Industrial collaborative environments integrating AI, Big Data and Robotics for smart manufacturing

ScienceDirect Available online at www.sciencedirect.com Procedia CIRP 128 (2024) 858–863 2212-8271 © 2024 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the scientific committee of the 34th CIRP Design Conference 10.1016/j.procir.2024.04.027 Keywords: AI, Smart manufacturing, Robotics; 1. Introduction Recent years have been marked by global scale occurrences including economic crisis and an unprecedented pandemic which have provided tremendous challenges for the manufacturing firms that had to rethink their production and business models[1].Under the new reality, reconfiguration has to consider many more dimensions than a simple rescheduling of pending orders. The production system itself needs to adopt both human and automated resources that can work together seamlessly or mutually exchange tasks, allowing the execution of any process plan in multiple, non-predetermined ways [2]. For an effective human automation collaboration though, not only the production resources such as robots have to adapt to humans, but the humans also need to learn how to benefit from automation [3]. Emerging technologies have been used to foster human-robot collaboration including Augmented Reality (AR) based tools for providing robot-related information (working volumes, trajectory visualization, sensing alerts), safety related information (danger zones, etc.),and step-bystep instructions [4]. Nevertheless, limited automation acceptance, misuse of automation, and automation failures remain relevant hindering the exploitation of the benefits of collaborative smart manufacturing. Thereby, approaches 34th CIRP Design Conference Industrial collaborative environments integrating AI, Big Data and Robotics for smart manufacturing Nikos Dimitropoulosa, George Michalosa, Zoi Arkoulia, George Kokotinisa, Sotiris Makrisa* aLaboratory for Manufacturing Systems and Automation, Department of Mechanical Engineering and Aeronautics, University of Patras, Patras, 26504, Greece * Corresponding author. Tel.: +30-2610-910160; fax: +30-2610-997314.E-mail address:[email protected] Abstract In recent years, global events have posed significant challenges to manufacturing firms' business models, necessitating adaptable production systems that integrate human and automated resources seamlessly. Fortunately, technological advancements over the past decade have facilitated flexible production solutions. This paper introduces the approach of the CONVERGING EU project for collaborative smart manufacturing systems aiming at increasing flexibility, efficiency, and operators’ satisfaction. The proposed approach is structured into three technical pillars: perception, adaptation, collaboration,enhanced by modules for fostering user experience. Perception involves identifying and recognizing features of parts, resources, and surroundings to infer and analyse their status, and formulate action plans. Automated adaptations through hardware and control system modifications allow for executing formulated plans while complying with user characteristics and needs. Collaboration focuses on the software and hardware interfaces to ensure safe and seamless interaction with collaborative robotic solutions. Training, ergonomics tracking, and adaptable interfaces are also proposed to pave the way toward social industrial environments. This approach is applied to four industrial scenarios from different manufacturing sectors: automotive, white goods, aeronautics and additive manufacturing. The design of systems incorporating AI and robotics is presented together with the expected impact including safer, more intuitive, and trustful human-robot interaction, but also more inclusive industrial environments. © 2024 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the scientific committee of the 34th CIRP Design Conference Nikos Dimitropoulos et al. / Procedia CIRP 128 (2024) 858–863 859 ensuring efficiency, reliability, and flexibility in collaborative environments together with reduction of health and safety risks, and positive user experience are needed. This paper discusses the CONVERGING approach (www.converging-project.eu) towards enabling industrial collaborative environments integrating AI, Big Data and robotics for smart manufacturing, trying to address the aforementioned bottlenecks. The state of the art is analysed in section 2, and the proposed approach is detailed in section 3. In Section 4, four reference use cases where the proposed approach is applied are analyzed. Section 5 discusses the expected impact of applying the proposed approach. Finally, Section 6 concludes the presented work and outlines next steps. 2. Literature review Over the last decade though, significant breakthroughs have been achieved on the enabling technology towards supporting production flexibility. Achievements in perception research include the use of vision laser-based sensors and stereo cameras capable of coping with changes in illumination both for indoors and outdoors applications. Software wise, model alignment, feature-based, descriptor-based (wavelets) or color filtering methods based on 2D or 3D datasets have been proposed. Deep learning-based methods (e.g., Yolo or TensorFlow) have prevailed though thanks to their convincing results for object recognition and rigid-body pose estimation[5]. Deep learning have also been deployed for multimodal fusion [6], and anomaly detection [7]. Furthermore, human motion detection has been used in the context of multimodal human-robot interaction and human aware robot control[8]. Simulation tools of digital world models have allowed for virtual commissioning and shorter lead times. Digital Twins [9],have enabled the development of virtual robot controllers for testing and verification of robot tasks. Commercial tools: Microsoft Robotics Developer Studio (MRDS), V-REP [10], SIEMENS Process Simulate and Dassault Systems Delmia can simulate single robot to complete lines. In Human Robot (HR) applications, task assignments have to vary dynamically at both planning and execution level. The fixed control logic (PLC) needs to be enhanced with an open architecture such as the one offered by services [11]. Methods for allocating assembly tasks to humans and using different models (Dual GSPN model, STRIPS operators, hybrid Bayesian networks etc.) and criteria (time, cost, changes etc.). Human aware task planners to generate plans for robots have been investigated [12]. Path planning and workspace sharing has been recently improved with robot trajectory re-planning for collision avoidance (e.g., RealTime Robotics), or to maintain a desired speed and separation with the human through MPC. Human-centered objectives of ‘legible’ robot motion have also been studied[13]. In terms of hardware, high-speed reconfigurable grippers for handling parts of different shapes and geometries[14], and robot end-effector with in-hand manipulation capabilities are among the latest trends.Robot programming is not yet userfriendly, but has been achieved in different ways, developing on motion-oriented robot languages, through visual guidance and imitation, voice commands and haptic interaction,or by using AR technology[15]. Currently, the development of a robot-agnostic, adaptive, user-friendly, and interactive programming environment for industrial robots exceeds the current state of the art. While progress is underway by various entities, particularly within the realm of the Robot Operating System (ROS), a comprehensive solution meeting all these criteria remains elusive[16]. ROSIN, a project funded by the EU, aimed to amplify ROS impact by making ROS-Industrial better and more business-friendly and accessible[17].The ability to share programs in a network of autonomous robots is also a widely sought breakthrough. 3. Approach The interaction of human machine interaction, and particularly in human robot interaction, entails bottlenecks such as technical limitations, safety concerns, task complexity, and inadequate human factors consideration. In response, CONVERGING introduces a comprehensive approach fostering the seamless operation of smart and reconfigurable production systems including multiple autonomous agents (collaborative robots, AGVs, humans) via three key pillars: perception, adaptation, collaboration enhanced by modules for a«social industrial environment» that ensures trustful, safe and inclusive user experience. More details on the technologies introduced will be presented in the following sub-sections. 3.1. Perception &data pipelines In current practice, autonomous operations rely on real-time sensor data for decision-making. However, the tools available for collecting, processing, and optimizing this data often require manual programming, leading to time-consuming and labor-intensive processes. Moreover, validating new changes in robotic systems is typically performed manually, resulting in prolonged deployment times. The CONVERGING approach aims to address these bottlenecks towards more streamlined and automated approaches in industrial settings.The CONVERGING perception pillar is structured into three modules (Figure 1) addressing gaps in existing systems. Figure 1. Perception involving data representation, storage and retrieval. Perception functionalities enable systems to interpret and effectively respond to their environment and hence are the 860 Nikos Dimitropoulos et al. / Procedia CIRP 128 (2024) 858–863 backbone of CONVERGING. The first module,"data at rest," establishes a digital thread for traceability across control systems and production resources, streamlining the management and orchestration of static information generated throughout the process flow. The "data in motion" module governs real-time data exchange, considering communication latencies crucial for timely actions on the shop floor. The "AI digital twin" module is a virtual replica of workstations to facilitate design iterations from conceptualization to deployment, while also integrating AI capabilities for the generation of key performance indicators (KPIs). CONVERGING is based on the OpenFlow reference architecture [18]. ROS will be used for data exchange during robot task execution or for robot task programming, whereas the potential of AutomationML for modelling the information related with the digital pipeline and the potential offered by OPC UA for modelling status information of production lines will be investigated. 3.2. Adaptation & AI Based Autonomy CONVERGING aims to implement a highly reconfigurable production system by deploying collaborative robotics and smart mechatronic devices, relying on multi-level AI to achieve autonomy (Figure 2). Figure 2. CONVERGING reconfiguration and AI-based autonomy. This will be achieved thought the development of modules, such as: i) “dynamic work reorganization” module, to balance the workload among the human and robotic resources, generating and simulating an number of alternative assignments of the tasks and selecting the scenario complying with the KPIs set, ii) “AI station controller”, responsible for orchestrating the production, monitoring the tasks execution progress and dispatching/distributing the follow up tasks when needed, iii) “perception and autonomy” module, responsible for the implementing updates based on the robots’perception capabilities to increase their autonomy. By integrating sensors installed at the shopfloor and directly on top of the robots, processing the relevant data in real-time using AI based computer vision techniques it is possible to adapt the robot behavior accordingly, iv) “collaborative robot control” module, that will provide path planning and online control of the robot capabilities, aiming to improve the ability of the robot to replan online to execute the assigned tasks, using direct/indirect human commands, existing resource models, simulation environments and data collected from the cell, v) “humanoid collaborative robot”, consisting of a two arms, responsible for the execution of collaborative tasks in close distance with the operator, working at simultaneously on the same part, assisting when needed like a human coworker, vi) “remote inspection robot”, responsible for the execution of tasks in hazardous and tight environments, allowing the operator monitor and control the execution safely, from a remote location, vii) “medium payload collaborative robot”, responsible for manipulating and holding heavy parts, presenting them to the operator in the appropriate way to perform additional tasks that require human dexterity, facilitating ergonomics, viii) “polishing robot”, responsible for the execution of polishing operators, equipped with the necessary sensors and algorithms. 3.3. Smart Human-Machine Collaboration A frequently reported drawback of HRC systems is the lack of flexibility to handle real-world uncertainties. The few industrial implementations allocate higher priority to the design and implementation of safety systems which often impose hard stops and dominate the operation. The control system must be equipped with greater intelligence and confidence to interpret various scenarios, such as individuals approaching the robot without the intention of contacting or interacting with it. CONVERGING focuses on software and hardware interfaces to ensure safe and seamless interaction with collaborative robotic solutions, minimizing learning curves and setup times. Three modules contribute to this goal (Figure 3): i) “safety assessment and monitoring”, responsible for monitoring the status and behaviors of the Safety Rated Solution/Control Systems (SRP/CS) of CONVERGING and allows the interface of the safety rated solution with other CONVERGING modules, ii) “teaching by demonstration”, enabling kinesthetic teaching of complex robot task, operating both off site and on site with the help of suitable user interfaces that can include VR and AR and are addressed even to novice in robotics users, iii) “autonomous robot behavior adjustment”allowing the adaptation of robot path and control strategies to improve task outcomes, including methods to update the human or task models with suitable parameters, either identified from data on the spot or provided by the AI station controller. Figure 3. Human Machine Collaboration and Social Industrial Environment. 3.4. Social - Industrial Environment Integrators lack of empirical knowledge and understanding for integrating user-centered requirements (i.e. subjective user Nikos Dimitropoulos et al. / Procedia CIRP 128 (2024) 858–863 861 experience and functional usability) in human-robot collaborative (HRC) systems while ensuring workforce trust, acceptance, comfort, and safety. Moreover, a widely accepted method for inclusiveness addressing the needs of individual human actors of different roles, backgrounds, skills, characteristics, etc. has not prevailed. Hence, CONVERGING aims to establish a human centered social-industrial environment where robot activities and interactions with humans are dynamically shaped to maximize user experience, trust, skills and safety. through the following three modules: i) “operator training”, responsible for training and upskilling Human Resources using state of the art technologies such as AR, and VR, ii) “user experience and ergonomics” which allows user-centered models and measurements to be considered in the robot strategy; these models may include factors related to human acceptance (e.g., individual preferences on speed/separation, or other parameters of the robot motion) or physical models which can evaluate the human ergonomics in a task specific way, iii) “multi-actor contextual interfaces”, providing a rich and flexible way that allows bi-lateral information exchange between human operators and the CONVERGING system, in a human-centric customizable manner via personalized interfaces. 4. Use cases The approach is tested and evaluated in four industrial scenarios inspired by the automotive, white goods, aeronautics and additive manufacturing sectors. Although the use cases derive from such diverse sectors, they share similarities in terms of the required technological solutions. 4.1. Autonomous tooling and inspection repair with robots Current State: This case derives from a stamping plant for car production. Single panels are created through stramping process by the use of different presses, dies and tools. Usually, a stamping process involves three main stations/operations: i) draw dies, responsible of making the main form of the panel, ii) trim dies (cutting the perimeter of the panel and performing different holes) and iii) flange/restrike dies (last operations responsible of reforming the panel and bending some flanges if needed). Focusing on the draw dies, they have to be perfectly polished with as low surface rugosity as possible to ensure compliance with the strict quality requirements of company. Currently, this process is manually performed by trained operators. However, this work is labor-intensive and induces ergonomic risks due to physically demanding postures and repetitive movements. Future Vision: CONVERGING envisages to increase the level of automation by introducing robots that will undertake the repetitive sanding task, while the operators can focus on visual quality inspection, reducing ergonomic risks and improving efficiency and performance of the stamping process. To ensure the proper functioning of the die, in the beginning an operator will inspect the die for any signs of wear or damage. This will be crucial as it will help to identify any potential issues that could impact the die's performance. If any wear or damage is detected, the operator will mark the area for repair, ensuring that the robot doesn't waste time repairing the wrong area. Once the damaged area is marked, the robot will take over and repair the die by polishing the marked area. This step help to restore the die's functionality and extend its lifespan, ensuring optimal performance and compliance with production requirements. Figure 4. Automotive production line. 4.2. AI enabled dexterous assembly of white goods Current State: Manufacturing of kitchen hobs involves highly heterogeneous and complex processes, including automated and manual assembly tasks. The attention will be focused on the hobs’ manufacturing line, in which the majority of the assembly work is manually performed. The factory does some production of individual components (gluing, enamel, pressing), but then they are assembled into their frame at the hob assembly line especially the electrical components for the heating elements. Future Vision: CONVERGING introduces a collaborative approach between a human and a dual arm humanoid robot. A human operator will load the coils into a feeding system, locating the terminals of the cables in a fixed position (e.g., spring-loaded clamp with referencing jig) to avoid uncertainty of coils feeding from the supplier boxes (Figure 4). Figure 5. White goods collaborative production station. The robot using a specially designed end-effector along with a smart multi-sensor system will confirm that the coil and connectors are correctly loaded. Then it will proceed with the assembly sequence, ensuring highly reliable screwing of the connectors to the terminals (torque and position), to guarantee that it will not be disconnected under nominal conditions with e.g., consistent pull test after assembly. AI methods for classification and anomaly detection will be used to monitor the assembly with force sensors and cameras, and human demonstrations or corrections will be requested when the perception system has low confidence. Human demonstrations (either through camera observation or kinesthetic teaching) allow the human to program the trajectory and correct robot 862 Nikos Dimitropoulos et al. / Procedia CIRP 128 (2024) 858–863 motion, while production and management will be capable of exchanging data and information thanks to the digital pipeline. The AI station controller will coordinate the Human and Robot, respecting the needs and mental state of the operator. 4.3. Improving Aircraft Fuel Tank Maintenance Current State: Aircraft fuel tank is a dangerous working environment. This makes the maintenance task very complex from a technical point of view, especially regarding the wing fuel tanks. There are several challenges in this maintenance procedure. Firstly, accessibility is limited due to related to the size of the entrance hole. Therefore, operators with specific body types are occupied for this job. Then, maneuverability inside the fuel tank is restricted due to the size of the tank itself. The technicians need to crawl inside the tank, which posed ergonomic risks. Special equipment should be used to eliminate the possibility of explosions due to any possible sparks. The safe duration a technician can remain in a fuel tank is limited to 30minutes leading to stress to timely complete the task. Finally, any task performed inside the tank requires two operators, one inside the tank and one supervising from outside. Future Vision: CONVERGING introduces new technologies and equipment for the inspection and maintenance of aircraft fuel tanks . A smart collaborative robot system with a flexible end effector able to extend will be used to inspect and perform several of the maintenance tasks when needed. The use of the appropriate “vision” sensors combined with AI based decisionmaking algorithms allows the detection of possible damages and identification of their characteristics (type, location, size). Figure 6. Aeronautics robotized inspection. Additionally, foreign object debris will be avoided with the use of anomaly detection methods, reviewed by the operator, before the fuel tank is reported as non-problematic. Data collected from each inspection will be used for perception improvements. This data will also be imported to an AI-driven suggestion system for most efficient repair distributions between the robot and the technician and suggestion for improvement of the fuel tank design. The HRC technology will increase the flexibility of the system as the transportation and configuration of the robot will be performed by the operator. Moreover, the operator will remotely monitor the robot’s actions and intervene when needed. Adaptive AR tools will facilitate HRI by providing workers with instructions, but also job verification and experts remote assistance capabilities. It is expected that exposure time in hazardous environmentswill be reduced thus ameliorating operator well-being. The cycle time of the process is expected to be reducedthanks to the smart robot, which will reduce the need of human intervention and the time for safety preparatory activities, respectively. 4.4. Enhancing Additive Manufacturing Post-Processing Current State: Additive manufacturing has attracted much attention due to its capability of producing parts with complex geometries. In most of the cases, supportive structures are needed on the parts during the manufacturing to enable the creation of several geometries. Therefore, post processing for support removal is necessary. This operation is currently performed manually by specialized operators. It is a difficult process as there is a high chance that the operator damages the part during the support removal as the tolerances are strict. Additionally, when the operation is performed manually, the remaining marks of the process are different on each part raising the surface finishing cost due to the high variability. The need for manipulation of parts of high weight and the difficulty of mounting them on milling machines due to their complex geometries are two additional issues of the current setup. Future Vision: Within CONVERGING, HRC and AI technologies will be implemented to the additive manufacturing use case. Robots will be used to initially blow the powder from the part after the additive manufacturing process, remove the part from the bed. Then they will assist with the final post-processing of the part in collaboration with a human operator. The robot will present it to the operator in the most appropriate way so that the operator can perform the necessary processing steps. The robot will identify the actions performed by the operator and adapt its pose accordingly in real-time to facilitate accessibility and improve ergonomics. AI and environment perception sensors (vision, force sensors) will be integrated to the robots to increase autonomy and safety for the operator during collaboration. The operator will have the ability to train the robot for the processing of different parts via hand-guiding and/or gestures and thus increase the variety of parts which can be handled. Using data gathered after processing each part the robot will be able to improve its precision and improve the process quality. The relevant data generated by the process will be stored in a database and retrieved in next iterations as needed. Figure 7. HRC processing of additive manufactured parts. 5. Results & Discussion The integration of emerging technologies as per the CONVERGING approach can radically change the production environments in manufacturing. Significant results are expected including faster, more robust perception capabilities Nikos Dimitropoulos et al. / Procedia CIRP 128 (2024) 858–863 863 leading to traceable and thus more trustful decision-making. Furthermore, ergonomics improvement will be improved both in terms of physical load and exposure to hazardous environments thanks to the allocation of dangerous and discomfortable tasks to robots, and in terms of cognitive load thanks to the advanced interfaces providing instructions and contact to the operators. Safety enhancements are expected thanks to advanced SRP/CS and dedicated operator training for familiarization with working in HRC environments and awareness for safety incidents. Finally, sensing, ai-algorithms, and robot control will ensure adaptation of robot behavior to operator preferences and characteristics improving user experience. Key Performance Indicators (KPIs) can quantify the expected benefits. Based on investigation for quantifying results in similar works [19] the following KPIs are suggested: •Reduction of production errors via AI and automation. •Reduction of average throughput time. •Improvement of end-product quality. •Reduction of material waste. •Reduction of rejection rate. •Reduction of non-value adding activities. •Reduction of operational costs related to manual labor. •Improvement of learning curves. •Improvement of ergonomics posture (e.g. RULA[20]) •Reduction of physical workload for parts handling •Improvement of employee’s satisfaction. •Reduction of monotonous tasks assigned to operators. •Reduction of human exposure to hazardous environments •Improvement of operator acceptance of HRC task. •Accessibility for operators with visual impairments. 6. Conclusion & Future work This paper presented the high-level design of a framework that will allow the formation of collaborative environments integrating AI, Big Data and Robotics for smart manufacturing. The enabling technologies were described as well as four indicative use cases where they will be applied, along with the expected impact and KPIs that enable to quantify the expected results. The technologies described in Section 3 are currently being implemented under the CONVERGING EU funded project and will be integrated into 4 pilots from the automotive, white goods, aeronautics and additive manufacturing sectors. Future work will focus on the development of the technologies discussed in Section 3, their optimization and their integration under a common production station. Moreover, future work will focus on the deployment of the developed production station in an industrial environment. This will allow to accurately measure the performance of the system as a whole and highlight any bottlenecks as well as demonstrate the importance of AI towards achieving advanced HRI. Acknowledgements This research has been supported by the Horizon Europe project “CONVERGING Social industrial collaborative environments integrating AI, Big Data and Robotics for smart manufacturing” (GA: 101058521) (www.convergingproject.eu, funded by the EC. The authors would like to express their gratitude to the CONVERGING consortium for the valuable information and assistance they have provided. References [1] Fowler, D. S., Epiphaniou, G., Higgins, M. D., & Maple, C. (2023). Aspects of resilience for smart manufacturing systems. Strategic Change, 32(6), 183-193.. [2] Chryssolouris, G., Alexopoulos, K., & Arkouli, Z. (2023). A perspective on artificial intelligence in manufacturing (Vol. 436). Springer Nature. [3] Natarajan, M., Seraj, E., Altundas, B., Paleja, R., Ye, S., Chen, L., ... & Gombolay, M. (2023). Human-robot teaming: grand challenges. Current Robotics Reports, 4(3), 81-100. [4] Dimitropoulos, N., Togias, T., Michalos, G., & Makris, S. (2021). Operator support in human–robot collaborative environments using AI enhanced wearable devices. Procedia Cirp, 97, 464-469. [5] Voulodimos, A., Doulamis, N., Doulamis, A., & Protopapadakis, E. (2018). Deep learning for computer vision: A brief review. Computational intelligence and neuroscience, 2018. [6] Ngiam, J., Khosla, A., Kim, M., Nam, J., Lee, H., & Ng, A. Y. (2011). Multimodal deep learning. In Proceedings of the 28th international conference on machine learning (ICML-11) (pp. 689-696). [7] Staar, B., Lütjen, M., & Freitag, M. (2019). Anomaly detection with convolutional neural networks for industrial surface inspection. Procedia CIRP, 79, 484-489. [8] Katsampiris-Salgado, K., Haninger, K., Gkrizis, C., Dimitropoulos, N., Krüger, J., Michalos, G., & Makris, S. (2024). Collision detection for collaborative assembly operations on high-payload robots. Robotics and Computer-Integrated Manufacturing, 87, 102708. [9] Ugarte, M., Etxeberria, L., Unamuno, G., Bellanco, J. L., & Ugalde, E. (2022). Implementation of digital twin-based virtual commissioning in machine tool manufacturing. Procedia Computer Science, 200, 527-536.. [10]Rohmer, E., Singh, S. P., & Freese, M. (2013, November). V-REP: A versatile and scalable robot simulation framework. In 2013 IEEE/RSJ international conference on intelligent robots and systems (pp. 1321-1326). IEEE. [11]Kumar, A., Pattanaik, L. N., & Agrawal, R. (2019). Optimal sequence planning for multi-model reconfigurable assembly systems. The International Journal of Advanced Manufacturing Technology, 100, 1719– 1730 [12]Evangelou, G., Dimitropoulos, N., Michalos, G., & Makris, S. (2021). An approach for task and action planning in human–robot collaborative cells using AI. Procedia Cirp, 97, 476-481. [13]Montreuil, V., 2007. Planning Human Centered Robot Activities. In 2007 IEEE Int. Conf. on Systems, Man and Cybernetics, 2618–23. [14]Spiliotopoulos, J., Michalos, G., & Makris, S. (2018). A reconfigurable gripper for dexterous manipulation in flexible assembly. Inventions, 3(1), 4. [15]Dimitropoulos, N., Togias, T., Michalos, G., & Makris, S. (2021). Operator support in human–robot collaborative environments using AI enhanced wearable devices. Procedia Cirp, 97, 464-469. [16]A. T. Angonese and P. F. F. Rosa, "Integration of People Detection and Simultaneous Localization and Mapping Systems for an Autonomous Robotic Platform," 2016 XIII Latin American Robotics Symposium and IV Brazilian Robotics Symposium (LARS/SBR), Recife, 2016, pp. 251256. [17]Schiffer, S., Ferrein, A., Bharatheesha, M., & Corbato, C. H. (2019). European Robotics Forum 2018 Workshop “Teaching Robotics with ROS”. [18]Koukas, S., Kousi, N., Aivaliotis, S., Michalos, G., Bröchler, R., & Makris, S. (2022). ODIN architecture enabling reconfigurable human–robot based production lines. Procedia CIRP, 107, 1403-1408.. [19]Lindberg, C. F., Tan, S., Yan, J., & Starfelt, F. (2015). Key performance indicators improve industrial performance. Energy procedia, 75, 17851790. [20]Arkouli, Z., Michalos, G., & Makris, S. (2022). On the selection of ergonomics evaluation methods for human centric manufacturing tasks. Procedia CIRP, 107, 89-94.