs) installed in delivery vehicles provide valuable telemetry data, including vehicle location, speed, fuel consumption, and maintenance status. This information supports dynamic route optimization, predictive maintenance, and improved fleet management. Together, smart tags, RFID, and other A-IoT devices form a distributed, intelligent infrastructure that enhances visibility, traceability, and efficiency across the entire logistics chain. By leveraging the capabilities of A-IoT and ENDs, businesses can achieve more resilient, adaptive, and sustainable logistics operations. A-IoT integration into logistics workflows, in this example RFID as enabling technology, can follow several strategic approaches, each offering varying levels of traceability, durability, and cost-efficiency: Page 72 of 94 D5.1 - Use cases, KPIs and KVIs Returns Diagnostics Production Warehouse Transport Delivery Secure Cloud Data Return Items Figure 4.17: Illustration of End-to-end Logistics use case •External Tagging: tags are affixed to the exterior of items, containers, or pallets. This method is widely used for inventory tracking and shipment verification due to its simplicity and low implementation cost. •Packaging-Integrated Tagging: Tags are embedded directly into product packaging or a reusable transport item (RTI). This approach enhances durability and reduces the risk of tag detachment during handling and transit. •Design-Integrated Tagging: tag functionality is incorporated into the product’s design, often in a detachable or semi-permanent form. This strategy supports traceability throughout the product lifecycle while allowing for tag reuse or removal after sale. •Embedded Item Tagging: tag components are fully integrated into the product itself during manufacturing. This method enables seamless tracking and authentication, particularly valuable for high-value goods. This tagging method can be further enhanced by combining the device function with the tag, for example, to read out diagnostic information of the device over its life cycle by the end customer. 4.10.2 Functional Requirements Communication RFID tags must support globally unique identification and be readable in dense tag environments, such as warehouses or shipping containers, where hundreds or thousands of tags may be present simultaneously. Bulk identification with a read accuracy exceeding 99% is essential to ensure reliability. Communication protocols must support anti-collision and error correction mechanisms, and operate across various frequency bands (e.g., ultra-high frequency (UHF),high frequency (HF)) depending on regional regulations and application needs. Positioning/Location While RFID passive tags do not possess active transmission capabilities, they can support positioning when deployed within a dense infrastructure of fixed reader. Same with A-IoT devices if they have no active communication capabilities. These systems estimate the position of tagged items using techniques such as signal triangulation, ToF,orRSSI. However, achieving sub-meter Page 73 of 94 D5.1 - Use cases, KPIs and KVIs accuracy with passive systems alone remains challenging and often requires environmental calibration and redundancy. To enhance both positioning precision and contextual location awareness, hybrid systems are increasingly employed. These combine RFID with complementary technologies such as UWB for high-accuracy indoor positioning, or GPS for outdoor tracking. A-IoT devices, which often include multiple harvesting sources, sensors and connectivity modules, further enrich this ecosystem by providing continuous data streams that support real-time localization and environmental monitoring. Management Effective logistics management requires seamless integration of A-IoT data into resource planning and warehouse management systems. This includes real-time inventory updates, automated stock level alerts, and exception handling (e.g., damaged or missing items). Middleware must support data filtering, aggregation, and event-based triggers to reduce network load and improve responsiveness. Further, the environment, tag number and type is constantly changing. Therefore, flexible network structures must be implemented to be able to add, remove, or reconfigure new devices. Collected Information and Network Exposure END tags enhance logistics by monitoring environmental factors such as temperature, humidity, vibration, and shock. These parameters are essential for maintaining product quality, especially in cold chain and fragile goods transport. Collected data must be securely transmitted to cloud or edge platforms using encrypted and authenticated channels. This protects against tampering and unauthorized access, ensuring data integrity throughout the supply chain. Energy END tags operate without internal batteries, relying solely on harvested energy from ambient sources such as RF WPT and/or others. Energy harvesting must be sufficient to power sensing, data storage, and communication functions. RFID readers (both fixed and mobile) must ensure reliable tag activation at every stage of the logistics process, including during transit and at delivery checkpoints. User-Safety and Robustness The system must be robust against environmental stressors such as dust, moisture, temperature fluctuations, and electromagnetic interference. Tags and readers should comply with relevant safety standards and be designed for long-term durability, but should also consider their end of life (EoL) with option to recycle or reuse the tags. Scalability is also essential, with systems expected to handle thousands of tags simultaneously without degradation in performance. Secure digital communication of END devices in end-to-end logistics also improves user privacy. Unlike printed labels, which expose static information, encrypted transmissions allow controlled access, supporting compliance and trust. Page 74 of 94 D5.1 - Use cases, KPIs and KVIs AI/ML AI/ML capabilities are primarily implemented on the infrastructure side (e.g., edge servers, cloud platforms) rather than on the passive RFID tags themselves. Due to their ultra-low power design, passive tags lack the computational resources for onboard processing. However, AI/ML can be used to analyse aggregated RFID data for anomaly detection, predictive maintenance, and route optimization. These insights can significantly enhance operational efficiency and decision-making across the logistics chain. 4.10.3 Key Performance Indicators Table 4.29: KPIs for End-to-end Logistics use case. Device KPIs Network KPIs KPI Value KPI Value Sensing accuracy > 95 % End-to-end latency < 1 s Device autonomy > 30 days Age of information < 1 s Avg. power consumption < 1 µW Packet loss rate <0.1 % Peak power consumption < 10 µW Message size < 10 kbits Communication range < 10 m Service area dimension < 100 m2 Max simultaneous devices < 1000 devices Transfer interval < 1 s Device speed < 5 m·s-1 User-experienced data rate 100 kbps Application KPIs AI/ML KPIs KPI Value KPI Value Communication service availability > 99 % Training complexity N.A. Positioning/Location service availability > 99 % Inferencing accuracy N.A. Positioning/Location service accuracy < 0.5 m Inferencing latency N.A. Command service availability Not required AI/ML capabilities Not required 4.10.4 Key Value Indicators Environmental KVIs Impact Score: 3 The use of RFID technology in end-to-end logistics highly depends on the exact implementation of the technology. If normal barcode labels are replaced one-to-one with RFID tags, there will not be a huge environmental impact. But if the entire supply chain is adapted for the new technology, and recyclable RTI are implemented, or tags are directly integrated into the devices, therefore reducing the need for further package labelling, and enabling additional information throughout the lifespan of the device, environmental impacts can be substantial. Page 75 of 94 D5.1 - Use cases, KPIs and KVIs Social KVIs Impact Score: 4 By implementing RFID technology in end-to-end logistics, improvements to the efficiency and accuracy of supply chain operations can be made. This leads to faster delivery times, reduced stock-outs, and improved customer satisfaction. The real-time visibility provided by RFID technology also enhances the transparency of the supply chain, allowing consumers to track the journey of their products from the manufacturer to their doorstep. Device implemented tags further allow the end user to detect the authenticity of the device, or enable to gather more information on the device throughout its lifespan. Economic KVIs Impact Score: 4 By reducing the need for manual inventory checks and improving the accuracy of inventory records, RFID technology helps to reduce labor costs and minimize losses due to stock discrepancies, leading to cost savings for businesses. The improved efficiency of logistics operations also leads to cost savings in transportation and warehousing, further enhancing the economic benefits of RFID technology. Innovation KVIs Impact Score: 3 The use of RFID technology in end-to-end logistics represents a moderate innovation in supply chain management. By providing real-time visibility and accurate tracking of products, RFID technology enables businesses to optimize their logistics operations and improve overall supply chain efficiency. Mostly the innovative aspects will be present in the ability of the end user interfacing with the device tags, for example, to check authenticity, read out device information very easily, like manuals or data-sheets, or even check device status, if the tag is integrated into the device functionality. Environmental Social Economic Innovation 12345 Figure 4.18: End-to-end Logistics. Page 76 of 94 D5.1 - Use cases, KPIs and KVIs 4.11 Industrial Predictive Maintenance Table 4.30: Use Case Families for Industrial Predictive Maintenance use case. Use Case Families Hard-to-reach devices Massive-scale sensor networks Extreme-lifetime applications Table 4.31: END classes for Industrial Predictive Maintenance use case. END Classes UP ENDs DB ENDs DP ENDs CO ENDs 4.11.1 Description Predictive maintenance supports the identification of early signs of equipment wear, reducing unplanned downtime, and maximizing service schedules. In industrial applications, energy-neutral sensing enables continuous, wireless monitoring of hard-to-reach or distributed assets without the hassle of battery maintenance. Figure 4.19: Illustration of Predictive Maintenance use case. Typical applications are as follows. •Stagnant water detection on factory or warehouse flat roofs, where stagnant water can cause structural damage if not detected. •Machine hour counters, with runtime monitoring determining maintenance planning based on actual usage instead of fixed intervals. •Vibration analysis of motors, fans, or compressors, enabling early detection of misalignment, imbalance, or mechanical wear through spectral processing. •Thermal monitoring, with gradual temperature drift in bearings or electrical components indicating impending faults. Page 77 of 94 D5.1 - Use cases, KPIs and KVIs These sensors often operate with periodic power supplied by harvesting sources such as vibration, indoor light, or temperature gradients. Event-driven updates and local processing reduce energy demands and enable batteryless or battery-assisted operation for years. 4.11.2 Functional Requirements Communication To conserve energy, devices communicate infrequently or on event detection. Low-throughput short-range wireless protocols (e.g., sub-GHz, BLE, or industrial LoRa variants) are sufficient to transmit alerts or status messages. Where backscatter or wake-up radios are used, communication is asynchronous and low-power. Gateways gather the data and forward them to the cloud or local control systems for diagnostics or visualization. The employed Medium Access Control (MAC) strategies favour energy efficiency over complexity. Simple contention-based approaches (e.g., ALOHA variants) are common but offer limited robustness, especially in dense deployments, due to their "fire-and-forget" nature and lack of collision avoidance. More reliable alternatives, such as duty-cycled time division-multiple access (TDMA), can provide predictable access with fewer collisions, but require synchronization and coordination. Positioning/Location Precise positioning is usually not required. The device installation is associated with specific machines or locations, and the metadata can be manually tagged during the installation. In distributed or mobile assets, the granularity to the zone or room level can be sufficient. Management Each sensor must have a unique ID and be associated with its intended asset in an asset management system or digital twin. Uptime, signal quality, and power availability as minimum health indicators can be checked from time to time. Devices must support fault detection and marking to allow scheduled intervention or replacement when necessary. In addition, support for OTA updates and remote parameter configuration is essential, enabling firmware updates or operational adjustments to be applied without physical access. These updates can be triggered manually by operators in response to diagnostics, or automatically pushed by cloud-based systems as part of policy-driven workflows. For example, if the cloud detects that a sensor is sending data more frequently than necessary, causing unnecessary energy drain, it can automatically issue a downlink to adjust the reporting interval or detection threshold. Beyond device-level management, the network must support effective commissioning of new devices and orderly decommissioning of obsolete or malfunctioning ones. Collected Information and Network Exposure The information gathered by ENDs used in industrial predictive maintenance context typically includes runtime logs such as hours active, event counts like the presence of water or vibration anomalies, and condition metrics including temperature and vibration spectrum analysis. A key challenge in this context is balancing the demand for near real-time access to data with the need to conserve device energy, as more frequent transmissions significantly increase energy consumption and thus reduce device lifetime. Page 78 of 94 D5.1 - Use cases, KPIs and KVIs Industrial telemetry is not required to be sensitive, but must be protected with encryption and secure authentication. Gateways and storage need to offer integrity, particularly in those instances where data affect maintenance choices or regulatory compliance. Energy Energy-neutral operation can rely on a variety of energy harvesting sources: most commonly ambient light, possibly combined with others such as vibration or thermal gradients to improve robustness by compensating for the variability of individual sources. Devices buffer harvested energy using (a combination of) energy storage elements such as supercapacitors (enabling rapid discharge for intermittent sensing, processing, and transmission tasks) and batteries (leveraging high energy density). Event-driven architectures avoid unnecessary energy consumption. Battery-assisted nodes can be acceptable for multi-year deployment but should be avoided where replacement is not possible. User-Safety and Robustness Devices must be industrial-grade: resistant to extremes of dust, moisture, vibration, and temperature to ensure reliable operation under harsh conditions. They must also function reliably amid electrical noise, metallic obstructions, and harsh electromagnetic environments typical of industrial settings, where wireless communication faces reflections, interference, and signal attenuation. To prevent disruption, devices should be designed so that their operation does not interfere with the assets they monitor. The physical size of the device must be appropriate for the use case: generally less restrictive in open deployments such as rooftop monitoring but potentially critical in confined spaces such as size-constrained machinery, where compactness is necessary to avoid impeding operation or maintenance. Finally, devices should incorporate fail-safes such as watchdog timers and fallback logic to handle communication loss or internal failures gracefully. AI/ML Basic signal processing techniques, such as peak detection or fast Fourier transform (FFT), can typically be performed locally at the edge to pre-filter or classify data, thus reducing the volume of data that needs to be transmitted. Local ML on the sensor is constrained by energy availability, which leads to a trade-off between performing lightweight local inference with potentially lower accuracy and offloading data to cloud-based ML models, which may offer higher accuracy, but require more energy for communication. Cloud-side AI models can aggregate data across multiple assets, identify long-term trends, and provide more accurate failure forecasts. Page 79 of 94 D5.1 - Use cases, KPIs and KVIs and maintenance operations. Transfer learning techniques may be applied to enable robots and standalone ENDs to refine AI models using shared environmental context, reducing redundant retraining. Further, apart from being used for situational awareness and robot control purposes (such as manoeuvring), AI/ML models can be used to assist with communication tasks, for example, by recommending different END duty cycling or signal transmission/ reception policies for interference mitigation or avoidance. The system must also enable issuance and transfer of predicted END energy consumption along intended robot paths, leveraging real-time and historical data to optimize routing and resource allocation. 4.12.3 Key Performance Indicators Table 4.35: KPIs for Cooperative Mobile Robots use case. Device KPIs Network KPIs KPI Value KPI Value Sensing accuracy > 95 % End-to-end latency < 0.5 s Device autonomy > 10 years Age of information < 0.5 s Avg. power consumption < 1 mW Packet loss rate <0.1 % Peak power consumption < 50 mW Message size < 1 kbit Communication range < 20 m Service area dimension < 10000 m2 Max simultaneous devices < 1000 devices Transfer interval On request Device speed < 5 m·s-1 User-experienced data rate 100 kbps Application KPIs AI/ML KPIs KPI Value KPI Value Communication service availability > 99.9 % Training complexity < 1 GFLOP Positioning/Location service availability > 99 % Inferencing accuracy > 90 % Positioning/Location service accuracy < 1 m Inferencing latency < 10 s Command service availability Not required AI/ML capabilities Yes 4.12.4 Key Value Indicators Environmental KVIs Impact Score: 3 ENDs equipped with energy harvesting and wireless power transfer can reduce reliance on conventional batteries, ensuring sustainable operations. However, robotic deployment still requires energy for communication, processing, and robot charging, contributing to environmental impact. Mitigation strategies must incorporate renewable energy sourcing, energy-efficient hardware/software design, and recyclable components, reducing lifecycle waste. Page 86 of 94 D5.1 - Use cases, KPIs and KVIs Social KVIs Impact Score: 2 A-IoT might contribute to longer operational uptime and reduced maintenance efforts in a cooperative robotic system, which could slightly influence workforce roles (e.g., fewer tasks related to battery replacement or device monitoring). However, this feature would not significantly alter the broader social considerations, as compared to a cooperative mobile robot system incorporating conventional IoT devices. Economic KVIs Impact Score: 4 Unlike conventional IoT-equipped robots, END-enabled systems leverage energy harvesting WPT, thereby reducing dependency on battery replacements and lowering long-term operational costs. This shift enhances cost-effectiveness, especially for smaller businesses that may struggle with high maintenance expenses. From a market perspective, A-IoT adoption could lower entry barriers for businesses by reducing hardware costs, but it may also intensify monopolization risks if only large corporations can afford advanced A-IoT-powered robotics. In addition, regulatory frameworks must evolve to address data security, energy harvesting policies, and A-IoT-driven automation governance, ensuring fair competition and ethical deployment. Innovation KVIs Impact Score: 4 The thermal energy dissipated from AI/ML workload processing can be partially captured through thermoelectric harvesting and reallocated to critical robotic tasks, such as continuing model training or ensuring uninterrupted sensor data transmission, hence, reducing dependency on external power sources. Therefore, a KVI in this context would be on the “energy reuse potential” training or inference data may carry. Also, before initiating Transfer Learning, the network must assess the usefulness of the AI/ML model in its current context, weighing this decision against an alternative decision on wirelessly transferring energy. A new KVI “AI/ML model usefulness” can be shaped to evaluate the most effective decision in space and in time. Page 87 of 94 D5.1 - Use cases, KPIs and KVIs Environmental Social Economic Innovation 12345 Figure 4.22: Cooperative Mobile Robots. Page 88 of 94 D5.1 - Use cases, KPIs and KVIs Chapter 5 Conclusions In this document, an analysis of use cases for A-IoT is conducted. Initially, A-IoT and the technologies supporting it are introduced, providing the context for understanding the rest of the document. A review of the state of the art on the use case definition for A-IoT is performed, focusing on activities carried out by SDOs and European projects. The core of the document describes several proposed use cases. These are examined through the identification of functional requirements, KPIs, and KVIs in four dimensions: environmental, social, economic, and innovative. With this approach, a view of the potential of A-IoT is provided, and insights into its practical implementation are generated. It is important to highlight that the KVI analysis performed by the experts contributing to this deliverable should be considered as preliminary, as the factors that it considers are difficult to quantify and anticipate. Further stages of this analysis and more research on END capabilities are needed to make it more accurate. Deliverable D5.1 uses D2.1 as one of its primary references, leveraging on the classification of ENDs. Moreover, D5.1 analyses use cases that will serve as the foundation for the upcoming deliverables D5.2 and D5.3, which will develop the corresponding proofs of concept and demonstrators. D5.1 is also a reference for deliverables D3.1, D3.2, D3.3, D4.1, D4.2, and D4.3, as it provides relevant KPIs and KVIs to support the development of protocols, infrastructures and algorithms to be described in those documents. D5.1 directly contributes to the achievement of Milestone 2: "Technology trade-offs, use cases, KPIs, and KVIs defined". Additionally, it contributes indirectly to the fulfilment of Specific Objective 4: "Implement, test, validate, and demonstrate end-to-end intelligent END capabilities, and 6G network functionality" by defining the use cases that will guide the development of the AMBIENT-6G’s proofs of concept and demonstrators. Page 89 of 94 D5.1 - Use cases, KPIs and KVIs References [1] AMBIENT-6G, D2.1 END architecture, technology trade-off, and characterization, May 2025. [2] T. Pirson, L. Golard, and D. Bol, “Evaluating the (ir)relevance of IoT solutions with respect to environmental limits based on LCA and backcasting studies,” in Ninth Computing within Limits 2023, LIMITS, Jun. 2023. [3] International Telecommunication Union. “Creating a circular economy for ICT equipment.” (2023), [Online]. Available: https://www.itu.int/en/mediacentre/backgrounders/Pages/e-waste.aspx (visited on 11/24/2023). [4] W. Li, Q. Liu, Y. Zhang, C. Li, Z. He, W. C. H. Choy, P. J. Low, P. Sonar, and A. K. K. Kyaw, “Biodegradable Materials and Green Processing for Green Electronics,” Advanced Materials, vol. 32, no. 33, p. 2 001 591, 2020. DOI: 10.1002/adma.202001591. [5] D. Danninger, R. Pruckner, L. Holzinger, R. Koeppe, and M. Kaltenbrunner, “MycelioTronics: Fungal mycelium skin for sustainable electronics,” Science Advances, vol. 8, no. 45, eadd7118, 2022. DOI: 10. 1126/sciadv.add7118. [6] A. S. Weddell, M. Magno, G. V. Merrett, D. Brunelli, B. M. Al-Hashimi, and L. Benini, “A survey of multisource energy harvesting systems,” in 2013 Design, Automation & Test in Europe Conference & Exhibition (DATE), 2013, pp. 905–908. DOI: 10.7873/DATE.2013.190. [7] H. Yang, M. Huang, M. Ren, and X. Li, “Piezoelectric energy harvesting interface circuit for small area and low power consumption‚ a review,” Measurement, vol. 242, p. 116 051, 2025, ISSN: 0263-2241. DOI: https : / / doi . org / 10 . 1016 / j . measurement . 2024 . 116051. [Online]. Available: https : / / www . sciencedirect.com/science/article/pii/S0263224124019365. [8] S.-E. Kim, T. Kang, K.-I. Oh, M. J. Park, H.-I. Park, I. G. Lim, and J.-J. Lee, “Energy Management Integrated Circuit for Multi-Source Energy Harvesters in WBAN Applications,” Applied Sciences, vol. 8, no. 8, 2018, ISSN: 2076-3417. DOI: 10.3390/app8081262. [Online]. Available: https://www.mdpi.com/ 2076-3417/8/8/1262. [9] C. Delgado, J. M. Sanz, and J. Famaey, “On the Feasibility of Battery-Less LoRaWAN Communications Using Energy Harvesting,” in 2019 IEEE Global Communications Conference (GLOBECOM), 2019. [10] C. Delgado, J. M. Sanz, C. Blondia, and J. Famaey, “Batteryless LoRaWAN Communications Using Energy Harvesting: Modeling and Characterization,” IEEE Internet of Things Journal, vol. 8, no. 4, pp. 2694–2711, 2021. [11] M. Capuzzo, C. Delgado, A. K. Sultania, J. Famaey, and A. Zanella, “Enabling Green IoT: Energy-Aware Communication Protocols for Battery-less LoRaWAN Devices,” in Proceedings of the 24th International ACM Conference on Modeling, Analysis and Simulation of Wireless and Mobile Systems, 2021, pp. 95–98. [12] A. K. Sultania, C. Delgado, and J. Famaey, “Enabling Low-Latency Bluetooth Low Energy on Energy Harvesting Batteryless Devices Using Wake-Up Radios,” Sensors, vol. 20, no. 18, 2020. [13] A. K. Sultania, C. Delgado, C. Blondia, and J. Famaey, “Downlink Performance Modeling and Evaluation of Batteryless Low Power BLE Node,” Sensors, vol. 22, no. 8, 2022. [14] A. K. Sultania and J. Famaey, “Batteryless Bluetooth Low Energy Prototype With Energy-Aware Bidirectional Communication Powered by Ambient Light,” IEEE Sensors Journal, vol. 22, no. 7, pp. 6685–6697, 2022. [15] A. K. Sultania and J. Famaey, “Batteryless NB-IoT prototype for bidirectional communication powered by ambient light,” Ad Hoc Networks, vol. 142, p. 103 100, 2023. [16] L. Moons, S. Nasser, A. Sabovic, R. K. Singh, and J. Famaey, “Evaluating Fast and Grant-Free Uplink Access in Next-Generation Cellular IoT Networks,” in Proceedings of the 2024 3rd International Conference on 6G Networking (6GNet), IEEE, 2024, pp. 19–24. DOI: 10.1109/6GNet63182.2024.10765754. [Online]. Available: https://doi.org/10.1109/6GNet63182.2024.10765754. Page 90 of 94 D5.1 - Use cases, KPIs and KVIs [17] N. A. Mitsiou, S. A. Tegos, P. D. Diamantoulakis, P. G. Sarigiannidis, and G. K. Karagiannidis, “Proactive Scheduling for Zero-Energy Device Networks With Fast Uplink Grant,” IEEE Wireless Communications Letters, vol. 12, no. 7, pp. 1184–1188, 2023. [18] “3GPP TR 38.769 V19.0.0 Study on solutions for Ambient IoT (Internet of Things) in NR.” [Online]. Available: https://www.3gpp.org/ftp/Specs/archive/38_series/38.769/38769-j00.zip. [19] R. Mishra and A. Mishra, “Current research on Internet of Things (IoT) security protocols: A survey,” Computers & Security, p. 104 310, 2025. [20] G. Callebaut, G. Leenders, J. Van Mulders, G. Ottoy, L. De Strycker, and L. Van der Perre, “The Art of Designing Remote IoT Devices–Technologies and Strategies for a Long Battery Life,” Sensors, vol. 21, no. 3, Jan. 2021, ISSN: 1424-8220. DOI: 10.3390/s21030913. [Online]. Available: https://www.mdpi. com/1424-8220/21/3/913. [21] P. Miranda, M. Siekkinen, and H. Waris, “TLS and energy consumption on a mobile device: A measurement study,” in 2011 IEEE Symposium on Computers and Communications (ISCC), IEEE, 2011, pp. 983–989. [22] K. O. Adefemi Alimi, K. Ouahada, A. M. Abu-Mahfouz, and S. Rimer, “A survey on the security of low power wide area networks: Threats, challenges, and potential solutions,” Sensors, vol. 20, no. 20, p. 5800, 2020. [23] X. Yang, L. Shu, Y. Liu, G. P. Hancke, M. A. Ferrag, and K. Huang, “Physical security and safety of IoT equipment: A survey of recent advances and opportunities,” IEEE Transactions on Industrial Informatics, vol. 18, no. 7, pp. 4319–4330, 2022. [24] M.-K. Oh, S. Lee, and Y. Kang, “Wi-sun device authentication using physical layer fingerprint,” in 2021 International Conference on Information and Communication Technology Convergence (ICTC), IEEE, 2021, pp. 160–162. [25] F. Alwahedi, A. Aldhaheri, M. A. Ferrag, A. Battah, and N. Tihanyi, “Machine learning techniques for IoT security: Current research and future vision with generative AI and large language models,” Internet of Things and Cyber-Physical Systems, vol. 4, pp. 167–185, 2024. [26] J. Van Mulders, S. Boeckx, J. Cappelle, L. Van der Perre, and L. De Strycker, “UAV-based solution for extending the lifetime of IoT devices: efficiency, design and sustainability,” Frontiers in Communications and Networks, vol. Volume 5 - 2024, 2024, ISSN: 2673-530X. DOI: 10.3389/frcmn.2024.1341081. [27] B. J. B. Deutschmann, U. Muehlmann, A. Kaplan, G. Callebaut, T. Wilding, B. Cox, L. V. der Perre, F. Tufvesson, E. G. Larsson, and K. Witrisal, “Physically Large Apertures for Wireless Power Transfer: Performance and Regulatory Aspects,” IEEE Wireless Communications Magazine, 2025, Accepted for publication. [Online]. Available: https://arxiv.org/abs/2503.06807. [28] U. Kunnath Ganesan, R. Sarvendranath, and E. G. Larsson, “BeamSync: Over-the-Air Synchronization for Distributed Massive MIMO Systems,” IEEE Transactions on Wireless Communications, vol. 23, no. 7, pp. 6824–6837, 2024. DOI: 10.1109/TWC.2023.3335089. [29] B. J. B. Deutschmann and P. Vouras, Spatiotemporal Synchronization of Distributed Arrays using ParticleBased Loopy Belief Propagation, 2025. arXiv: 2504 .00744 [eess.SP]. [Online]. Available: https :/ / arxiv.org/abs/2504.00744. [30] K. Wu, J. A. Zhang, X. Huang, R. W. Heath, and Y. J. Guo, “Green Joint Communications and Sensing Employing Analog Multi-Beam Antenna Arrays,” IEEE Communications Magazine, vol. 61, no. 7, pp. 172– 178, 2023. DOI: 10.1109/MCOM.007.2200495. [31] A. Azarbahram, O. L. A. López, and M. Latva-Aho, “Waveform Optimization and Beam Focusing for NearField Wireless Power Transfer With Dynamic Metasurface Antennas and Non-Linear Energy Harvesters,” IEEE Transactions on Wireless Communications, vol. 24, no. 2, pp. 1031–1045, 2025. DOI: 10.1109/TWC. 2024.3503908. [32] V. Jamali, A. M. Tulino, G. Fischer, R. R. Müller, and R. Schober, “Intelligent Surface-Aided Transmitter Architectures for Millimeter-Wave Ultra Massive MIMO Systems,” IEEE Open Journal of the Communications Society, vol. 2, pp. 144–167, 2021. DOI: 10.1109/OJCOMS.2020.3048063. [33] L. D’Angelo, B. J. Deutschmann, and K. Witrisal, “Combined Wideband Channel Estimation and Direct Link Interference Mitigation in Bistatic Backscatter Systems,” in 2024 IEEE 25th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC), 2024, pp. 506–510. DOI: 10.1109/ SPAWC60668.2024.10694033. [34] A. Kaplan, J. Vieira, and E. G. Larsson, “Direct Link Interference Suppression for Bistatic Backscatter Communication in Distributed MIMO,” IEEE Trans. Wireless Commun., vol. 23, no. 2, pp. 1024–1036, 2024. Page 91 of 94 D5.1 - Use cases, KPIs and KVIs [35] O. L. A. López, H. Alves, R. D. Souza, and S. Montejo-Sánchez, “Statistical Analysis of Multiple Antenna Strategies for Wireless Energy Transfer,” vol. 67, no. 10, pp. 7245–7262, 2019. DOI: 10.1109/TCOMM. 2019.2928542. [36] R. M. Dreifuerst and R. W. Heath, “Massive MIMO in 5G: How Beamforming, Codebooks, and Feedback Enable Larger Arrays,” IEEE Communications Magazine, vol. 61, no. 12, pp. 18–23, 2023. DOI: 10.1109/ MCOM.001.2300064. [37] B. J. B. Deutschmann, T. Wilding, M. Graber, and K. Witrisal, “XL-MIMO Channel Modeling and Prediction for Wireless Power Transfer,” in 2023 IEEE International Conference on Communications Workshops (ICC Workshops), 2023, pp. 1355–1361. DOI: 10.1109/ICCWorkshops57953.2023.10283480. [38] TensorFlow, TensorFlow Lite for Microcontrollers,https://www.tensorflow.org/lite/microcontrollers, 2024. [39] Arm Ltd., CMSIS-NN: Efficient Neural Network Kernels for Arm Cortex-M CPUs,https://developer. arm.com/solutions/machine-learning-on-arm/developer-material/how-to-guides/cmsis-nn, 2024. [40] A. Adi et al., “Energy-Aware Machine Learning for IoT Devices: A Survey,” IEEE Internet of Things Journal, 2020. [Online]. Available: https://arxiv.org/pdf/2309.00022. [41] A. Dhar et al., “On-Device Machine Learning: A Survey from the Algorithmic and Learning Theory Perspective,” arXiv preprint arXiv:2301.12345, 2023. [Online]. Available: https://arxiv.org/abs/2301.12345. [42] Y. Zhang et al., “Tiny Continual Learning on Energy-Harvesting Devices,” Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, vol. 6, no. 3, pp. 1–25, 2022. [43] B. Sudharsan, J. G. Breslin, M. Tahir, M. Intizar Ali, O. Rana, S. Dustdar, and R. Ranjan, “OTA-TinyML: Over the Air Deployment of TinyML Models and Execution on IoT Devices,” en, IEEE Internet Computing, vol. 26, no. 3, pp. 69–78, May 2022, ISSN: 1089-7801, 1941-0131. DOI: 10.1109/MIC.2021.3133552. [Online]. Available: https://ieeexplore.ieee.org/document/9811289/. [44] 3GPP, Specification # 38.848; Study on Ambient IoT (Internet of Things) in RAN; Release 18, Sep. 2023. [Online]. Available: https://portal.3gpp.org/desktopmodules/Specifications/SpecificationDetails. aspx?specificationId=4146. [45] 3GPP, Specification # 22.840; Study on Ambient power-enabled Internet of Things, Dec. 2023. [Online]. Available: https://portal.3gpp.org/desktopmodules/Specifications/SpecificationDetails. aspx?specificationId=4045. [46] 3GPP, Specification # 22.369; Service requirements for Ambient power-enabled IoT, Sep. 2024. [Online]. Available: https://portal.3gpp.org/desktopmodules/Specifications/SpecificationDetails. aspx?specificationId=4232. [47] IEEE, IEEE 802.11-23/0436r0; Technical Report on support of AMP IoT devices in WLAN, Mar. 2023. [Online]. Available: https://mentor.ieee.org/802.11/dcn/23/11-23-0436-00-0amp-technicalreport-on-support-of-amp-iot-devices-in-wlan.docx. [48] ADROIT6G, D2.1 Requirements, use cases and system architecture – Initial, Jul. 2023. [Online]. Available: https://adroit6g.eu/wp-content/uploads/2023/12/ADROIT6G_D2.1_Requirements-Use-casesArchitecture_Final.pdf. [49] ADROIT6G, D2.3 Requirements, use cases and system architecture, Jan. 2024. [Online]. Available: https: //adroit6g.eu/wp-content/uploads/2025/01/ADROIT6G_D2.3_Final.pdf. [50] HEXA-X, Deliverable D1.1 6G Vision, use cases and key societal values, Feb. 2021. [Online]. Available: https://hexa-x.eu/wp-content/uploads/2021/02/Hexa-X_D1.1.pdf. [51] HEXA-X-II, Deliverable D1.2 6G Use Cases and Requirements, Dec. 2023. [Online]. Available: https: //hexa-x-ii.eu/wp-content/uploads/2024/01/Hexa-X-II_D1.2.pdf. [52] IntellIoT, Deliverable D2.1 Use case specification & Open Call definition (first version), Mar. 2021. [Online]. Available: https://intelliot.eu/wp-content/uploads/2022/09/Deliverable-D2.1-Use-casespecification-and-Open-Call-definition-1.pdf. [53] IntellIoT, Deliverable D2.4 Use Case specification & Open Call definition (final version), Apr. 2022. [Online]. Available: https://intelliot.eu/wp-content/uploads/2022/10/IntellIoT_deliverable_D2.4. pdf. [54] REINDEER, D1.1. Use case-driven specifications and technical requirements and initial channel model, Oct. 2021. [Online]. Available: https://zenodo.org/records/5561844. [55] SUPERIOT, D1.4 Description of selected scenarios, applications, and their requirements (2), Dec. 2024. [56] “How the Bluetooth®Electronic Shelf Label Standard Will Impact the Smart Retail Market,” ABI Research, Tech. Rep., Jun. 2023. Page 92 of 94 D5.1 - Use cases, KPIs and KVIs [57] J.-S. Park and B.-J. Jang, “Electronic shelf label system employing a visible light identification link,” in 2016 IEEE Radio and Wireless Symposium (RWS), 2016, pp. 219–222. [58] E. E. Dikel, Y. E. Li, M. Vuotari, and S. Mancini, “Evaluating the standby power consumption of smart LED bulbs,” Energy and Buildings, vol. 186, pp. 71–79, 2019, ISSN: 0378-7788. DOI: https://doi.org/ 10.1016/j.enbuild.2019.01.019. [59] J. Cappelle, L. V. der Perre, E. Fitzgerald, S. Ravyts, W. Gajda, V. D. Smedt, B. Cox, and G. Callebaut, IoT on the Road to Sustainability: Vehicle or Bandit? 2024. arXiv: 2405.20706 [eess.SP]. [Online]. Available: https://arxiv.org/abs/2405.20706. [60] B. Aliakbarian, S. Ghirlandi, A. Rizzi, R. Stefanini, and G. Vignali, “Life cycle assessment of plastic and paper-based ultra high frequency RFID tags,” International Journal of RF Technologies, vol. 14, no. 1, pp. 17–32, Jun. 2024. DOI: 10.3233/RFT-230044. [61] J. Sennesael, A. Moerman, H. Rogier, and P. Van Torre, “Carbon-Ink Sensing Patterns for a Contactless Smart Diaper System,” in 2024 IEEE International Conference on Flexible and Printable Sensors and Systems (FLEPS), 2024, pp. 1–4. DOI: 10.1109/FLEPS61194.2024.10603972. [62] J. Li, Z. Xie, Z. Wang, Z. Lin, C. Lu, Z. Zhao, Y. Jin, J. Yin, S. Mu, C. Zhang, W. Gui, X. Liang, J. Wang, and W. Ding, “A triboelectric gait sensor system for human activity recognition and user identification,” Nano Energy, vol. 112, p. 108 473, 2023, ISSN: 2211-2855. DOI: https : / / doi . org / 10 . 1016 / j . nanoen.2023.108473. [Online]. Available: https://www.sciencedirect.com/science/article/pii/ S2211285523003105. [63] H. Hassan, M. Abdullah, M. Z. Hussain, M. Zulkifl Hasan, M. Mustafa, A. Khalid, R. Awan, U. Hussain, Z. A. Khan, and A. Javaid, “TinyML-Based Gait Recognition for User Detection on an STM32L4 IoT Node,” in 2024 IEEE 9th International Conference for Convergence in Technology (I2CT), 2024, pp. 1–6. DOI: 10.1109/I2CT61223.2024.10543752. [64] W. Chantaweesomboon, “Bluetooth Geo-Fence for Elderly and Patient Care,” in 2021 25th International Computer Science and Engineering Conference (ICSEC), 2021, pp. 252–255. DOI: 10.1109/ICSEC53205. 2021.9684648. [65] A. N. Arun, B. Lee, F. A. Castiblanco, D. R. Buckmaster, C.-C. Wang, D. J. Love, J. V. Krogmeier, M. M. Butt, and A. Ghosh, “Ambient IoT: Communications Enabling Precision Agriculture,” IEEE Communications Magazine, vol. 63, no. 4, pp. 137–143, 2025. DOI: 10.1109/MCOM.003.2400508. [66] S. Yang, Y. Bénédic, D.-T. Phan-Huy, J.-M. Gorce, and G. Villemaud, “Indoor Localization of Smartphones Thanks to Zero-Energy-Devices Beacons,” in 2024 18th European Conference on Antennas and Propagation (EuCAP), 2024, pp. 1–5. DOI: 10.23919/EuCAP60739.2024.10501308. [67] J. Liao, X. Wang, K. Ruttik, R. Jäntti, and D.-T. Phan-Huy, “In-Band Ambient FSK Backscatter Communications Leveraging LTE Cell-Specific Reference Signals,” IEEE Journal of Radio Frequency Identification, vol. 7, pp. 267–277, 2023. DOI: 10.1109/JRFID.2023.3280108. [68] J. Liao, T. Zhang, K. Ruttik, R. Jäntti, and D.-T. Phan-Huy, “Ambient Backscatter Communication in LTE Uplink Sounding Reference Signal,” arXiv preprint arXiv:2501.10952, 2025. [69] A. Fascista, B. J. B. Deutschmann, M. F. Keskin, T. Wilding, A. Coluccia, K. Witrisal, E. Leitinger, G. Seco-Granados, and H. Wymeersch, “Joint Localization, Synchronization and Mapping via Phase-Coherent Distributed Arrays,” IEEE Journal of Selected Topics in Signal Processing, vol. 19, no. 2, pp. 412–429, 2025. DOI: 10.1109/JSTSP.2025.3533111. [70] B. J. B. Deutschmann, E. Leitinger, and K. Witrisal, Geometry-Based Channel Estimation, Prediction, and Fusion, 2025. arXiv: 2503.17868 [eess.SP]. [Online]. Available: https://arxiv.org/abs/2503.17868. [71] A. Fascista, B. J. B. Deutschmann, M. F. Keskin, T. Wilding, A. Coluccia, K. Witrisal, E. Leitinger, G. SecoGranados, and H. Wymeersch, “Uplink Joint Positioning and Synchronization in Cell-Free Deployments with Radio Stripes,” in 2023 IEEE International Conference on Communications Workshops (ICC Workshops), 2023, pp. 1330–1336. DOI: 10.1109/ICCWorkshops57953.2023.10283650. [72] C. Nelson, X. Li, A. Fedorov, B. Deutschmann, and F. Tufvesson, “Distributed MIMO Measurements for Integrated Communication and Sensing in an Industrial Environment,” Sensors, vol. 24, no. 5, 2024, ISSN: 1424-8220. DOI: 10 . 3390 / s24051385. [Online]. Available: https : / / www . mdpi . com / 1424 - 8220/24/5/1385. [73] B. Deutschmann, C. Nelson, M. Henriksson, G. Marti, A. Kosasih, N. Tervo, E. Leitinger, and F. Tufvesson, “Accurate Direct Positioning in Distributed MIMO Using Delay-Doppler Channel Measurements,” in 2024 IEEE 25th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC), 2024, pp. 606–610. DOI: 10.1109/SPAWC60668.2024.10694279. Page 93 of 94 D5.1 - Use cases, KPIs and KVIs [74] L. Görtschacher, J. Grosinger, H. N. Khan, and W. Bösch, “Fast two dimensional position update system for UHF RFID tag tracking,” in 2017 IEEE MTT-S International Microwave Symposium (IMS), 2017, pp. 1331–1334. DOI: 10.1109/MWSYM.2017.8058857. [75] A. Pirnat, B. Bertalanič, G. Cerar, M. Mohorčič, M. Meža, and C. Fortuna, “Towards Sustainable Deep Learning for Wireless Fingerprinting Localization,” in ICC 2022 - IEEE International Conference on Communications, 2022, pp. 3208–3213. DOI: 10.1109/ICC45855.2022.9838464. [76] A. Abedi, M. H. Mazaheri, O. Abari, and T. Brecht, “Witag: Rethinking backscatter communication for wifi networks,” in Proceedings of the 17th ACM workshop on hot topics in networks, 2018, pp. 148–154. [77] E. Soltanaghaei, A. Dongare, A. Prabhakara, S. Kumar, A. Rowe, and K. Whitehouse, “Tagfi: Locating ultra-low power wifi tags using unmodified wifi infrastructure,” Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, vol. 5, no. 1, pp. 1–29, 2021. [78] I. O. for Standardization, Robots and robotic devices — Safety requirements for industrial robots — Part 1: Robots, 2011. [Online]. Available: https://www.iso.org/standard/51330.html. [79] R. I. Association and A. N. S. Institute, Industrial Robots and Robot Systems — Safety Requirements, 2012. [Online]. Available: https://webstore.ansi.org/standards/ria/ansiriar15062012. Page 94 of 94