Research On Collaborative Computing And Resource Allocation Strategy Of Internet Of Vehicles Based On Artificial Intelligence

In order to alleviate the computational burden of vehicle,mobile edge computing is used as a promising computing paradigm in the Internet of Vehicles,allowing vehicles to offload computation-intensive and time-sensitive tasks to servers at edge nodes which are able to provide computing services for vehicles.Under such a background,in order to reduce the latency of task execution and improve the utilization of system resources,combined with mobile edge computing technology,this thesis mainly studies the collab-orative computing and resource allocation strategies in the Internet of Vehicles based on artificial intelligence.The research is mainly conducted in two scenarios,namely the col-laborative computing and resource scheduling in the V2I-based offloading scenario and the V2V-based offloading scenario.Vehicles in the adjacent area will perceive a large amount of the same data and gen-erate many similar tasks.This thesis focuses on the repeated communication and repeated calculation problems of vehicles in the adjacent area during the offloading process,and proposes a cooperative bandwidth,computing and storage resources allocation mecha-nism with the optimization goal of minimizing system cost.Finally,the joint optimization problem is constructed as a mixed nonlinear integer programming problem,and the sub-optimal solution is obtained by using the multi-dimensional resource allocation algorithm based on DDPG reinforcement learning.The simulation results show that the above strat-egy can effectively reduce the total delay of system task execution and improve resource utilization.Previously,we mainly studied the problem of V2I computing offloading under the wide-area coverage of RSU.With the ultra-dense deployment of 5G base stations,the coverage of RSU will become smaller and smaller,which brings great challenges to the continuity of traditional V2I computing services.In order to provide uninterrupted com-puting offloading services,this thesis proposes a V2V-based opportunistic task offloading strategy.In view of the huge vehicle density on urban roads,the effective communication time of front and rear vehicles is usually close to the minute level.Based on the op-portunistic V2V communication and task calculation brought by mobility,vehicles with computational migration can allocate tasks to vehicles with rich resources.In addition,V2V short-distance communication can opportunistically reuse V2I spectrum resources,and provide V2V low-latency and high-reliability computing offloading services without affecting the communication needs of the primary user.In order to solve the problem of re-source allocation,this thesis explores resource allocation strategies based on single-agent DDPG and multi-agent DDPG,rationally scheduling vehicle tasks and allocating commu-nication resources.Compared with other association and channel selection strategies and the joint resource allocation optimization strategy based on single-agent DDPG,numer-ical simulations verify that the task offloading and joint resource allocation optimization strategy based on multi-agent DDPG proposed in this thesis have a better service experi-ence.