Research On Heterogeneous Spectrum Sharing Technology In Cognitive Vehicular Networks

With the explosive growth of vehicular service types and the increasing density of urban vehicles,limited spectrum resources have been difficult to meet the needs of diverse communication services for vehicles.To this end,the vehicular network introduces cognitive radio technology to alleviate the shortage of spectrum resources through the sharing of rich spectrum resources.At present,cognitive vehicular network usually focuses on the use of sub-6 GHz with good penetration characteristics.However,the spectrum resources in the sub-6 GHz cannot fully meet the requirements of diverse vehicular services due to the limited available bandwidth.Compared to sub-6 GHz,millimeter-wave with rich bandwidth resources have begun to attract people’s attention.However,because of the extremely high path loss and penetration loss of millimeter-wave,it is difficult to form a stable,reliable and long-term communication connection.Therefore,the use of sub-6 GHz and millimeter wave in vehicular network has become an important way to meet the communication needs of vehicular network.It is necessary to study the heterogeneous spectrum sharing technology in the cognitive vehicular network.This thesis is supported by Chinese Natural Science Foundation Project "Sensing and Utilization of Polarization Resource in a Dense Heterogeneous Wireless Environment"(No.61571162)and Beijing Natural Science Foundation Project "Research on Dynamic Spatiotemporal Data driven Video Data Understanding based Vehicular Network Resource Allocation"(No.4202049).This paper studies the heterogeneous spectrum sharing technology in the cognitive vehicular network integrating the sub-6 GHz and millimeter-Wave,the main contents are as follows:1.Aiming at the problem of heterogeneous spectrum interference caused by vehicles using sub-6 GHz and millimeter-wave,an interference multi-graph based heterogeneous spectrum sharing algorithm and an interference-aware multi-graph based heterogeneous spectrum sharing algorithm are proposed.First,a heterogeneous spectrum sharing model is constructed,and the latency constraint of this model is converted into signal-to-interferencc-plus-noise ratio constraints through network calculus theory,thereby reducing the difficulity of the packet latency acquisition.Then,an interference multi-graph based heterogeneous spectrum sharing algorithm is proposed.An interference multi-graph is constructed to quantify heterogeneous spectrum interference among vehicles,and a multi-graph coloring algorithms is designed for heterogeneous spectrum allocation.Since the above algorithm only use the system sum rate as optimization goal,vehicles with poor channel quality are facing a loss of communication reliability.Therefore,the original problem of system sum rate maximization is transformed into the problem of maximization of the minimum transmission rate,and an interference-aware multi-graph based heterogeneous spectrum sharing algorithm is proposed to improve the communication reliability of vehicles.Simulation shows that compared with the traditional game based heterogeneous spectrum sharing algorithm,the above two algorithms better guarantee the reliability and delay performance requirements,and can protect the coummnuication outage probability and the packt latency outage probabilbity less than 0.0001 and 0.0009,repectively.Moreover,the interference-aware multi-graph based heterogeneous spectrum sharing algorithm can provide better reliability.2.Aiming at the problem of diverse QoS of vehicular communication links caused by the coexistence of diverse services in the vehicular network,an interference-aware multi-agent deep reinforcement learning based heterogeneous spectrum sharing algorithm is proposed to ensure multi-service QoS requirements.At the same time,the proposed algorithm adopts a distributed scheme,which also solves the problems of high signaling overhead and high computational complexity of the existing spectrum sharing technology in cognitive vehicular network.First,we introduce an effective capacity model and build a multi-service QoS guaranteed heterogeneous spectrum sharing model.Then,based on the interference-aware multi-agent deep reinforcement learning model,the vehicle as an agent is equipped with cognitive device to autonomously select QoS guaranteed parameter and transmission spectrum to obtain a near-optimal heterogeneous spectrum sharing scheme.The algorithm we designed considers the state space dimension change problem of the existing reinforcement learning model due to the change in the number of observed neighbor vehicles.The vehicle with interference sensing ability can build state space by using the interference it senses in each spectrum resource to maintain the stability of state space dimension.The interference can characterize the spectrum resources,distance,and other information used by neighbor vehicles in state space.Simulation shows that the proposed algorithm can guarantee the diverse QoS requirements of vehicular communication links for different services.It is guaranteed that the probability of collision avoidance information transmission service packet latency exceeding 10 ms is lower than the specified 0.01,and the probability of autonomous driving information transmission service packet latency exceeding 100 ms is lower than the specified 0.05.