Data-driven Based Reliable Communications For Vanets

With the improvement of the national economy,vehicles have become an indispensable part of people's lives.They have also brought about social problems such as traffic congestion and accidents.The development of information technology represented by the vehicle ad hoc network technology provides a solution to these problems.Since this technology is mostly related to the practical application of personal safety,it imposes very strict constraints on the safety and reliability.Specifically,it includes high accuracy,low latency and strong anti-interference constraints in the process of transmitting information between vehicles.At the same time,thanks to the rapid development of computer computing power in recent years,data-driven methods have been rapidly developed and become an increasingly important role in many fields.Therefore,we try to adopt the data driven methods to the data transmissions in vehicular applications,especially in the issues of wireless channels and anti-interference scenarios,which are widely considered in traditional investigations.The high dynamics of the vehicle ad hoc network environment puts forward higher requirements for the channel estimation.This thesis designs a network model based on convolutional neural network based on the characteristics of vehicle network wireless communication to recover the transmitted information.Different from the existing conventional OFDM system processing algorithm based on the pilot to explicitly estimate the channel state information and then recover the transmitted information,the proposed VNET model does not explicitly estimate the channel state information and directly carry out the recovery of the transmitted information.The training process of the model uses the offline mode and the trained model is used for online testing.The performance of the proposed VNET model on the simulated data set is better than that of the traditional algorithm and the feedforward neural network algorithm,which further proves the great potential of the data-driven methods to estimate the wireless channel in complex environments.To solve the serious signal interference problems,this thesis proposes an anti-interference strategy based on roadside communication unit,which is intended to alleviate the self-signal interference between vehicles in vehicle-intensive areas and the intelligent interference existing in the vehicle networking environment.Specifically,when the receiving vehicle signal is severely blocked or attacked,the roadside communication unit relays the transmitted information.From the perspective of the roadside communication unit,we combine the idea of confrontation learning,based on the solution of the strategy gradient algorithm and the state value evaluation algorithm,and finally both the bit error rate and the utility function performance have improved significantly,verifying the effectiveness of the proposed strategy.