Machine Learning Based Resource Allocation For Vehicular Networks

With the rapid development of communication networks,the future network will become an autonomous system.Massive data and intelligent algorithms will definitely become the important basics for the further development of wireless networks.In terms of transportation,with the increasing demands for intelligent vehicles,vehicular network is an urgent need of research and development.In vehicular networks,all kinds of vehicle services are of great significance to relieve traffic pressure and reduce traffic accidents.Unfortunately,the resources of vehicular networks are limited.With the increase in vehicle service demands,the reasonable resource allocation strategies in vehicular networks are of great significance of ensuring the high reliability and low latency of vehicle services.Firstly,the research background of vehicular networks and how to apply machine learning in vehicular networks are introduced.The resource allocation strategies in vehicular networks are described in details.Further,the applicability of machine learning methods is described.It is proposed that the machine learning methods would be used to allocate the resources of vehicular networks.Secondly,a three-layer vehicular network architecture is proposed,including local computing layer,MEC layer and cloud computing layer.On the basis of this architecture,considering the low delay requirement of vehicular networks,the objective function is constructed,which takes the delay of completing tasks as the system cost.In view of this objective function,the state,action and reward are formulated for the system,and a resource allocation algorithm based on reinforcement learning method is proposed to realize the efficient dynamic resource allocation among layers.Simulation results show that the proposed strategy can minimize the total cost and make the service tasks meet their low latency requirements.Thirdly,a multi-objective resource allocation strategy based on reinforcement learning method is proposed.Considering the low latency and high reliability requirements of the vehicular networks,the total latency and rate are calculated and used as the total system cost to construct the objective function.Then,a resource allocation algorithm is proposed to solve this objective function.For local computing layer,a resource allocation algorithm is proposed.For mobile edge computing layer,the states,actions and rewards are formulated for the system,and a resource allocation algorithm based on multi-objective reinforcement learning method is proposed.Simulation results show that the proposed strategy can realize inter-layer and intra-layer communication and computing resource allocation efficiently,as results,the service tasks in the vehicular networks can meet its low latency and high reliability requirements.Finally,the main work and innovation points of this thesis are summarized,and the future research work is prospected.