Research On Mobile Parking Incentive Based On Reinforcement Learning In Vehicle Network

With the rapid development of wireless devices,more and more vehicles are now equipped with a large number of wireless devices,making it possible to use large-scale vehicles networks.Therefore,the industry and academia pay more attention in vehicular ad hoc networks(VANETs).Due to the fast moving speed of vehicles,the rapid and dynamic changes in the topology of vehicles nodes,and the low bandwith of wireless devices in vehicles,the vehicles network quality is poor,which is different from traditional wireless networks.In this paper,the solution is to use the vehicles parked on the roadside,add the vehicles to the network,and use the wireless devices in the parked vehicles for data distribution in order to realize network communication between mobile vehicles.Vehicles parked on the roadside have the characteristics of large amounts,wide distribution and long parking time.According to research,the vehicles clusters formed by parking at specific locations are stable,which can be used as roadside infrastructure to improve the quality of vehicle communication.Adding parking vehicles to the vehicles network requires car owners to share the wireless equipments on their vehicles.According to the survey,less than 30% car owners are willing to share the wireless equipments in their vehicles.Except,many owners are accustomed to park in the parking lot,which can not increase the connectivity of the vehicle network.The solution to these problems is usually to design an incentive mechanism to motivate users to park their vehicles at specific locations.This paper proposes a pricing system,considering the design of a dynamic pricing algorithm based on reinforcement learning,which aims to maximize the connectivity of road sections and minimize system costs.(1)Reinforcement learning does not rely on the model,it can learn the optimal dynamic pricing strategy through the dynamic interaction between the agent and the environment.(2)Reinforcement learning has strong adaptability.Due to the rapid changes in the road traffic environment,the overall environment is uncertain.Reinforcement learning can continuously learn to adapt to dynamic changes towards the optimal reward.Due to the large state space caused by reinforcement learning in the highdimensional situation,the learning speed is slow.This paper proposes a pricing strategy based on deep Q network(DQN),which can compress the state space and estimate the Q value through the deep neural network.Based on real road and parking data,the results of the vehicle trajectory simulation of Vanet Mobi Sim and the network simulation of NS2 show that the scheme improves the completion rate of tasks,and the parking result improves the communication quality in vehicles.