Periodic Reporting for period 1 - VESAFE (Enhanced 5G Vehicle to Vehicle Technologies for Safe and Reliable Connected Autonomous Vehicles Applications)
Reporting period: 2021-12-09 to 2023-12-08
With increasing urbanization and more cars on roads, the global transportation sector is facing a big safety crisis. There are more than 1 million people killed on the world’s road every year, with more than 90% road accidents attributing to human errors.
Connected and autonomous vehicles (CAV) holds great potentials for road safety and efficiency, but two challenges need to be tackled before its potentials can be fully unleashed: significantly enhanced vehicle to vehicle (eV2V) services and reliable cooperative perception with autonomous vehicles (AV) sensors.
One important technology to improve eV2V is using millimetre wave (mmWave) communications. However, the directional connectivity of mmWave presents new challenges for the whole protocol stack of V2V networks. On the other hand, reliable cooperative perception requires novel design of deep learning (DL) models for object detection and fusion of sensor data from neighbour AVs. As AVs require extremely high safety and reliability, the above challenges underpinned by CAV demand a holistic design and tight integration of connected vehicles (CV) and AV technologies.
This innovative and multi-disciplinary project VESAFE will tackle the key challenges with complementary expertise and realize the full potentials of CAV. A novel cross-layer design is adopted for the research of 5th generation (5G) eV2V schemes with mmWave and reliable cooperative perception for advanced CAV safety applications: the eV2V schemes are AV oriented and greatly assisted by the AV cooperative perception, and the reliable cooperative perception is boosted by the awareness of V2V network conditions.
And this project proposes a novel integrated evaluation of 5G eV2V and CAV applications to support system design and performance characterization.
Through cutting-edge research, this project will produce robust eV2V schemes and reliable cooperative perception solutions, make breakthrough on 5G V2V and CAV safety applications.
It will significantly strengthen knowledge transfer, boost research and innovation (R&I) capacity and quality, therefore contributing to Europe’s competitiveness and growth in the critical 5G/6G and AV sectors.
Connected and autonomous vehicles (CAV) holds great potentials for road safety and efficiency, but two challenges need to be tackled before its potentials can be fully unleashed: significantly enhanced vehicle to vehicle (eV2V) services and reliable cooperative perception with autonomous vehicles (AV) sensors.
One important technology to improve eV2V is using millimetre wave (mmWave) communications. However, the directional connectivity of mmWave presents new challenges for the whole protocol stack of V2V networks. On the other hand, reliable cooperative perception requires novel design of deep learning (DL) models for object detection and fusion of sensor data from neighbour AVs. As AVs require extremely high safety and reliability, the above challenges underpinned by CAV demand a holistic design and tight integration of connected vehicles (CV) and AV technologies.
This innovative and multi-disciplinary project VESAFE will tackle the key challenges with complementary expertise and realize the full potentials of CAV. A novel cross-layer design is adopted for the research of 5th generation (5G) eV2V schemes with mmWave and reliable cooperative perception for advanced CAV safety applications: the eV2V schemes are AV oriented and greatly assisted by the AV cooperative perception, and the reliable cooperative perception is boosted by the awareness of V2V network conditions.
And this project proposes a novel integrated evaluation of 5G eV2V and CAV applications to support system design and performance characterization.
Through cutting-edge research, this project will produce robust eV2V schemes and reliable cooperative perception solutions, make breakthrough on 5G V2V and CAV safety applications.
It will significantly strengthen knowledge transfer, boost research and innovation (R&I) capacity and quality, therefore contributing to Europe’s competitiveness and growth in the critical 5G/6G and AV sectors.
So far we conducted a thorough survey on V2X communications and disscussed enabling technologies and challenges faced by V2X for advanced cooperative driving applications. We proposed ideas for next generation (6G) V2X based advanced driving applications. We have set up a system V2X simulator with the channel model for mmWave communications, link management and beam allocation algorithms for mmWave communication. The impact of road infrastructure and algorithm design have been instigated with connected and intelligent vehicles. To address the computing and bandwidth limitations at the vehicles, we proposed deep neural networks (DNN) driven compressive offloading for edge assisted semantic video segmentation. We have publised or submitted more than 10 research papers in the top journals or conferences in the areas of mobile computing, wireless networks and cooperative CAV.
Due to the highly congested frequency spectrum at sub-6G used by existing cellular systems, it is foreseen mmWave spectrum will be needed to provide the high data rate needed, which has been set as an active V2V activity for the 3GPP Release 17. mmWave refers to radio frequency spectrum between 24 GHz and 100 GHz, which is already a major part of 5G new radio (NR). With its significantly wider bandwidth and high spatial multiplexing gains, multi-Gigabit and low-latency connectivity could be available. Extensive research has been done for infrastructure based 5G mmWave communication, such as channel modelling and massive MIMO beamforming. As mmWave suffers from excessive propagation loss and susceptibility to blockage by obstacles such as vehicles and buildings, it necessitates the use of directional beamforming and precludes broadcast. The directional connectivity makes V2V operation with mmWave very challenging, including link configuration and beam management, contention based channel access, sidelink autonomous scheduling, distributed congestion control and interference management at MAC layer. This project will move beyond the SOTA by providing robust and innovative solution with mmWave and non-orthogonal multiple access, which will exploit context from AV cooperative perception.
On the other hand, the reliability of the object detection and tracking is critical for data fusion and cooperative perception, but there is lack of investigation in the existing research from the DL models and data fusion aspects. Furthermore, accurate evaluation and modeling of the V2V, cooperative perception and CAV applications are essential steps for protocol design and performance characterization. While there are increasing efforts on evaluating CAV over real testbeds, the existing works are limited in the scale of performance evaluation and depth of understanding the potential system issues. There is no reported evaluation of advanced CAV applications over 5G V2V. This project will move beyond the SOTA by developing novel reliable cooperative perception solutions with new DL models for object detection and data fusion, and novel integrated simulation of CV and AV safety applications
On the other hand, the reliability of the object detection and tracking is critical for data fusion and cooperative perception, but there is lack of investigation in the existing research from the DL models and data fusion aspects. Furthermore, accurate evaluation and modeling of the V2V, cooperative perception and CAV applications are essential steps for protocol design and performance characterization. While there are increasing efforts on evaluating CAV over real testbeds, the existing works are limited in the scale of performance evaluation and depth of understanding the potential system issues. There is no reported evaluation of advanced CAV applications over 5G V2V. This project will move beyond the SOTA by developing novel reliable cooperative perception solutions with new DL models for object detection and data fusion, and novel integrated simulation of CV and AV safety applications