Research On Efficient Communication And Trustworthy Scheduling For Vehicle-Infrastructure Collaboration In Cognitive Internet Of Vehicles

The combination of wireless communication and Artificial Intelligence(AI)technology has greatly contributed to the development of connected autonomous driving,online traffic services,and intelligent transportation.Cognitive Internet of Vehicles(CIoVs)is oriented to the connected autonomous driving scenario,by introducing cognitive engines to deeply mine multi-dimensional sensory information in physical and cyberspace and obtain cognitive results to assist vehicle decision making and network resource scheduling.It promotes the evolution of individual vehicle intelligence to connected and collaborative autonomous driving,and becomes one of the key technologies in the current mobile communication and transportation fields.In order to fully exploit the advantages of CIoVs,there are still some key issues that need to be addressed,including the problems of frequent communication switching due to fast-changing network topology in CIoVs,the problem of unable to achieve trustworthy sharing of traffic information under malicious node attacks,and the problems of inefficient utilization of multidimensional resources due to imbalance of traffic flow and computational load,etc.To address the above issues,this thesis focuses on the efficient communication and trustworthy scheduling for vehicle-infrastructure collaboration in CIoVs.Combining key technologies such as AI,blockchain,and edge computing,this thesis proposes a vehicle clustering network based on driving behavior characteristics,a blockchain-based traffic information trustworthy sharing and online prediction scheme,and a distributed multi-agent reinforcement learning-based vehicle collaborative scheduling scheme,respectively.The main contributions and innovations of this thesis are as follows:1.A communication mechanism based on driving behavior characteristics for clustering networking in cognitive internet of vehicles is proposed.In CIoVs scenario,the highly dynamic network topology will lead to frequent switching and interruption problems in vehicle communication.The communication mechanism based on driving behavior characteristics of the clustering network is proposed considering the driving motion state and driving behavior of of vehicles.First,a driving pattern model containing multiple feature parameters is analyzed and established to accurately capture driving behavior features.Furthermore,a cluster head selection strategy based on driving pattern stability and channel quality is designed.Mathematical modeling and performance analysis are performed for cluster stability and cluster survival time,which are known to be negatively correlated with the discrete degree of cluster members’ driving patterns.To improve the accuracy of pattern recognition,a Genetic Algorithm-based Neural Network(GANN)pattern recognition algorithm is proposed for driving pattern recognition and classification of Connected Autonomous Vehicles(CAVs),thus supporting efficient vehicle clustering.Finally,the simulation results show that the proposed mechanism can effectively reduce the communication interruption switching frequency and improve the communication throughput and cluster survival time.2.A blockchain-enabled online traffic congestion duration prediction method in cognitive internet of vehicles is proposed.In CIoVs scenarios,the reliable cognition and prediction of traffic situations can assist vehicles in path planning,alleviate traffic congestion,and improve traffic efficiency.Trustworthy sharing of traffic information and data security in CIoVs are crucial to supporting online traffic situation prediction and ensuring traffic safety and efficiency.Blockchain technology with its advantages of decentralization,traceability and tamper-proof can facilitate the trustworthy sharing of traffic information to support online traffic services,but the traditional consensus mechanism still suffers from large computational energy consumption and consensus latency.To support online prediction of traffic situations,a blockchain-based cognitive segment sharing system is proposed,and a traffic efficiency cognitive model based on anomaly detection and filtering mechanisms is designed.Second,to improve the consensus efficiency,a credibility evaluation mechanism is designed and the Credit-Delegated Byzantine Fault Tolerance algorithm(CDBFT)is proposed.Furthermore,an online multi-step prediction algorithm based on Long Short-Term Memory(LSTM)network is proposed for traffic congestion duration prediction.Finally,the simulation results show that the proposed online prediction method effectively improves the accuracy of multi-step prediction.3.A distributed multi-agent reinforcement learning based collaborative path planning and scheduling strategies in cognitive internet of vehicles is proposed.In CIoVs scenarios,there exist vehicles with different computation and communication capabilities with diverse service requirements.However,imbalance in traffic flow distribution and computational load will lead to severe degradation in traffic efficiency and computational performance of vehicles.Collaborative path planning and scheduling can overcome the limitations of individual decision and obtain the global optimal scheduling strategy to improve the quality of service.In addition,leveraging edge computing technology can provide computational resources close to the terminal for collaborative decision-making tasks and reduce computational latency.First,to facilitate vehicle trustworthy collaboration,a blockchain-based distributed collaborative decision-making framework is designed.Second,to meet the diverse service requirements of vehicles,communication,vehicle mobility,and computing task processing models are established and collaborative decision latency is analyzed,and the joint optimization problem of minimizing travel time and computation latency is formulated.Furthermore,we formulate the scheduling problems of different types of vehicle as markov decision processes and a Q-Learning based Distributed Multi-Agent Reinforcement Learning(DMARL)algorithm is proposed for collaborative path planning and scheduling.Finally,the simulation results show that multidimensional resources can be efficiently utilized through collaborative scheduling in CIoVs,and the proposed trustworthy collaborative scheduling strategy improves the load balancing of road infrastructure and edge computing nodes,effectively reducing travel time and computational latency.