Towards Unobstructed Millimeter-Wave 6 G: Coordinated Multipoint Transmission Over a Fiber Wireless Distributed Antenna System.

1 Towards Unobstructed Millimeter-Wave 6G: Coordinated Multipoint Transmission over a Fiber Wireless Distributed Antenna System Argyris Kokkinis, Maria Vargemidou, Chris Vagionas, Kostas Siozios, Nikos Pleros, Amalia Miliou Abstract—Millimeter wave (mmWave) frequencies offer broad channel-bandwidths in 6G Radio Access Networks (RANs), but can be severely limited by channel outages and blockage phenomena in realistic non-line of sight field-environments. This work experimentally presents the first, outdoor 3×1 Fiber Wireless (FiWi) mmWave Distributed Antenna System (DAS) across 7 km fiber-distances and 30m wireless radio links, Software Defined Controlled (SDN) Coordinated Multipoint (CoMP) scheduled transmissions from three rooftop antennas to a mobile UE. The proposed system relies on wavelength-tunable transmissions through an all-passive, energy-efficient 32×32 AWGR and an algorithmic workflow that continuously monitors the FiWi link status, allowing for a scalable solution with even higher number of antennas and interfaces. The system is thoroughly evaluated in the physical layer, exhibiting up to 4.5 Gb/s data-rate at a25odegree beam-width and latency down to less than 0.2 ms. Continuous, SDN-controlled CoMP transmission is initially presented in pre-defined time-slots that alternate the three FiWi paths every 1s with high 14.4 dB isolation and 25 µs switching time. Unobstructed connectivity scenario is also experimentally presented, where an intentional mmWave blockage and link outage triggers a zero-touch re-routing and scheduling of upcoming traffic to the UE via an alternate path, selected in a round robbin manner. Finally, real-time streaming services and computation offloading of typical 6G applications at a nearby MEC server confirm the capability of the proposed system to provide high-availability services, shaping a promising roadmap for self-organizing 6G mmWave networks. Index Terms—6G, Optical Switching, mmWave, Radio over Fiber, SDN control, DAS, MEC I. INTRODUCTION In the era of 5G/6G, there is an increasing demand for ubiquitous broadband connectivity to several smart devices. 6G use cases span across a wide range of applications, from enhanced Mobile Broadband (eMBB) services, Augmented/Virtual Reality (AR/VR), Internet of Things/Everything (IoT/IoE), Industry 4.0 or Ultra-Reliable Low-Latency Communications (URLLC), often providing access to critical services or requiring computation offloading at nearby Multi-Access Edge Computing (MEC) servers [1]. This has been fueling a drastic surge of wireless traffic-capacities [2] with Key Performance Indicators (KPIs) of the 2030 vision of International Mobile Telecommunications (IMT) promoting extreme user data rates of 1 Gb/s anywhere in the Radio Access Network (RAN), user channels ≥400 MHz, reliable links with ≥99.999% availability and latency lower than ≤10ms. Millimeter-wave (mmWave) frequencies have been heralded as an enabling technology with large available spectrum, [3], yet they suffer from higher propagation loss. This has spurred rapid progress in mmWave antennas with directional beams of high gain and increasing network densification that places multiple access points closer to the User-Equipment (UE) Fig. 1: Schematic of the proposed 6G FiWi mmWave architecture with CoMP schemes providing unobstructed connectivity to mobile terminals and 6G applications. [4]. Co-locating multiple mmWave beams presents significant challenges [5], as it is necessary to properly steer the beam to the correct UE-location, handle the broader noise bandwidth, parasitic side lobes, out-of-band leakage of multi-antenna devices [6], [7], beam-misalignments, fading, channel-outages and blocking events [8]. Consequently, unwanted radiation from adjacent cells can cause Inter Cell Interference (ICI) and degrade Signal to Noise Ratio (SNR) [9] in realistic RANs, especially close to cell edges and multi-user environments [10], when the presence of various UEs and obstacles can cause frequent ICI and blockage phenomena [11], [12]. To address the connectivity-issue from a network perspective, multiple mmWave antennas and Remote Radio Head (RRHs) can be scattered around the field, building up Distributed Antenna Systems (DAS) or Distributed MultipleInput Multiple-Output (D-MIMO) Systems, adopting validated spatial mutliplexing techniques from Sub-6GHz networks. Strategically placing several antennas across the RAN service area allows radiating from different locations and transmission angles to the UE, as shown in the conceptual schematic of Fig. 1. Such mmWave DASs can offer ubiquitous, unobstructed real-time coverage with 360° field of view (FoV), provided that these are combined with flexible network coordination and Software Defined Control (SDN) of multiple paths to the UE, known as Coordinated MultiPoint (CoMP) transmissions or network-MIMO [13]. 2 TABLE I: State-of-the-Art Comparison in reconfigurable FiWi mmWave networks with CoMP functionalities. Ref. Technology Insertion Loss (dB) Channels Channel Rate (Gb/s) Radio (GHz) Distance (m) Polar. Depend. Indoor /Outdoor SDN control CoMP Real Time [14] MEM 8.4 4 2.5 – 4.3 Yes Indoor No No No [15] SiPho – 20 – – – Yes Indoor No No No [16] OXC – 8 20 60 – Yes Indoor No No No [17] SiPho 10–15 10 0.75 – – Yes Indoor No No No [18] OXC – 8 0.155 60 – Yes Indoor No No No [19] WSS 5.5 9 1.485 60 5 Yes Indoor No No No [20] SiPho 5 4 1 60 1 Yes Indoor No No No [21] SiPho 10.52 8 – – – Yes Indoor No No No [22] SiPho 12 12 10 – – No Indoor No No No [23] SiPho 18.5 24 50 – – No Indoor No No No [24] SiPho ∼15 8 100 – – No Indoor No No No [25] D-MIMO 15 4 – – 1.5 No Indoor No No No [26] D-MIMO 15.7 3 25 28 0.55 No Indoor No No No [27] DAS 10-15 8 4 28 3 No Indoor No No No [28] D-MIMO – 9 130 88 1 Yes Indoor No No No [29] DAS 15 2 60 – 2 No Indoor No No No [30] RRH-BBU - 4 50 – - No Indoor Yes CoMP-ST No [31] A-RoF 5 2 – 28 20 No Outdoor No CoMP-ST Yes [32] DAS - 4 48 24.2-29.5 20 No Indoor No CoMP-ST Yes [33] DAS - 4 – 40 20 No Outdoor No CoMP-ST Yes [34] RRH-BBU 7 32 10 – – No Indoor No CoMP-ST No [35] IFoF DAS 10 32 1 28 5.0 No Outdoor No CoMP-CJT No This work multi DAS 4 32 4.5 61-72 30 No Both Yes CoMP-ST Yes Introduced in 2006 [36], CoMP promotes continuous monitoring of the wireless-channel, reporting the Received Signal Strength Indicator (RSSI) or Channel State Information (CSI) to a centralized network-controller. In turn, the controller properly schedules transmissions, so as to ensure low ICI and circumvent channel-outages [13]. CoMP is classified into two categories [37]: i) CoMP Scheduled Transmission (CoMP-ST), or sometimes referred to as CoMP scheduled beamforming, where data traffic is transmitted via one antenna only at a time, while neighboring cells are silenced or steered towards different directions, minimizing ICI [38]; and ii) CoMP Coherent Joint Transmission (CoMP-CJT), where data traffic is transmitted simultaneously through multiple antennas [12], combining the different paths for constructive interference and enhanced SNR, provided the probabilistic channel response is accurately managed. After successful deployments in sub-6GHz networks, CoMP is being strongly considered in mmWave 5G/6G [13], although the challenges associated with intensive processing and power-hungry switching of broad mmWave radios remain to be addressed or implemented on efficient hardware. Specifically, CoMP-ST requires constructing a sparse, permutation matrix with the CSIs per time-slot between any of the Tx antennas and a single Rx UE antenna, selecting the optimum Fiber Wireless (FiWi) mmWave path and deactivating the other paths, prior to transmitting the data in the slot. In contrast, CoMP-CJT requires that the network controller is fully aware of the CSI of all FiWi paths at each time-slot to properly precode the signal constituents with the correct weights that cancel out ICI. Eventually, fully populated channel matrices must be constructed between all Tx antennas and the single Rx using CoMP-CJT [39], [40], achieving higher SNR gains. However, it imposes tight-synchronization requirements between the transmitting interfaces, appearing to this end as rather impractical for mmWave RANs. Lately, aiming to unleash multi-antenna mmWave connectivity, several optical transport and circuit-switching solutions are interfaced with multi-beam prototypes. Early profound demonstrations of N×N Optical Cross-Connects (OXCs) and 1×N Silicon Photonics (SiPho) Reconfigurable Optical Add/Drop Mutliplexers (ROADMs) in FiWi links have validated the high capacity credentials in transporting and routing up to 32 radio channels and 100 Gb/s channelrates, as summarized in TableI. Yet, most of these solutions focused solely on low-loss OXC prototypes, often with strong polarization dependence [14]- [21] and offline transmissions, routing traffic-links to different mmWave UEs of high-mobility or in pre-scheduled time-slots [16]- [23], and without considering any multi-point connectivity to a UE or possible linkoutages. Here, it is also worth noting that 6G is promising to deliver programmable network operations [41], reliable connectivity and high service availability in an automated fashion, eliminating human interventions and contributing to fast, autonomous, self-organizing RANs functioning at all outdoor conditions. Faced with the mmWave blockages and high ICI, recent global research efforts have steered focus towards developing practical CoMP schemes in mmWave and high carrier frequencies. CoMP-ST offers the anticipated reduced complexity, being the most suitable candidate solution for mmWave 6G. The first demonstrations of CoMP-ST over joint FiWi links for LTE signals focused on handling user channels as low as 100 MHz only [30]. Subsequently, CoMP-ST was rapidly 3 extended in mmWave prototypes, including carrier frequencies at 28 GHz [31], [32] and 40GHz [33], yet still presented only in indoor laboratory environments [31], [33], [34] and even anechoic chambers with well-controlled channel conditions [32]. However, recent advances in telecom equipment vendors, such as Ericsson and Keysight, enabled the implementation of a 4×1 CoMP-CJT FiWi transmission to a UE, yet user bandwidths are still as low as 100 MHz at carrier frequencies up to 3.5 GHz offline transmissions [42]. To the authors’ knowledge, outdoor trials of real-time FiWi mmWave DAS systems have been thus limited only to carriers up to 28 GHz, yet without any automated, SDN-controlled scheduling [35], rendering all FiWi CoMP solutions as simply, parallel operational links with independent Tx hardware and manual circuit switching operations, not achieving to truly showcase an outdoor working solution with automated SDN-controlled hand-overs for real-time elimination of link-outages. The current work builds on our recent SDN-coordination mechanism between an outdoor hybrid FSO/mmWave link [43] and the detailed theoretical investigations of CoMPST policies for alternating between known FiWi mmWave links [37], [44], to experimentally present for the first time outdoor FiWi CoMP-ST across an obstructed 3×1 mmWave DAS. Specifically, it is based on the data-plane of our recent demonstration of an outdoor, dual-band (O-/C-band), hybrid FSO/mmWave, static point-to-point link [45] towards a scalable 32-channel operation via an all-passive 32×32 AWGR, in order to handle also the mobility of users in a pointto-multipoint mmWave network architecture. Moreover, the control-plane and SDN-coordination workflow of our previous FSO/mmWave communication-resources allocation [45] is now extended to a unified SDN workflow that can also allocate different MEC-resources towards a joint management of FiWi and computing resources. Therefore, a complete dataplane and control-plane architecture of an end-to-end CoMP operation is experimentally presented for 5G/6G mobile users with diverse performance and power efficiency requirements, whose relative positions change dynamically. As conceptually depicted in Fig. 1, the proposed FiWi DAS relies on mmWave antennas operating at 61-72 GHz and scattered at three different rooftop locations of the university campus achieving a programmable, SDN-controlled, scheduled transmission of a 2Gb/s real-time traffic through 7km of Single Mode Fiber (SMF) and 30m radio distance towards a mmWave UE in the parking lot. Upon a degraded link-quality or outage, automated SDN handovers to alternate paths are triggered and coordinated via an overarching SDN workflow that reconfigures the FiWi path-connectivity via tunable wavelength transmitters and a 32x32 AWGR. Eyediagram, RSSI and antenna-radiation patterns are reported, revealing open eye diagrams and 25obeam width, while FiWipath switching times are measured to be 25 µs. Finally, uninterrupted, real-time connectivity is demonstrated by offloading and executing three different 6G UE applications at a nearby MEC server, including an eMBB, an IoT and a URLLC scenario, providing an end-to-end FiWi mmWave link latency of only 0.195 ms, confirming that the respective user data-rate and latency KPIs are met, thus shaping the technology towards unobstructed, broadband 6G with mmWave SDN-controlled, CoMP scheduled transmission capabilities. II. EXPERIMENTAL SETUP &DEVICES In order to showcase the developed FiWi mmWave CoMP scheduled transmission, an experimental testbed is set-up outdoors between neighboring building of the university-campus as illustrated in Fig 2(a), including three mmWave RRH antennas placed atop the roofs of two neighboring buildings and transmitting towards the final mmWave UE terminal placed in the parking lot, as shown in the mapping of Fig. 2(b). Targeting to serve various 6G networks applications, e.g. eMBB, URLLC or IoT/M2M, with varying computing requirements served by a MEC server, a single-board pod server with heterogeneous computing platforms is collocated at the Edge Box (EB) unit. The EB with the MEC server are shown in Fig. 2(c), integrating a Xilinx Alveo U200 acceleration card and an NVIDIA Tesla T4 GPU, connected to the server’s PCIe bus, in extend to the standard CPU. The Alveo U200 FPGA has a maximum power rating of 225W, contains 225K LookUp Tables, and offers an off-chip bandwidth of 77 GB/s [46]. On the other hand, the NVIDIA Tesla T4 has a maximum power of 70W and operates at a clock frequency of 585 MHz [47], while the MEC server’s CPU is an Intel Xeon Gold 5218R with 40 cores, running at 2.10 GHz. Collocating different platforms in the MEC server enables a selective execution of a 6G application to the best matching platform based on specific optimization criteria, such as energy, latency, or performance. In addition to hosting MEC platforms, the EB incorporates Mellanox ConnectX-5 NICs that implement the FiWi network control plane, providing three interface ports with tunable 10GbE SFP+ wavelength transceivers. The control plane selectively activates or deactivates any of the FiWi mmWave paths and reconfigures the optical connectivity to the UE terminal in the RAN based through proper management of the wavelength of the transceiver interfaces. The wavelength tuning operation is achieved by programming the EEPROM memory of the transceiver module, accessed via an i2c serial interface provided by the Mellanox NIC, according to the SFF8690 standard for the Tunable SFP+ memory map of the ITU frequencies [48]. The three optical X-haul interfaces of the EB are flexibly interconnected at up to 32 FiWi mmWave DAS locations. The optical outputs of the SFPs are each connected to a 7-km sMF spool with 1.7 dB attenuation, emulating realistic X-haul linkdistances, before being connected to three input ports of the 32×32 AWGR, shown in Fig. 2(c). The selected wavelengths emit at λ1=1544.7 nm, λ2=1549.5 nm and λ3=1554.9 nm, corresponding to the AWGR output ports 12, 18 and 25, while the average insertion losses of the AWGR for all possible combinations of the channels are found to be around 4 dB, with polarization dependence of less than 0.8 dB, indicating its suitability in optical X-haul transport networks and optical DAS solutions. In the wireless segment of the setup, a set of mmWave antenna units is deployed and connected to the AWGR in the 4 Fig. 2: a) Experimental setup of the 3×1 distributed FiWi mmWave X-haul CoMP, b) overall map of the outdoor setup placement, photos of c) the MEC server and AWGR, d),e),f) the three mmWave transmitter locations and g) the mmWave UE terminal. downlink direction. The antenna unit interfaces are equipped with SFP+ transceivers that up-convert the signal at frequencies between 61 GHz and 72 GHz to transmit it over the air up to 30m. After wireless transmission, the signal is received by a mmWave UE terminal, comprising an Rx antenna and a desktop PC with another Mellanox NIC (NIC # 2), requesting typical 6G application use-cases. Note that only the FiWi transmission in the downlink direction is demonstrated in the presented experimental setup, shown in Fig. 2. This setup incorporates all MEC and FiWi hardware components as well as the SDN CoMP control processes at the EB that handles real-time traffic exchange. The uplink transmission is not demonstrated in order to reduce the complexity of the experimental setup, as this would require implementing an identical SDN control agent at the uplink interfaces of the mmWave RRH units, in order to control the uplink transmission of the active antennas similar to the TWDM-PONs [49] To demonstrate the functionality of the FiWi DAS system in realistic 6G Use Cases, three characteristic applications are selected and developed for all three heterogeneous MEC platforms. The three examined applications are developed using Xilinx High-Level Synthesis (HLS) and Vitis 2023.1 [50], CUDA, and C for deployment on the FPGA, GPU and CPU platforms, respectively. The eigenface algorithm is trained offline on images of five individuals using the open-source ORL faces dataset [51] and executed on the three platforms for inference on a 0.5 MB dataset. The Savitzky-Golay filter is implemented based on the HLS and CUDA architectures detailed in [52] and executed to smooth 10 MB of temperature time-series data from temperature sensors. For the Sobel filter, images from the open source KTTI lane detection dataset [53] are used to detect edges in images captured by vehiclemounted cameras. III. SDN WORKFLOW &CONTROL PLANE To implement the FiWi DAS system with dynamic path selection and MEC platform reservation, a comprehensive SDN workflow for the CoMP scheduled transmission scenario is designed and integrated into the experimental setup, as detailed in this section. The SDN workflow developed consists of 16 sequential time steps, termed T0to T15 schematically represented in Fig. 3(a), allowing joint allocation of the communication and computing resources. Initially, the process begins at the time step T0 with the identification of user’s requested KPI, such as low latency or low power execution related to the 6G application use case, e.g. eMBB, URLL, IoT on the MEC server. In the current work, it is assumed that the task offloading decision has already been made by the user, and no optimization criteria is applied to determine which tasks should be offloaded. Consequently, upon receiving an application request, the SDN controller proceeds with resource reservation steps T1–T3, initiating requests for computing and communication resources and verification of the availability of the suitable computing platforms (CPU, GPU, FPGA) and the FiWi path, in alignment with 6G application optimization criteria and KPIs. Toward identifying the optimal computational platform in steps T2–T3, the SDN controller uses an execution metrics 5 Fig. 3: a) Algorithmic workflow for the allocation of the three processors of the MEC server and the 3×1 FiWi mmWave CoMP connectivity with simple Round Robin path selection, b) Stages of two parallel SDN processes for the control of the wavelength and selection of the MEC processor, c) The spectrum of the 3×3 AWGR and the three selected wavelengths for the multiple antenna transmission. Look-Up Table that contains indicative latency and power metrics for the three platforms and selects the one that best meets the KPI requirements for latency or power consumption. If the requested execution platform is unavailable or occupied by another task, a MEC conflict arises, shown in steps T10–T12 of Fig. 3(a), which is reported back to the SDN. The SDN controller resolves the conflict by selecting the next platform that fulfills the KPI. If neither the GPU nor the FPGA is available, the default CPU is used for execution. In parallel, to support communication resource monitoring and allocation, the mobile device periodically reports the status of the downlink channels to the edge box controller. The process that is illustrated as a feedback-loop in step T15, trasnferring the SDN control from step T9back to step T5, is intended to periodically assess the availability of the active FiWi CoMP channel selected in step T5. Similar to the MEC conflict resolution scheme, if a single FiWi mmWave link outage occurs-caused by any possible reason, for example, user beam rotation to a misaligned angle, user movement outside the beam coverage area, obstruction of the wireless link by an object or caused by signal quality lower than some selected threshold, such as a case that does not allow any data-traffic to be circulated due to high ICI—the status of the selected FiWi downlink channel is not reported and as a result, the channel is marked unavailable and malfunctioning. In such cases, the EB activates the next available FiWi path in the downlink direction towards the user, shown in steps T13–T14 relying on a round-robbin selection scheme. It queries the neighboring FiWi paths in a sequential manner, until it detects the first available one. Once the MEC compute platform and the FiWi channel are allocated, the requested application is executed, and the resulting output data is transmitted to the UE. This streamlined workflow ensures efficient resource allocation and seamless application execution. To implement SDN control in our experimental setup, a set of actions to ensure the continuous communication with the SDN agent are developed and its components are integrated in the EU, resulting in the workflow presented in Fig. 3(b). The SDN agent communicates through the NETCONF [54] protocol and interacts with the MEC server through two custom YANG model files. One model defines the computating platform, while the other configures the FiWi downlink topology by defining the required adjustment of the wavelength of the SFP+ transceivers and routing it via the AWGR. The open-source software tools Sysrepo [55] and Netopeer2 [56] are utilized for SDN management, with sysrepo serving as a database for storing the received NETCONF packets. The Netopeer2 client accesses Sysrepo to retrieve the YANG model .xml files that define the network topology. Note that the time required to reconfigure the FiWi paths and adjust the wavelength of the SFP+ transceivers depends on the time needed to detect the dropped path, to communicate with the SDN agent, and for the SDN agent to issue the new YANG configuration. The submission time of the new YANG configuration is programmable and was manually and arbitrarily set to 30 seconds for demonstration purposes in the experiments presented in Section IV. IV. EXPERIMENTAL FIELD TRIAL DEMONSTRATIONS To fully evaluate and demonstrate the FiWi mmWave DAS with CoMP features and real-time MEC allocation, we experimentally implemented and evaluated three typical challenging FiWi mmWave scenarios, as illustrated in the top, middle and bottom insets of Fig. 4, which respectively include: i) the characterization of the FiWi channels and the antenna radiation pattern with a static user rotating across 360oaround 6 Fig. 4: Conceptual schematic of the three experimentally demonstrated FiWi CoMP scheduled transmissions and realistic mmWave link-outage scenarios. itself, causing beam-misalignment and link-outage, ii) a linear movement of a mobile user with its antenna-beam constantly directed towards a 1D-arrangement of three Tx antennas with different area-coverage, evaluating also a pre-defined scheduled transmission in time-slots, and iii) a scenario with all three antennas directed towards the area of a single static user with a FiWi mmWave channel blockage, automatically triggering a dynamic re-routing of the traffic to the UE via an alternative FiWi path. A. Single FiWi mmWave channel with UE rotation Aiming to showcase the performance of a single FiWi mmWave channel of the proposed DAS system, we initially characterized the deployed AWGR device in terms of its spectrum and insertion losses, followed by an assessment of the radiation pattern of the utilized mmWave antennas. The AWGR characterization is performed using a tunable laser diode connected to one of its input ports, with the wavelength swept from 1540 nm to 1560 nm, while the output spectra are recorded at all of the AWGR’s output ports. The captured spectra for all 32 channel outputs were overlapped together and colored as shown in Fig. 3(c). Based on this figure, all 32 AWGR-channels seem to exhibit similar response, with an average insertion loss of 4 dB and a channel-bandwidth of around 0.3 nm. Featuring a channel spacing of 0.8 nm between two neighboring channels, the AWGR demonstrates clear channel separation, with crosstalk levels between -35 dB and -30 dB, supporting a 100 GHz spacing of the ITU grid, while the rest of the input ports 0° 22° 45° 68° 90° 112° 135° 158° 180° 202° 225° 248° 270° 292° 315° 338° 30 40 50 60 70 (a) Rx RSSI -80 -70 -60 -50 -40 -30 -20 -100 10 20 30 40 50 60 70 80 Angle (degrees) 0 2 4 Rx Bandwidth (Gbps) (b) Fig. 5: a) RSSI in dBm of the received signal at the Rx antenna when statically 360oacross its axis and b) the received signal bandwidth for the same setup. exhibit a similar behavior, effectively allowing to route and interconnect any of the input ports to any of its output ports, based exclusively on its wavelength. Having characterized the AWGR, we subsequently select three arbitrary wavelengths, namely λ1= 1544.7 nm, λ2= 1549.5 nm and λ3= 1554.9 nm, in order transmit data and route them to three different AWGR output ports. After the AWGR characterization, a static FiWi mmWave transmission is established across a 7km fiber distance and 10m wireless link-distance from the Tx antenna. On the other hand, the mobile terminal is placed on a portable rack, allowing rotation across a complete 360ocycle around its own axis, as depicted in the scenario at the top-inset of Fig. 4. Using the proposed setup, the mmWave power received by the portable Rx antenna terminal is measured at thirty different angles of rotation, allowing to represent the radiation pattern of the antenna, as shown in Fig. 5(a), where the maximum peak power of the RSSI is measured to be -31 dBm for the boresight angle with perfect alignment between the Tx and Rx beams. The 10 dB beam width of the main radiation lobe is around 25o(−10oto +15o), while some undesired sidelobes in the radiation pattern are also noticed, e.g., at −35o,−130oand in the opposite direction of −180o, which can cause degradation of the SNR to other neighboring UEs in the case of a 6G multi-beam hotspot scenario. The transmission data-rate from the Tx to the UE through different transmission angles was also evaluated within a sector 7 (b) (c) (d) (a) (b) (c) (d) 200 ps25 μs Fig. 6: a) RSSI of the received signal when the Rx antenna is moved perpendicular to the transmitted antennas, b) timetraces of the slotted downlink tranmission operation, c) risetime of the time traces and d) eye-diagram analysis. between −90oto +90owith a step of around +10ousing the iperf tool. The results are shown in Fig. 5(b), achieving a data rate between 4Gb/s to 4.5 Gb/s within the main lobe between −10oto +15o, and a user data-rate higher than 2 Gb/s within a broader −60osector from −30oto +30o. B. Linear Movement in 1D array The second testing scenario includes a mobile UE with a beam directed towards a 1D array of three parallel radiating Tx antennas, moving towards a path perpendicular to the beam direction, while the UE beam is being constantly directed towards the Tx antennas without any rotation or steering, as shown in the middle inset of Fig. 4. The three Tx antennas are positioned in a linear configuration inside a room at perpendicular coordinate-distance values of 0.8 m, 3.2 m and 6.4 m from respectively the closest wall of the room, thus being spaced as follows: 2.4 m between the 1st and the 2nd and 3.2 m between the 2nd and 3rd. The portable Rx antenna is being moved in a linear line, starting from the entrance door at the wall of the room with 0 m coordinate, i.e. before entering the coverage of the beam-width of the first Tx antenna, and moving across an 8 m distance in parallel with the arrangement of the Tx antenna, i.e. after exiting the coverage of the beamwidth of the third Tx antenna, while maintaining a constant mmWave link distance of 10 m. During this gradual movement, the RF power at the Rx antenna is recorded and plotted as shown in Fig. 6(a). The gradual movement results in having the Rx displaced by -0.8 m outside of the boresight axis of Tx #1, towards a perfect placement at the boresight axis of Tx #1 and within the radio coverage, and then again moving outside of the beam, with the RSSI ranging from -60 dBm up to -32.5 dBm, with the 10 dB beam-width radio coverage corresponding to 1.1 m. Equivalently, roughly equal peak RSSI values around -33 dBm and -31.5 dBm, with beam-widths covering a 1.4 m and 1.5 m are observed for the second and third Tx antenna radiations. Using this 3×1 FiWi mmWave DAS configuration, a periodic CoMP transmission is implemented over time for a duration of 6s, where the data traffic is directed to the three different FiWi downlink paths in slots with a duration of 1 s each. In order to evaluate its dynamic switching operation, the time-traces of the three slotted downlink transmissions are depicted in Fig. 6(b), with 4 s being allocated to path 1, and 1 s being allocated to paths 2 and 3, at the time windows between [2s, 3s] and [4s, 5s], respectively. The slots are used to characterize the switching time by capturing a zoomedin trace of their rise times, as shown in Fig. 6(c), revealing a switching-time of 25µs. Furthermore, characterization of signal quality includes an eye diagram analysis, as shown in Fig. 6(d), which exhibits an open eye with an extinction ratio (ER) of 6.9 dB and a Q factor of 9.7. These slotted transmission traces demonstrate clear On/Off isolation of at least 14.5 dB between the optically transmitted data traffic across different paths, attributed to the AWGR’s low crosstalk levels, thereby confirming the high quality of the CoMP transmission performance. C. Dynamic FiWi CoMP with zero-touch Tx scheduling Finally, a dynamic FiWi CoMP transmission scenario is operated on top of the 3×1 DAS, where a sudden signal degradation and blockage of an active, established path, automatically triggers a re-routing of the traffic to the UE, via an alternative path. At the beginning of this test, the Tx antennas are placed on the roof and on the windows of the 3rd and 2nd floors of two campus buildings, as shown in Fig. 2(d)- (f), while the Rx UE terminal is placed in the parking lot with a distance of 30 m. The Tx antennas are all directed towards the mobile Rx terminal, prior to any transmission. Blockage and link-quality degradation are emulated manually by introducing a scattering object into the link path, which reduces the received signal-power and thus degraded the SNR at the endpoint. This link-degradation is periodically detected by the SDN-agent through the link-quality monitoring loop associated with the time-steps T15 and T5of the workflow in Fig. 3(a), which automatically triggers a zero-touch path reconfiguration, scheduling the re-transmission of the dropped packets and the subsequent traffic of the FiWi downlink to flow to the same end-point UE via an alternative path. The operation of the dynamic CoMP FiWi transmission with path reconfiguration and handover to a neighboring mmWave 8 (b) (d) Outage #1 Restore Handover to Tx #1 Handover to Tx #2 (a) (c) Fig. 7: a) dynamic FiWi CoMP handover b) real-time video tranmission to the static Rx user c) latency of downlink link when 1,000 packets are transmitted and d) achieved tranmission bandwidth. TABLE II: MEC Platform Execution Metrics 6G-Apps⋆Metrics CPU GPU FPGA Eigen-faces Latency (ms) 57 1.6 35 Power (W) 15.4 26.9 12.5 Savitzky Golay Latency (ms) 20.3 0.5 1.8 Power (W) 23.5 28.4 15.3 Sobel filter Latency (ms) 25 0.02 3.3 Power (W) 21.6 26 12.5 ⋆Data sizes: 0.5 MB, 10 MB, 0.26 MB. Tx beam is evaluated through a time-trace of the downlink trasnmission, as captured by a 192 GS/s Real Time Oscilloscope (RTO). The time-trace is shown in Fig. 7(a), where the fast RTO is used to capture a 2min-long recording of the signal relying on a setting of ultra-small 5 MS/s sampling rate, that makes the recorded time-trace appear slightly ’blurry’ in the vertical axis, but still being clearly seen. The indicator-markers enlisted in the trace can map the channel outages with the steps T13 -T14 of the developed workflow. In this demonstration, the UE reported the status of the FiWi link every 30s, therefore the high signal quality at the marker-steps T13 -T14 of the FiWi and MEC reservation, are followed by a low-signal quality transmission, after interleaving the scattering object in the link at T13 -T14. After reporting the status to the controller, a dynamic CoMP mechanism triggers a rescheduling of the FiWi transmission, which is then reconfigured to use path 2, selected via a round-robin approach. After repeating steps T13 -T14 for path 2, the traffic is successfully restored to path 1. The FiWi DAS system transmission is finally benchmarked using iperf-traffic statistics, real-time video streaming, and realistic 6G MEC applications offloaded to the MEC server, to showcase uninterrupted operation. Traffic statistics collected via continuous ping measurements are shown in Fig. 7(c) using a 1 Kbyte ping packet size, revealing a normal distribution with an end-to-end latency of only 0.793 ms. Moreover, further latency and bandwidth measurements were performed also using the iperf tool, as shown in Fig. 7(d), which indicated a terminal bitrate of 2 Gb/s and minimum latency of 0.195 ms, which validate the low-latency transport credentials of the proposed solution. Finally, the FiWi DAS system is evaluated with real-time services, initially including video streaming services provided by the MEC server and displayed at the screen of the UE, as depicted in the photo of Fig. 7(b). An 8K-UHD video with 9-min duration is continuously streamed by the client’s VLC media player without any service interruption, log or lost video-frame, concluding to smooth media broadcasting. Finally, to showcase the capability of the FiWi DAS system to deliver realistic 6G applications to an end-user with uninterrupted, real-time MEC operations, three distinctive 6G applications are offloaded by the mobile UE terminal to the MEC server. Specifically, the UE terminal requested the execution of all three applications, namely the Eigenfaces for the eMBB scenario, the Savgol filter for the IoT scenario, and the Sobel video-filter for the URLLC scenario, to any of the three available computing platforms (CPU, GPU and FPGA) prior to transmitting the generated application-data output files to the mobile terminal. A table with the benchmarks of the three applications in terms of execution-latency and power consumption for each of the three platforms is shown in the Table II, validating the capability of the developed system to offload and execute real-time MEC applications on top of the FiWi DAS system. V. CONCLUSIONS The first demonstration of a 3×1-distributed FiWi mmWave DAS is experimentally presented with zero-touch SDNcontrolled mechanisms that can centrally coordinate and schedule multipoint transmission to a mmWave mobile UE. The physical layer metrics are thoroughly evaluated in terms of physical layer metrics, achieving an up to 4.5Gb/s data rate and latency down to 0.195 ms, while real-time MEC services of 6G applications and video streaming services confirm the capability of providing end-to-end connectivity and highavailability services, circumventing low signal qualities and mmwave blockage phenomena at cell edges and non-LOS environments, shaping a promising roadmap for 6G FiWi multi-antenna X-haul mmWave RANs. REFERENCES [1] Samsung Research, “6G: The next hyper-connected experience for all,” Samsung Research, White Paper, 2021, available online from Samsung Research. 9 [2] R. Singh, A. Kaushik, W. Shin, M. Di Renzo, V. Sciancalepore, D. Lee, H. Sasaki, A. Shojaeifard, and O. A. Dobre, “Towards 6g evolution: Three enhancements, three innovations, and three major challenges,” IEEE Network, 2025. [3] T. S. Rappaport, S. Sun, R. Mayzus, H. Zhao, Y. Azar, K. Wang, G. N. Wong, J. K. Schulz, M. Samimi, and F. Gutierrez, “Millimeter wave mobile communications for 5g cellular: It will work!” IEEE access, vol. 1, pp. 335–349, 2013. [4] N. Bhushan, J. Li, D. Malladi, R. Gilmore, D. Brenner, A. Damnjanovic, R. T. Sukhavasi, C. Patel, and S. Geirhofer, “Network densification: the dominant theme for wireless evolution into 5g,” IEEE Communications Magazine, vol. 52, no. 2, pp. 82–89, 2014. [5] L. Zhu, W. Ma, B. Ning, and R. Zhang, “Movable-antenna enhanced multiuser communication via antenna position optimization,” IEEE Transactions on Wireless Communications, vol. 23, no. 7, pp. 7214– 7229, 2023. [6] T. Sowlati, S. Sarkar, B. G. Perumana, W. L. Chan, A. Papio Toda, B. Afshar, M. Boers, D. Shin, T. R. Mercer, W.-H. Chen, A. Grau Besoli, S. Yoon, S. Kyriazidou, P. Yang, V. Aggarwal, N. Vakilian, D. Rozenblit, M. Kahrizi, J. Zhang, A. Wang, P. Sen, D. Murphy, A. Sajjadi, A. Mehrabani, E. Kornaros, K. Low, K. Kimura, V. Roussel, H. Xie, and V. Kodavati, “A 60-ghz 144-element phased-array transceiver for backhaul application,” IEEE Journal of Solid-State Circuits, vol. 53, no. 12, pp. 3640–3659, 2018. [7] A. M. Elbir, K. V. Mishra, S. A. Vorobyov, and R. W. Heath, “Twentyfive years of advances in beamforming: From convex and nonconvex optimization to learning techniques,” IEEE Signal Processing Magazine, vol. 40, no. 4, pp. 118–131, 2023. [8] A. V. Lopez, A. Chervyakov, G. Chance, S. Verma, and Y. Tang, “Opportunities and challenges of mmWave NR,” IEEE Wireless Communications, vol. 26, no. 2, pp. 4–6, 2019. [9] V. Petrov, M. Komarov, D. Moltchanov, J. M. Jornet, and Y. Koucheryavy, “Interference and sinr in millimeter wave and terahertz communication systems with blocking and directional antennas,” IEEE Transactions on Wireless Communications, vol. 16, no. 3, pp. 1791– 1808, 2017. [10] C.-N. Chen, Y.-H. Lin, L.-C. Hung, T.-C. Tang, W.-P. Chao, C.-Y. Chen, P.-H. Chuang, G.-Y. Lin, W.-J. Liao, Y.-H. Nien et al., “38-GHz phased array transmitter and receiver based on scalable phased array modules with endfire antenna arrays for 5G MMW data links,” IEEE Transactions on Microwave Theory and Techniques, vol. 69, no. 1, pp. 980–999, 2020. [11] J. Palacios, N. Gonz´ alez-Prelcic, C. Mosquera, T. Shimizu, and C.-H. Wang, “A hybrid beamforming design for massive MIMO LEO satellite communications,” Frontiers in Space Technologies, vol. 2, p. 696464, 2021. [12] G. Nigam, P. Minero, and M. Haenggi, “Coordinated multipoint joint transmission in heterogeneous networks,” IEEE Transactions on Communications, vol. 62, no. 11, pp. 4134–4146, 2014. [13] G. R. MacCartney and T. S. Rappaport, “Millimeter-wave base station diversity for 5g coordinated multipoint (comp) applications,” IEEE Transactions on Wireless Communications, vol. 18, no. 7, pp. 3395– 3410, 2019. [14] B. Schrenk, T. Lor¨ unser, and T. Zemen, “Towards spectrumprogrammable, mesh-enabled mobile xhaul through reconfigurable wdm overlay in fully-passive networks,” in 2015 17th International Conference on Transparent Optical Networks (ICTON). IEEE, 2015, pp. 1–4. [15] M. S. Dahlem, C. W. Holzwarth, A. Khilo, F. X. K¨ artner, H. I. Smith, and E. P. Ippen, “Reconfigurable multi-channel second-order silicon microring-resonator filterbanks for on-chip wdm systems,” Optics express, vol. 19, no. 1, pp. 306–316, 2010. [16] G.-K. Chang, A. Chowdhury, Z. Jia, H.-C. Chien, M.-F. Huang, J. Yu, and G. Ellinas, “Key technologies of wdm-pon for future converged optical broadband access networks,” Journal of Optical Communications and Networking, vol. 1, no. 4, pp. C35–C50, 2009. [17] C. Browning, Q. Cheng, N. C. Abrams, M. Ruffini, L. Y. Dai, L. P. Barry, and K. Bergman, “A silicon photonic switching platform for flexible converged centralized-radio access networking,” Journal of Lightwave Technology, vol. 38, no. 19, pp. 5386–5392, 2020. [18] T. Kuri, H. Toda, J. J. V. Olmos, and K.-i. Kitayama, “Reconfigurable dense wavelength-division-multiplexing millimeter-waveband radio-over-fiber access system technologies,” Journal of Lightwave Technology, vol. 28, no. 16, pp. 2247–2257, 2010. [19] Z. Jia, “Wireless high-definition services over optical fiber access networks,” in 2009 18th Annual Wireless and Optical Communications Conference. IEEE, 2009, pp. 1–4. [20] A. Tsakyridis, E. Ruggeri, G. Kalfas, R. M. Oldenbeuving, P. W. L. van Dijk, C. G. H. Roeloffzen, Y. Leiba, A. Miliou, N. Pleros, and C. Vagionas, “Reconfigurable fiber wireless ifof fronthaul with 60 ghz phased array antenna and silicon photonic roadm for 5g mmwave crans,” IEEE Journal on Selected Areas in Communications, vol. 39, no. 9, pp. 2816–2826, 2021. [21] D. Wu, Y. Wu, Y. Wang, J. An, and X. Hu, “Reconfigurable optical add-drop multiplexer based on thermally tunable micro-ring resonators,” Optics Communications, vol. 367, pp. 44–49, 2016. [22] V. Sorianello, G. D. Angelis, T. Cassese, M. V. Preite, P. Velha, A. Bianchi, M. Romagnoli, and F. Testa, “Polarization insensitive silicon photonic roadm with selectable communication direction for radio access networks,” Optics Letters, vol. 41, no. 24, pp. 5688–5691, 2016. [23] G.-H. Chen, C.-W. Peng, M.-W. Cheng, P.-C. Guo, J.-F. Tsai, Y. Tong, C.-W. Chow, C.-H. Yeh, and H. K. Tsang, “Silicon-photonics based remote-radio-head using mode and wavelength division multiplexing with capacity of 4.781 tbit/s for radio-over-fiber massive mimo,” in 45th European Conference on Optical Communication (ECOC 2019). IET, 2019. [24] S. Nakamura, S. Yanagimachi, H. Takeshita, A. Tajima, T. Hino, and K. Fukuchi, “Optical switches based on silicon photonics for roadm application,” IEEE Journal of Selected Topics in Quantum Electronics, vol. 22, no. 6, pp. 185–193, 2016. [25] F. Olofsson, L. Aabel, T. Eriksson, M. Karlsson, and C. Fager, “Sigmadelta-over-fiber with wdm serial connection for distributed mimo,” Journal of Lightwave Technology, 2024. [26] H. Bao, F. Ponzini, and C. Fager, “Wideband mm-wave 6 x 2 distributed mimo transmitter using sigma-delta-over-fiber,” Journal of Lightwave Technology, vol. 42, no. 9, pp. 3107–3117, 2024. [27] J. Kim, M. Sung, S.-H. Cho, Y.-J. Won, B.-C. Lim, S.-Y. Pyun, J.-K. Lee, and J. H. Lee, “Mimo-supporting radio-over-fiber system and its application in mmwave-based indoor 5g mobile network,” Journal of Lightwave Technology, vol. 38, no. 1, pp. 101–111, 2019. [28] P. T. Dat, F. Rottenberg, A. Kanno, N. Yamamoto, and T. Kawanishi, “3 x 3 mimo fiber–wireless system in w-band with wdm/pdm rof transmission capability,” Journal of Lightwave Technology, vol. 39, no. 24, pp. 7794– 7803, 2021. [29] S.-R. Moon, M. Sung, E.-S. Kim, J. K. Lee, S.-H. Cho, and J. Kim, “Rof-based indoor distributed antenna system that can simultaneously support 5g mmwave and 6g terahertz services,” Optics Express, vol. 30, no. 2, pp. 1521–1533, 2022. [30] J. Zhang, Y. Ji, H. Yu, X. Huang, and H. Li, “Experimental demonstration of fronthaul flexibility for enhanced comp service in 5g radio and optical access networks,” Optics express, vol. 25, no. 18, pp. 21 247– 21 258, 2017. [31] NTT DOCOMO, Inc. (2022) World’s first successful demonstration of distributed mimo that continues wireless connections in the 28 ghz band by eliminating shielding issues. [Online]. Available at: https://group.ntt/ en/newrelease/2022/10/31/221031a.html. [32] A. Moerman, J. Van Kerrebrouck, O. Caytan, I. L. de Paula, L. Bogaert, G. Torfs, P. Demeester, H. Rogier, and S. Lemey, “Beyond 5g without obstacles: mmwave-over-fiber distributed antenna systems,” IEEE Communications Magazine, vol. 60, no. 1, pp. 27–33, 2022. [33] NTT DOCOMO, Inc. (2023) Successful 40 ghz band distributed mimo demonstration of maintaining radio communication quality even when interference occurs due to movement of multiple users. [Online]. Available at: https://www.docomo.ne.jp/english/info/media center/. [34] J. Zhang, Y. Ji, S. Jia, H. Li, X. Yu, and X. Wang, “Reconfigurable optical mobile fronthaul networks for coordinated multipoint transmission and reception in 5g,” Journal of Optical Communications and Networking, vol. 9, no. 6, pp. 489–497, 2017. [35] M. Sung, J. Kim, E.-S. Kim, S.-H. Cho, Y.-J. Won, B.-C. Lim, S.-Y. Pyun, J. K. Lee, and J. H. Lee, “5g trial services demonstration: Ifofbased distributed antenna system in 28 ghz millimeter-wave supporting gigabit mobile services,” Journal of Lightwave Technology, vol. 37, no. 14, pp. 3592–3601, 2019. [36] M. Karakayali, G. Foschini, and R. Valenzuela, “Network coordination for spectrally efficient communications in cellular systems,” IEEE Wireless Communications, vol. 13, no. 4, pp. 56–61, 2006. [37] M. Gatzianas, G. Kalfas, A. Mesodiakaki, C. Vagionas, and N. Pleros, “Traffic-aware coordinated beamforming for mmwave backhauling of 5g dense networks,” IEEE Transactions on Wireless Communications, vol. 21, no. 7, pp. 5019–5034, 2021. [38] Y. Liu, J. Lee, and W. Chen, “Buffer-aware scheduling and power allocation for comp transmission with large-scale antennas,” IEEE Transactions on Wireless Communications, vol. 21, no. 6, pp. 4168– 4182, 2021. [39] H. Chergui, A. Ksentini, L. Blanco, and C. Verikoukis, “Toward zerotouch management and orchestration of massive deployment of network