Research On Resource Allocation And Optimization For Joint Communication,Sensing,and Computation Enabled Vehicular Networks

High-level autonomous driving and advanced vehicle-road coordination,as the keys for improving driving safety,enhancing traffic efficiency,and realizing intelligent transportation,put forward new requirements for the future development of vehicular networks:1)Efficient,real-time,low-latency,and highly reliable interaction capability.2)Networked high-precision,highresolution sensing capabilities.3)Ubiquitous intelligent computation capabilities.In this case,the joint communication,sensing,and computation enabled vehicular networks are born.In such networks,communication,sensing,and computation are deeply integrated and mutually beneficial,making wireless networks simultaneously have physical-digital space sensing,ubiquitous communication,and intelligent computation capabilities,which provide network support to enable high-level autonomous driving and advanced vehicle-road collaboration.However,the current domestic and international research on joint communication,sensing,and computation enabled vehicular networks is mainly focused on the concept proposal,basic theory research and network architecture design.The research on the multi-domain resource allocation and optimization of the joint communication,sensing,and computation enabled vehicular networks is still in the initial and exploratory stage.Within this perspective,this thesis is dedicated to exploring the new paradigm of resource allocation and optimization for the joint communication,sensing,and computation enabled vehicular networks,and provides theoretical guidance and methodological support to realize the collaborative optimization and efficient utilization of communication,sensing,and computation multi-domain resources.In this way,system-level joint optimization of communication,sensing and computing performances can be achieved,enabling the network to simultaneously support highly reliable,low-latency information interaction,high-precision,wide-range sensing and adaptive,on-demand computing power allocation,effectively meeting the new demand of smart transportation applications such as high-level automatic driving and advanced vehicle-road coordination for future vehicular networks.The innovations and contributions of this thesis are mainly included in the following four aspects.1.Queue-Aware Dynamic Downlink Resource Allocation and Optimization StrategyIn the joint communication,sensing,and computation enabled vehicular networks,RSU is required to have both well sensing and high bandwidth communication capabilities.To this end,this thesis proposes a queue-aware dynamic power and subcarrier allocation strategy.Specifically,this thesis firstly introduces orthogonal frequency division multiple access technique into the joint communication,sensing,and computation enabled vehicular networks to support communication function for massive vehicles.For the huge amount of packets to be forwarded during downlink communication,this thesis designs a queue scheduling model of data packets to perceive the status of current queue backlog and schedule the queue dynamically.Further,this thesis formulates a transmit power minimization problem,while considering network stability,downlink communication and sensing performances of RSU.Since the problem is a non-convex problem,this thesis exploits the Lyapunov optimization technique,successive convex approximation method and parameter transformation method to convert this problem into a convex problem.Finally,this thesis derives an optimal closed-form power allocation expression by using the Lagrangian dual decomposition method.Simulation results show that the proposed strategy can achieve the optimal power consumption of RSU while meeting the requirements of downlink high bandwidth communication and well sensing performances.2.Dynamic V2I Uplink Resource Allocation and Sensing Information Processing StrategyIn the joint communication,sensing,and computation enabled vehicular networks,the vehicles upload the massive sensing information to the RSU via V2I uplink communication.Then,the RSU can gather sensing information from different vehicles and combine it with its own sensing information to perform data fusion,dimensionality reduction and redundancy reduction at the mobile edge computing(MEC)server,to provide vehicle-road cooperation services for vehicles.However,the wireless channel status changes rapidly.In the case of bad channel conditions,uploading the sensing information from the vehicles results in high uplink transmission latency.To deal with it,the vehicles can process the information locally and just deliver the results to the RSU.However,the processing accuracy of the sensing information on the vehicles is lower than that on the MEC server due to limited computing power of the onboard server.Besides,processing locally leads to higher vehicle energy consumption.Thus,this thesis proposes a two-stage deep reinforcement learning(DRL)-based resource allocation and sensing information processing strategy.Specifically,this thesis firstly designs the terminal machine learning(ML)task model and the edge ML task model on the vehicle side and RSU side.Then,this thesis formulates a long-term multi-objective optimization problem to jointly optimize the sensing information processing accuracy,the total sensing information processing delay,and the vehicle energy consumption.Notably,owing to the stochastic traffic and time-varying communication conditions,this thesis reformulates it as a Markov decision process(MDP).Simulation results show that the proposed strategy is able to effectively jointly optimize the sensing information processing accuracy,the total sensing information processing delay,and the vehicle energy consumption,which can realize high sensing information processing accuracy and low total sensing information processing delay while minimizing vehicle energy consumption as much as possible.3.Dynamic Vehicle-Side Dual-Beam Power Resource Allocation and Computation Offloading StrategyThe vehicles can be equipped with a beam-sharing based joint communication and sensing system to realize V2I uplink communication and sensing functions in different directions.Further,to achieve high-level autonomous driving,the vehicles also need to process massive sensing information.Due to the limited computing power and energy of the vehicles,the vehicles can offload tasks to the MEC server via V2I uplink communication to improve the processing speed and accuracy of the tasks.However,there is mutual interference between the V2I uplink communication beam and the sensing beam,and the high mobility of the vehicle leads to real-time dynamic changes of wireless channel conditions.Thus,this thesis proposes a two-stage DRL-based resource allocation and offloading strategy.Specifically,this thesis firstly models the inter-beam interference in two cases based on the time-varying relative positions between the vehicles and the RSU.This thesis also designs the local execution model and the MEC server execution model for the computation offloading problem.Then,this thesis formulates a long-term multi-objective problem that jointly optimizes the task execution latency and the sensing performance of multiple vehicles.Taking into account the stochastic traffic,the time-varying V2I channel gain and the time-varying impulse response of sensed target,this thesis reformulates the multi-objective optimization problem as an MDP.Simulation results show that the proposed strategy can effectively achieve the joint optimization and performance tradeoff between the task execution latency and the sensing performance.4.Queue-Aware Dynamic Vehicle-Side Multi-Beam Resource Allocation and Optimization StrategyBased on the above-mentioned two-beam based integrated communication and sensing systems,three-beam based integrated communication and sensing systems are introduced,i.e.,rearward V2V communication is achieved by transmitting beams directed to the rear of the vehicles to assist the rear vehicles in achieving high-level autonomous driving functions such as over-the-horizon sensing.Nevertheless,in this case,both the inter-beam interference of a single vehicle and the inter-vehicle interference are induced.Therefore,this thesis proposes a queue-aware dynamic beam power allocation strategy.Specifically,based on the above inter-beam interference model of two beams,this thesis further investigates the inter-beam interference problem of multi beams,and this thesis also proposes an interference model during the inter-vehicle communication and sensing process.Then,this thesis formulates a stochastic programming problem,which optimizes the sensing performance,subject to constraints on the network stability,power limits and quality-of-service requirements.Leveraging the Lyapunov optimization technique,this stochastic programming problem is transformed into a single-time slot non-convex problem.Further,an improved hybrid meta-heuristic algorithm is designed to solve the non-convex problem.Simulation results verify that the proposed strategy is able to effectively joint optimize the multi-vehicle communication performance(V2I uplink communication performance,V2V communication performance)and sensing performance,which can satisfy the high bandwidth V2I and V2V communication requirements while making the vehicle-side sensing performance optimal as much as possible.