Research On Data Distribution For VANET Based On Learning

Since the invention of the automobile in the first industrial revolution,with the development of economy and technology,more and more vehicles have been produced.However,with the increasement of vehicles,the traditional wireless cellular network can not meet the bandwidth requirements of communication between vehicles any longer.So,the concept of VANET is proposed.In the VANET,each vehicle can communicate with other vehicles and RSUs through OBU to obtain road information,so as to reduce the accident rate,realize real-time navigation and improve driving efficiency.Therefore,improving the efficiency of data distribution plays an important role in VANET.In the existing research,most of them only use V2 I for data distribution,which is inefficient.Even if V2 I is combined with V2 V,deep learning and reinforcement learning are not used,so their dynamic adaptability to VANET is poor.In this project,the method of cluster head node selection based on deep reinforcement learning combining V2 I and V2 V communication is adopted to improve the efficiency of data distribution.In particular,Software-defined network(SDN)controller uses deep reinforcement learning(the combination of CNN and DND)to choose edge cluster head nodes and gateway cluster head nodes from all vehicles.Data is distributed by these cluster-head nodes to avoid congestion caused by direct communication between all vehicles and the RSU,so as to improve the efficiency of data distribution.In addition,the existing research does not consider that data transmission between vehicles may encounter obstacles such as high-rise tunnels.Therefore,in order to further improve the connectivity and stability of VANET and improve the efficiency of data distribution,this project introduces UAV as an auxiliary relay node for data distribution.When data transmission is carried out among non-cluster-head nodes,each node will evaluate the effectiveness of the next hop node and the distance between the next hop node and the destination node through Q-learning,and then select the optimal next-hop forwarding node to ensure the stability and efficiency ofdata transmission.The simulation results show that by using the cluster-based method combined with V2 I and V2 V,the packet delivery rate of the system is higher than that of V2 I communication alone and that cluster-head nodes are selected randomly.In addition,in the case of UAV collaboration and using Q-learning to select the optimal next-hop forwarding node,the packet delivery rate of the system is improved compared with AODV and GPSR,and the end-to-end delay is reduced.These show the effectiveness of the proposed scheme in improving system performance...