Research On Intelligent Routing Protocol Of Vehicle Network

The vehicular ad hoc network is a wireless ad hoc communication network composed of mobile vehicle nodes distributed on the road and roadside infrastructure nodes with a fixed location.The mobile vehicle node in the network is not only used as a terminal node to send and receive data,but also as a relay node to achieve multi-hop routing and forwarding of data.As one of the important parts of the Intelligent Transportation System,it has received extensive attention at home and abroad.Different from the traditional ad hoc network,the vehicle nodes in the vehicular ad hoc network move faster.This makes the network environment of the vehicle network change more frequently,and the time for maintaining a stable link between communication nodes is also shorter.The traditional routing protocol cannot be efficiently applied to multiple types of vehicular ad hoc network scenarios.Therefore,it is necessary to customize matching and efficient routing protocols according to different application scenarios.This article has made the following work for the longdistance and cross-street communication scenarios in the vehicular ad hoc network:(1)In order to achieve efficient and stable communication from the source vehicle node to the destination vehicle node,an ant colony optimization vehicle network routing protocol based on the cell attractor selection model is proposed,which transmits data in the center of the street.First,the street evaluation method is designed according to the connectivity and delay in the street,and based on this,a path search algorithm based on the ant colony optimization algorithm is proposed to find the street path to the intersection of the destination.Then,the communication quality of neighboring nodes is evaluated according to the reception of HELLO messages,and a next-hop decision-making mechanism based on the cell attractor selection model is proposed for forwarding decision-making between neighboring vehicles.Finally,the data packet will be forwarded along the route found by the path search algorithm,and the node with the highest weight in the set of neighbor nodes will be selected as the next hop according to the next hop decision mechanism until it reaches the destination node.(2)Scheme(1)will generate more control data packets,leading to increased routing overhead.Aiming at this problem,a vehicle network routing protocol based on multi-agent reinforcement learning is proposed.First,a distributed multi-agent network system model is established to describe the working process of the protocol.Then,the Actor-Critic method suitable for the vehicle network environment is designed.The next hop intersection node is selected through the Actor network,and the Critic network is used to evaluate the quality of the decision,so as to make the Actor's decision more accurate.Finally,a method that uses distributed reward strategies to consider the entire network information is proposed.On the one hand,the reward function includes a local(direct)reward describing the interaction between individuals,that is,the environmental information about the street between two adjacent intersection nodes.On the other hand,global rewards are introduced to reflect the quality of the actions performed,that is,the direction of data packet transmission.Experimental results prove that the proposed protocols can maintain a relatively good data packet delivery rate and a low average end-to-end delay under different number of vehicle nodes,different number of CBR connections,different maximum vehicle speeds and real traffic trajectories.And scheme(2)has lower routing overhead.