Research On Resource Allocation Algorithm Of Internet Of Vehicles Based On DDQN

In recent years,intelligent transportation systems have received extensive attention from industry and academia.With the continuous development of communication technology,the number of communication devices has increased sharply,how to use the limited spectrum resources to ensure the communication quality between communication devices has become a major problem in internet of vehicles communication.With the successful application of deep learning and reinforcement learning in image processing,unmanned driving,target recognition and other fields,researchers try to combine deep learning and reinforcement learning technology with internet of vehicles communication technology.Vehicles or base stations are used as agents,after learning,agents can make more reasonable and significant resource allocation decisions,improve the availability of spectrum resources and ensure the quality of communication.Combined with deep reinforcement learning technology,the thesis designs two internet of vehicles resource allocation algorithms based on DDQN.Firstly,this thesis designs a multi-agent distributed resource allocation algorithm based on DDQN under the urban system model.The algorithm regards each vehicle as an agent,and the vehicle can make resource selection decisions independently without base station scheduling.The optimization goal of the algorithm is to maximize the total capacity of V2 I link and the information transmission success rate of V2V link within the allowable transmission delay,and ensure the transmission reliability of V2V link at the same time.Therefore,the algorithm takes V2 I link capacity requirements,V2V link delay and reliability requirements as the constraints of the reward function.Each vehicle inputs the observed CSI to DDQN,and uses DDQN to select transmission resources for the V2V link.The DDQN algorithm is introduced into the target network,which alleviates the overestimation problem of the DQN algorithm and enables the agent to make more efficient decisions and reduce interference between links.Secondly,this thesis designs a centralized resource allocation algorithm based on DDQN under the highway system model.The algorithm regards the base station as an agent and allocates the optimal transmission channel for each V2V link through the reasonable scheduling of the base station.There are two major problems in the centralized allocation algorithm: the rapid movement of vehicles makes it difficult for the base station to obtain accurate CSI in time,and frequent information feedback leads to a large amount of signaling overhead.To solve these two problems,firstly,the observation information of the vehicle is corrected by the delayed prediction CSI model,and then the compression network at the vehicle end is used to compress the corrected observation information of the vehicle and send it to the base station to reduce the signaling overhead in the transmission process.The base station inputs the decompressed information to DDQN,makes resource allocation decisions according to the strategy learned by DDQN,and broadcasts the allocation results to each V2V link.The simulation results show that both the distributed resource allocation algorithm and the centralized resource allocation algorithm based on DDQN can enable the agent to make effective resource allocation decisions,improve the utilization of spectrum resources,and meet the QoS requirements of different services in the Internet of vehicles system...