| In recent years,research activities in the field of vehicular ad hoc network have been very active and tremendous progress has been made.These studies have deeply analyzed how to design data dissemination protocols and related algorithms in vehicular ad hoc network environment.Nowadays,vehicular ad hoc network is still an active field that remains rapid development.At the same time,there are also a lot of research challenges to be resolved in the field of vehicular ad hoc network.In addition,the continuous innovation of wireless communication technology has also brought unprecedented challenges to the application of vehicular ad hoc network.The main purpose of data dissemination between vehicles and vehicle and roadside infrastructure through wireless communication technology in vehicular ad hoc network is to further improve people's driving safety and daily traffic efficiency.Considering the rapid change of the network topology caused by the high-speed movement of vehicles,the intermittency of wireless link connections between vehicle nodes and the uneven distribution of vehicles' location,this paper study the data dissemination protocol based on reinforcement learning in different vehicular ad hoc network environments from different perspectives and the main works are as follows:1.Considering the highly dynamic characteristics of VANETs,a mobile adaptive unicast data dissemination protocol is proposed based on reinforcement learning.With the help of the carefully designed learning HELLO packet structure,a new distributed dynamic adaptive learning algorithm is proposed to learn and perceive the dynamic change of the network in real time,which improves the dynamic adaptability of the protocol in the VANETs environment.By jointly considering multiple key evaluation metrics of link connection,a new learning reward strategy is designed.In order to realize that the protocol can quickly respond to the high mobility of vehicle nodes,different learning strategies are designed for different types of data message packets in the network.In addition,in order to accelerate the convergence speed of the learning algorithm,a reactive path probe learning strategy is adopted in the initial stage of learning.Finally,a new MAC layer assisted learning strategy is proposed to further improve the mobile adaptability of the protocol in the VANETs environment.The simulation experiment results show that while the network overhead is kept within the acceptable range,the comprehensive dissemination performance of the protocol in terms of packet delivery success rate,end-to-end delay and average routing hops outperforms existing related protocols.2.Considering the uniqueness of the urban VANETs environment,an RSU-assisted vehicle traffic-aware unicast data dissemination protocol is proposed based on reinforcement learning.In urban VANETs scenarios,vehicle nodes traveling fast along the street pose a severe challenge to the data dissemination protocol,and the high mobility of vehicle nodes has a great impact on network performance.In addition,the local network fragment caused by the uneven distribution of vehicle nodes in the urban environment further puts forward higher requirements on the distribution protocol.More importantly,the high vehicle density and various natural obstacles during peak traffic hours,such as tall buildings,shelter from adjacent vehicles,and tall trees beside the road,further increase the difficulty of implementing efficient and reliable data dissemination protocol.To resolve the above problems,by utilizing the regularity of urban road network topology,addressing the problem of explosion of state space in traditional learning algorithms,and at the same time,in order to further reduce the network overhead introduced in the learning process and reduce the protocol's sensitivity to the high mobility of vehicle nodes,a new traffic-aware unicast dissemination strategy is proposed.This strategy realizes the transfer of the traditional multihop forwarding between the vehicle nodes with weak dynamic adaptability to the reliable and efficient multi-hop efficient forwarding with traffic-aware perception and strong dynamic adaptability between the vehicle nodes within the road segment.In order to achieve reliable and efficient forwarding from the source vehicle node to the first intersection node and the last intersection node to the destination vehicle node,a vehicle traffic-aware and dynamic adaptability learning strategy is proposed based on the redesigned learning V2 VHELLO packet structure to achieve reliable and efficient forwarding between vehicle nodes within the road segment.Considering the uneven distribution of vehicle nodes within the road segment and the limitation of the movement trajectory,combined with Q greedy geographic location forwarding and store-and-carry forwarding under local optimal conditions,a new vehicle traffic-aware and learning reward assignment strategy is proposed by jointly considering multiple vehicle link connection evaluation metrics.Based on the redesigned R2 R HELLO learning packet structure and the learned road segment traffic flow information,a new vehicle traffic-aware learning algorithm between RSU nodes is proposed to achieve efficient forwarding between road segments.The simulation experiment results show that the comprehensive performance of the protocol under different traffic density is better than the existing related dissemination protocols.3.Considering the high mobility,uneven location distribution and instability of communication link connections of vehicle nodes in urban VANETs environment,an RSUassisted vehicle traffic-aware multicast data dissemination protocol based on reinforcement learning is proposed.In order to minimize the impact of the rapid mobility of vehicle nodes on the performance of the multicast dissemination protocol,a new multicast dissemination model is proposed based on the reinforcement learning algorithm with the help of the static road topology information of the urban road network.This model transforms the traditional high-dynamic data dissemination process that directly focuses on the forwarding process from the multicast source node to the multicast group member nodes to a reliable dynamic data dissemination process which focuses on the forwarding process from the multicast source RSU node to the multicast group member RSU nodes to further improve the dynamic adaptability of the protocol.The multicast source vehicle node moving in a road segment first quickly disseminates the data packet to the multicast source RSU node,and then the multicast source RSU node continues to efficiently disseminate it to the multicast group member RSU nodes,and finally each member RSU node reliably disseminates the data packet to its associated multicast member vehicle nodes.As to the problem of poor dynamic adaptability of traditional multicast protocols,combined with the help of the redesigned V2 V and R2 R learning HELLO packet structure,different traffic-aware multicast learning strategies within and between road segment are designed.The experimental results show that,compared with the traditional tree-and-grid-based multicast dissemination protocols,our protocol has well dynamic adaptability and better multicast dissemination performance.4.In order to further improve the performance of emergency message dissemination in the highway environment in VANETs,a practical distributed mobility adaptive clustering protocol suitable for highly dynamic network scenarios is proposed through the exchange of three types of clustered data message packets HELLO,CH and JOIN,which improves the stability of the cluster structure.By sending and receiving clustered data message packets with the unified format to obtain the latest neighbor nodes and clustering information,the protocol is more adaptable to high network dynamics.By introducing the periodic broadcast of CH data message packets for the cluster head node with the absolute highest weight,it has a high tolerance to the clustering packets loss and also avoids the butterfly effect. |
PDF Full Text Request