5G Perspective Of Connected Autonomous Vehicles: Current Landscape and Challenges Toward 6G

Connected and autonomous vehicles (CAVs) have emerged as a promising paradigm in the automotive industry, revolutionizing transportation systems with their potential to improve safety, efficiency, and overall user experience. The communication aspect of CAVs is vital to enable their advanced capabilities. In this article, we focus on analyzing the current landscape, challenges, and future vision of CAVs from a 5G perspective. Unlike previous theoretical studies, our approach is grounded in empirical evidence obtained through trials conducted over a commercial 5G deployment, considering two representative use cases: “Remote Driving” and “Quality of Service (QoS)-based Level-of-Automation Selection.” Based on the quantitative findings, we identify and discuss several open challenges that need to be overcome to seamlessly integrate CAVs into the 5G ecosystem. Furthermore, we outline prospective enablers for solutions and enhancements that can address the identified shortcomings and provide a solid foundation for the future of CAVs in the context of 6G. Overall, this research contributes to the ongoing efforts to optimize communication systems for CAVs, facilitating their safe and efficient integration into the future smart transportation ecosystem.


IntroductIon
The last decade has witnessed a great evolution in the technology of Connected and Autonomous Vehicles (CAVs), with several companies and governments investing heavily in their development and adoption [1].Key advancements and trends in the current state of CAVs include: • Connectivity: CAVs rely on advanced communication technologies to connect with other vehicles, infrastructure, and the cloud.This allows for sharing data, coordinating movements, and receiving real-time updates on environmental conditions.Undoubtedly, the deployment of 5G networks has further enhanced CAVs' connectivity.• Automation: CAVs use various sensors, cameras, and Machine Learning (ML) algorithms to per-ceive their surroundings, make decisions, and execute maneuvers without human intervention.
There are currently five levels of vehicle automation, with Level 0 corresponding to no automation and Level 5 to full automation [2].• Testing and Deployment: Several companies and governments are testing CAVs in various real-world scenarios, such as ride-sharing, delivery services, and public transportation as a step toward transforming the way we travel and transport goods by building on the emerging lessons learned.Nevertheless, the widespread deployment of CAVs is still in its infancy, and there remain several technical, regulatory, and societal challenges that need to be addressed.
From a technical perspective, C-V2X (Cellular Vehicle-to-Everything) and 5G/6G mobile communications are crucial technologies for enabling the full potential of CAVs, including improved safety (e.g., faster detection of potential hazards and quicker response times), increased efficiency (e.g., optimizing traffic flow and reducing congestion), and new mobility services (e.g., ride-sharing, delivery services, and immersive entertainment experiences).5G has made substantial progress in developing a reliable, low-latency access network by creating new pathways for advanced spectrum and radio resource management and rethinking the core network's orchestration.5G New Radio (NR) secures high data rates and extended capacity leveraging on a large spectrum of wide bandwidth.The combination of the high-band spectrum with advanced antenna technologies, for example, Multiple-Input Multiple-Output (MIMO), allows service provisioning to a large number of vehicles.At the same time, lower frequency bands guarantee coverage in Non-Line-of-Sight (NLOS) communication scenarios [3], often found in dense urban environments.Finally, the softwarized 5G core has been designed to deliver consistently low response times by creating and managing slices throughout a contiguous virtualized infrastructure, ameliorating the network's latency and reliability [4].
Regrettably, though, most of the existing com-mercial 5G deployments operate in 5G NSA (Non-Standalone) mode [5], a preliminary deployment phase to 5G SA (Standalone) mode that does not fully integrate the aforementioned technological enablers.Actually, the 5G NSA deployment option involves using the legacy 4G LTE core network (EPC, Evolved Packet Core) for control plane functions while integrating 5G NR Radio Access Network (RAN) in the existing 4G base stations, such that user plane data traverse both 4G and 5G networks.5G NSA cannot fully realize the potential of 5G networks, introducing latency in the communication between 5G devices and the 4G core.Additionally, advanced features, such as network slicing, are limited or not fully optimized, restricting the expected gains in latency and reliability.Moving beyond this interim deployment phase, even fully standalone 5G networks are expected to eventually be pushed to their limits by the increasing scale and pervasiveness of computation and data, becoming inadequate to achieve the necessary data rate, latency, and number of connections per km2 [6].Thus, the next generation of 6G mobile communications will be required to provide the scalability, flexibility, and optimized performance mandated by the advanced applications and use cases of CAVs.Motivated by the above, in this article, we focus on the communications aspect of CAVs and review the current landscape and open challenges of CAV applications by capitalizing on recent trials over a representative commercial 5G deployment.Unlike previous works that study and analyze the advancements in CAVs from theoretical viewpoints [7], our approach is grounded on empirical evidence obtained through trials conducted over Vodafone's, Germany, 5G NSA network considering two diff erent tangible use cases: "Remote Driving" and "Quality of Service (QoS)-based Level-of-Automation Selection."First, data-driven analysis is performed based on the ideal conditions and network requirements that guarantee the feasibility and eff ectiveness of the two examined applications and the actual performance metrics measured during the conducted validation trials.The performance discrepancies observed between the two are then extrapolated to determine the technological inadequacies that must be addressed to close the gap, providing substantive feedback before entering the new era of 6G mobile communications.Therefore, our scope extends across all recent phases of mobile communications evolution, encompassing 5G NSA, 5G SA, and contemplating the theoretical landscape of 6G.
The remainder of this article is organized as follows.First, we provide an overview of the two CAV applications considered, along with the necessary network requirements and desired optimal conditions.The validation trials are then elaborated, and performance gaps are defined.Finally, emerging key enablers are identifi ed and discussed to bridge the observed divide, especially as we move forward to the foundation of 6G systems and platforms.

ILLustrAtIng the potentIAL of cAvs: tWo tAngIbLe use cAses
In the context of CAVs, two fundamental use cases are remote driving and the dynamic selection of automation levels based on the perceived QoS.These use cases can be seen as complementary extensions, collectively supporting a wide spectrum of CAV autonomy, ranging from remote control and assistance to fully autonomous operation.A remote driver can control a Level 4 autonomous vehicle (e.g., semi-autonomous vehicle [2]) equipped with sensors and cameras, while the QoS-based LoA selection adjusts the vehicle's reliance on remotely-processed assistance or comfort features according to the network connection quality.A transition scenario from remote to assisted driving is depicted in Fig. 1, along with the constituent elements of the end-to-end network and the vehicle's hardware to allow for the support of the two services.Within the context of autonomous driving, remote driving can serve as an interim measure and a backup solution to full automation, fostering improved safety and cultivating public trust.Moreover, the teleoperation of heavy machinery, drones, and industrial vehicles can offer precise control in hazardous environments or even disaster-stricken areas while revolutionizing vertical sectors, such as mining and construction.In urban environments, remote driving can facilitate services like last-mile delivery and valet parking, whereas, in suburban areas, the teleoperation of drones has a multifold contribution to precision agriculture and wildfire detection and monitoring, to name just a few examples.
Ideal Conditions and Requirements: To measure the quality of the remote driving service offered, a set of Key Performance Indicators (KPIs) has been defined and used in 5G networks [2].The probability of successfully transmitting data packets and the number of correctly received bits within a period are captured by reliability and throughput metrics.Given the load burden and criticality of timely video transmission in the uplink from the vehicle to the remote location, the metric of Uplink (UL) jitter is of particular interest.Conversely, regarding the timely downlink control-commands transmission for proper maneuvering, the achieved Downlink (DL) latency is measured, assessing the performance of the network including wireless transmission and core network processing.It should be noted that the remote driver's reaction time and the translation of the control commands to appropriate signals for the vehicle's actuators are not relevant to the particular DL latency metric.Finally, the maximum vehicle speed at which the optimal conditions can be achieved is expressed by a speed-related mobility metric.
The specific ideal conditions and requirements that have been defined by the latest standardization and research activities are listed as follows [8,9]

Qos-bAsed LeveL-of-AutomAtIon seLectIon
Description: QoS-based LoA selection refers to the dynamic adjustment of the functional scope of the available technologies or features in a given vehicle based on real-time monitoring of the underlying network conditions.This is particularly useful for arbitrating vehicle reliance on remotely-processed assistance or comfort features and, hence, determining the driving mode.In this way, the continued provision of features outside of a vehicle's Operational Design Domain (ODD), for example, remote perception, route optimization, and other edge or cloud-based applications, can be sustained instead of simply reducing the LoA and forfeiting these features.
In greater detail, upon reaching the end of the ODD, a client application hosted on the vehicle requests active support for maneuvering from a remotely deployed service that can handle tasks like object detection in incoming video streams.The service accepts the request depending on the expected QoS over the planned trip and continues monitoring the connection throughout the offered service.In case the link quality degrades, the client is prompted to switch to a lower LoA.The client and the server communicate over two distinct channels: control (initial negotiation and QoS monitoring) and video.In this way, the vehicle can intelligently toggle between LoAs in urban mobility and mixed traffic scenarios to ensure a safe ride and smoother interaction with human-driven vehicles.
Ideal Conditions and Requirements: Critical KPIs for this use case are End-to-End (E2E) latency, reliability, and throughput.E2E latency measures the time it takes to transmit packets between the vehicle and the remote service provider in both directions (UL/DL).At this point, the E2E latency of both the control and data transmissions is considered, and the performance of the network is assessed, including the wireless transmission part and the processing at the core.The signal processing times on both endpoints are excluded.Continuing, UL throughput is critical for ensuring the timely transmission of video streams, and hence timely detection of objects by the service.Reliability is equally vital in ensuring safety since lost packets give rise to incomplete state or control information and potentially lead to the complete abortion of the service.Finally, mobility relates to the ability of the network to support UEs at different speeds.
Based on the analysis of the QoS-based LoA selection use case for an auxiliary perception service, the following targeted KPIs values were defined [8] The required software for each use case was deployed on two Dell Optiplex-7070 with Intel i9-9900 3.10 GHz 8-core CPUs and 32 GB RAM -one for each of the communication endpoints between the vehicle and the corresponding remote service -along with a uBlox EVK-M8T GNSS device for clock synchronization.
Specifically, the developed service related to the remote driving use case has two endpoints that communicate over the network: the Remote Operations Center-Gateway (ROC-GW) node residing in the vehicle and the Remote Operations Center (ROC) GUI application at the remote location, in Chemnitz, Germany, which is approximately 36 km away from the test site in Schlettau.The former aggregates the data published by the OBU and coordinates the communication between the vehicle and the ROC, whereas the latter acts as an interface with the remote driver [10].Similarly, the service developed for the QoS-based LoA selection use case comprises a QoS-client at the vehicle's side and a QoS-server located also in Chemnitz.The QoS-client node is responsible for requesting assistance from the service running at the QoS-server.The acceptance of the request is based on an apriori Reference Signal Received Quality (RSRQ) mapping of the trial area, while both endpoints are responsible for the live monitoring of the network conditions once a connection is established.
The remote driving use case experiments were carried out over five weekdays during working hours, with the vehicle moving at a speed of up to 25 km/h.The distance between the vehicle and the base station ranged from 60 m to 120 m, while a varying Reference Signal Received Power (RSRP) between -62 dBm and -87 dBm was achieved due to NLOS propagation and coverage holes, causing the 5G signal to degrade.To evaluate the operation of the remote driving service under real-world conditions of different precision requirements, the following four test cases were defined and tested, depending on the maneuver performed in a controlled open-space environment with varying network conditions: • Straight maneuver (TC1) • Turn-right maneuver (TC2) • Lane-change maneuver (TC3) • Parking maneuver (TC4).
The tests focused on measuring the mean latency, mean throughput, jitter, and loss rate for each data stream exchanged between the ROC and the ROC-GW node installed within the vehicle.To this end, the incoming and outgoing traffic at the ROC-GW and ROC measurement points were captured and post-processed [11].
Contrary to remote driving, the QoS-based LoA selection use case allows for larger-scale experimentation, and for this purpose, an approximately 11 km test area was used.This resulted in a distance of 60 m to 2.5 km between the vehicle and the base station, while the vehicle's speed ranged up to 70 km/h, as imposed by the public roads regulation.A single test case was considered this time, according to which a vehicle performing its route is about to exit its ODD and requests assistance from the QoS-based LoA selection service.If the request is accepted, the video stream from the vehicle's front camera is transmitted from the QoS-client to the QoS-server endpoint.The video data are processed and routing information is transmitted back to the vehicle along with a stream of detection information for its assisted maneuvers.Network quality monitoring is initiated from the request acceptance until the completion of the assisted maneuver, indicating the vehicle's fall back into a reduced LoA in the event of link degradation.The tests focused on measuring the E2E latency, throughput, and loss rate for each data stream exchanged between the two QoS endpoints by post-processing the relevant network traffic captures.

performAnce gAps
The performance of the developed remote driving service is summarized in Table 1, which contains the measured KPIs for five iterations of each test case.Similarly, Fig. 2 illustrates the performance gaps identified by comparing the optimal and mean measured KPI values over the five iterations of the conducted trials.The range of values of the radial axes has been determined based on the minimum acceptable and optimally achievable values [8,9].However, in some KPI cases (e.g., mobility, reliability) this range has been further modified to facilitate better visualization.The numerical results verified that 5G NSA deployments support low-speed remote driving scenarios.In these instances, the effect of jitter is crucial, directly impacting the stability and consistency of the remote driver's interaction and control over the vehicle.Figure 2 demonstrates that a near-optimal performance in terms of jitter can be achieved, particularly enabling applications that involve haptic control, such as remote machinery operation.Nevertheless, a non-negligible performance gap in the achieved DL latency and UL throughput is identified.The significant deviation of the measured DL latency from the optimal value can be attributed to the inter-working of 5G NR with the legacy 4G LTE core, adding delays during the execution of core network functions.On top of that, the non-optimized RAN deployment that has been inherited from 4G LTE networks is further prohibiting the achievement of high throughput while using 5G NR frequency bands, abruptly reducing the RSRP due to obstacles or limited coverage in general.Regarding the achieved levels of reliability, a fair performance is observed, reaching approximately a percentage of 98 percent which, however, refers to the lowspeed use case scenario and has the potential to be ameliorated by full SA network deployments.
Table 2 presents the measured KPIs values for five representative iterations of the QoS-based LoA Selection use case, while their mean values are further illustrated in the form of a radar chart in Fig. 3 to indicate the performance gaps arising.Overall, during the trials, the developed service was shown to successfully accept or reject incoming assistance requests according to a static QoS estimate based on an apriori RSRQ mapping and provide an early indication of link degradation.Nevertheless, a sig-The significant deviation of the measured DL latency from the optimal value can be attributed to the inter-working of 5G NR with the legacy 4G LTE core, adding delays during the execution of core network functions.nificant performance disparity can be observed between the optimal and the achieved KPIs values under a 5G NSA network deployment, especially regarding the measured E2E latency and UL throughput.This behavior is consistent with the results of the remote driving use case and can be attributed to the factors analyzed earlier.
brIdgIng the dIvIde And roAd to 6g: emergIng Key enAbLers Two pillars are identified that could enable bridging the divide between the currently achieved and the optimal performance required by CAV use cases: rethinking the architecture of the core and radio access network under the anticipated 6G paradigm with a particular focus on the seamless integration of Artificial Intelligence (AI); and introducing storage and intelligence across the endto-end network.The former concerns exploiting the full capabilities of 5G technology and additionally evolving them toward the sixth generation.This will require transitioning from 5G NSA to SA deployments and adopting AI capabilities directly into the infrastructure and devices of 6G networks (Fig. 4).The latter regards the provisioning of beyond pure communication functions, such as sensing and perception [12], which are envisioned to be seamlessly delivered by the 6G technology enablers under a novel architectural approach and end-to-end system design currently being developed within the context of the Hexa-X-II project (https://hexa-x-ii.eu/).

ArchItecturAL Aspects
5G Core: The 5G Core (5GC) represents the embodiment of 5G technology, enabling reliable, secure, and ultra-low latency communications, and is the distinctive difference between 5G NSA and SA deployments.This aspect is particularly critical for time-sensitive applications, such as those associated with CAVs.The core network handles several essential functions that are entirely softwarized, and their design is agnostic of the underlying network infrastructure, allowing for flexible and on-demand routing that can adapt to the time-varying nature of road traffic.This capability of the 5GC network is further materialized by network slicing, based on which the physical network infrastructure can be logically partitioned into multiple virtual networks to account for the distinct needs of CAVs (e.g., high-bandwidth infotainment, time-critical collision avoidance) in a customized and isolated way.In addition to the above, 5GC supports Multi-access Edge Computing (MEC) architecture, bringing computing and storage resources closer to the network's edge, where data is generated and consumed.Consequently, location-aware requests (e.g., 3D road map creation) can be instantaneously processed, contributing to closing the gap between the currently achieved and the optimal performance in terms of latency.
Native AI in 6G: The integration of AI technology for enabling the comprehensive automation and optimization of the network and service management as part of the 6G network architecture will further contribute to closing the existing performance gaps.With the deployment of distributed agents, AI native control mechanisms will be introduced into the network, providing a multitude of benefits, including optimization of network slicing operations, intelligent resource allocation, and prioritization of energy-efficient operations.Real-time network slicing configurations will be optimized by automating the analysis of network traffic patterns, service demands, and user behavior.Additionally, AI algorithms will support closed-loop control across various scenarios, such as predictive orchestration and autonomous resource management, ensuring real-time responsiveness aligned with the 6G performance targeted KPIs, while security and privacy measures will be also strengthened.Particularly, in the context of CAVs, the aforementioned AI algorithms and control mechanisms will enable seamless and efficient communication between CAVs and the surrounding infrastructure, delivering ultra-low latency and high reliability, while ensuring high security and privacy measures.
Heterogeneous RANs: Space, Air, and Ground Integrated Network (SAGIN) is a revolutionary concept that aims to create a seamless and highly efficient communication infrastructure.In the context of CAVs and 6G, this integration opens up new possibilities regarding ubiquitous connectivity, efficiency, and global operational viability.CAVs will stay connected regardless of their location or their navigating environment.Moreover, spacebased components facilitate global coverage and precise positioning, eliminating communication gaps in remote or challenging areas where traditional networks may struggle.Concurrently, the inclusion of Infrastructure WiFi in the discourse of CAV connectivity broadens the spectrum of available solutions, enhancing the resilience, versatility, and cost-effectiveness of the communication infrastructure for addressing the evolving demands of the Connected Autonomous Vehicle landscape.

dIstrIbuted computAtIon And InteLLIgence
Edge Intelligence: Sensing and perception at the edge are key enablers in the realm of CAVs for proactively detecting and accurately interpreting information from the surrounding.Different functions related to the operation of CAVs can be executed at the edge, ranging from typical sensor data processing to the remote operation of a vehicle by software.Based on the amount of data to be processed and the level of reasoning that is required by the sensing and perception function, the computation can be placed at different tiers of the computing continuum.At the extreme-edge tier, vehicles near each other can form a short (or even wide) area network by utilizing unlicensed spectrum resources to exchange real-time information and perform collaborative decision-making similar to the use case of platooning.Other functions related to quick detection and response to changing road conditions, traffic patterns, and other environmental factors, for example, collision avoidance, can be executed deeper at the edge -however, still very close to the vehicles -by edge servers, where larger volumes of data can be collected and processed, gaining a more holistic view of the local situation.More complex functions characterized by continuity and duration, like remote and autonomous driving, should be instantiated closer to the cloud to avoid running into multiple edge instances that should coordinate seamlessly.This will be one of the open challenges to be investigated and determined under the 6G paradigm, namely to optimally place the right functions at the right network locations, mainly toward restricting the endto-end jitter and minimize the associated latency for CAV functionality [13].
In addition to improving the speed and efficiency of data processing, sensing and perception at the edge can also help to address privacy and security concerns.
By processing data locally, sensitive data can be kept within the vehicle or the network edge, reducing the risk of data breaches or unauthorized access.Security will be one of the main advantages of 6G compared to 5G, potentially addressed even at the physical layer, but also in higher protocol layers, especially in terms of core topologies over virtualized infrastructure.
Vehicular Cloud: The concept of vehicular cloud [14] holds significant potential as an envisaged enabler for CAVs.By combining the power of distributed computation and intelligence, the vehicular cloud leverages the collective computing capabilities and resources of vehicles within a network.In other words, it acts as a virtual ecosystem where vehicles collaboratively share computing resources, facilitating effi cient processing of complex algorithms, machine learning models, and sensor fusion tasks.This distributed computation and intelligence within the vehicular cloud is expected to enhance the overall perception, decision-making, and control capabilities of CAVs, ultimately improving their safety, efficiency, and user experience.In the context of 6G, the vehicular cloud can leverage advanced network capabilities to enable seamless and effi cient distribution of computation and intelligence among vehicles.
The ultra-high-speed connectivity offered by 6G will allow for rapid and reliable data exchange between CAVs and the cloud infrastructure, facilitating real-time processing and analysis of vast amounts of sensor data.The ultra-low latency can ensure timely and responsive decision-making, enabling CAVs to navigate complex environments with enhanced safety and effi ciency.Furthermore, the massive device connectivity of 6G is expected to enable the large-scale deployment of the vehicular cloud, integrating a multitude of vehicles into the distributed computing infrastructure.This will promote collective intelligence, where CAVs can collaborate and share resources, optimizing individual and collective performance.
Caching: Apart from computation and intelligence placed at the diff erent tiers of the computing continuum, caching should also span across the end-to-end network, reducing latency and improving data transfer efficiency between vehicles and the cloud.Specifically, in the context of CAVs, data such as maps, traffi c information, and sensor data must be continuously transmitted to the cloud.However, this can be challenging due to the limited bandwidth and congested backhaul links, the high volume of data, and the low tolerance for latency.Caching is a technology that stores frequently accessed data closer to the vehicles, reducing the need to retrieve data from the cloud whenever needed.By caching data in the vehicle, collaboratively using multiple vehicles' storage units (i.e., at the extreme edge tier) or at the edge tier of the network, these can be retrieved faster, with lower latency, and by incurring less network congestion [15].We believe that (distributed) caching will be one of the prime enablers within 6G, and even 5G SA, for achieving the required performance for such applications.

concLusIon
This article presented an empirical analysis of the current landscape, challenges, and future vision of Connected and Autonomous Vehicles (CAVs) from a 5G perspective.By conducting trials over a commercial 5G deployment, focusing on two representative use cases, namely "Remote Driving" and "Quality of Service (QoS)-based Level-of-Automation Selection," we have gained valuable insights into the communication requirements and performance metrics of CAVs.Through our quantitative analysis, we have identifi ed several open challenges that must be addressed to ensure the seamless integration of CAVs into the 5G ecosystem.This requires optimizing the underlying network architecture by transitioning from 5G Non-Standalone (NSA) to fully Standalone (SA) deployments.Furthermore, in the context of 6G, incorporating distributed intelligence, computation, and caching across the network is envisioned to further close the gap, contributing to the ultimate development of a smart and sustainable transportation environment.

FIGURE 4 .
FIGURE 4. Evolution of mobile networks for Connected Autonomous Vehicles (CAVs).
High-level overview of the examined use cases.

TABLE 1 .
Performance overview of conducted remote driving test cases.
FIGURE 2. Optimal versus measured KPIs values for the remote-driving trials.

TABLE 2 .
Performance overview of conducted QoS-based LoA selection experiments.Optimal versus measured KPIs values for the QoS-based LoA trials.
GriGorios KaKKavas [S'19] is a Ph.D. candidate and a research assistant at NETMODE Lab, NTUA, Greece.He received a diploma in ECE from NTUA in 2010 and an MSc in Management and Economics of Telecommunications Networks from UOA in 2012.His research interests lie in the areas of 5G communications, cognitive radio, and network monitoring techniques.Maria DiaManti [S'21, M'23] is a Ph.D. candidate and a research assistant at NETMODE Lab, National Technical University of Athens, Greece.She received a diploma in Electrical and Computer Engineering from the Aristotle University of Thessaloniki in 2018.Her research interests lie in the areas of 5G communications, resource management and optimization, and game theory.vasileios Karyotis (S'07, M'11] is an Assoc.Prof. with the Dept. of Informatics, Ionian Univ.Since 2009 he is a research associate of NETMODE Lab of NTUA, and since October 2017 an adjunct lecturer with the Hellenic Open University.He received a Diploma in ECE from NTUA in 2004, an MSc in EE from UPenn in 2005, and a PhD in ECE from NTUA in 2009.He has co-authored more than 90 papers and 2 books, and received a best paper award in ICT 2016.KwaMe nseboah nyarKo is a Ph.D. candidate and research assistant at the Professorship for Communications Engineering at the Technical University of Chemnitz (TUC), Germany.He received an M.Sc. in Embedded Systems from TUC in 2021.His research interests lie in autonomous driving and cooperative vehicles.Matthias Gabriel is a Ph.D. candidate and research assistant at the Professorship for Communications Engineering at the Technical University of Chemnitz (TUC), Germany.He received an M.Sc degree in ICT from TUC in 2014.His research interests lie in modern algorithms for CAVs such as robust localization, cooperative perception, and scene understanding.anastasios Zafeiropoulos [M] received the Diploma and Ph.D. degrees from the School of Electrical and Computer Engineering, National Technical University of Athens (NTUA).He is a Postdoctoral Researcher at NTUA.He has participated in multiple research projects, focusing on the fields of 5G/6G, cloud/ edge computing, IoT and machine learning technologies.syMeon papavassiliou [SM] is currently a Professor with the School of Electrical and Computer Engineering, National Technical University of Athens (NTUA).Prior to joining NTUA he was a Senior Technical Staff Member at AT&T Laboratories, NJ, USA, and Associate Professor at the ECE Department, New Jersey Institute of Technology, USA.He has an established record of publications in his field of expertise, with more than 400 technical journals and conference published papers.Klaus Moessner [SM] is currently a Professor of communications engineering with the Chemnitz University of Technology and also a Professor in cognitive networks with the Institute for Communication Systems, University of Surrey.He was the Founding Chair of the IEEE DYSPAN Working Group (WG6) on sensing interfaces for future and cognitive communication systems.He has also led several EU-funded ICT projects. bIogrAphIes