Channel Estimation Based On Deep Learning In Vehicle-to-everything Environments

With the advancement of science and technology and the development of in-vehicle electronic equipment,vehicle communication technology has been widely studied for its potential of enhancing road safety,improving traffic efficiency and providing rich in-vehicle information and entertainment services.The IEEE 802.11 p standard,known as dedicated short range communication standard,is one of the competitive technologies in realizing vehicle communication.However,due to the highly mobile and the relativly complex communication environment,the wireless channel has been influenced by the Doppler shift and multipath effects.The performance of channel estimation directly affects the demodulation of subsequent signals.Therefore,the research of channel estimation algorithms in vehicle communication scenarios is of great significance to the reliable communication of vehicles.This dissertation is mainly based on the IEEE 802.11 p protocol to design channel estimation algorithm for vehicle communication and try to use software radio technology to implement the algorithm and verify its performance.First,the dissertation presents the theoretical basis of channel estimation in the vehicle communication scenario.We introduce the wireless channel as well as the parameters used to measure the characteristics of the wireless channel and we analyze the general transmission model under the orthogonal frequency division multiplexing(OFDM)technology.Besides,according to the statistical characteristics of the vehicular wireless channel and the parameters in the IEEE 802.11 p standard,the wireless transmission model based on the IEEE 802.11 p standard has been obtained.Afterwards,we summarize some channel estimation algorithms and vehicular communication channel models.Secondly,the data pilot aided(DPA)method can tackle the problem incurred by insufficient number of pilots,but it has error propagation issue which is influenced by the noise and channel time variation.To address this issue,we design a neural network based on long short-term memory(LSTM)and multilayer perceptron(MLP)to track time-varying channels and eliminate noise.Simulation results illustrate that the proposed scheme exhibit better performance than the previous DPA schemes especially for transmitting large-length packets with high-order modulation schemes and/or in fast time-varying channels in simulation.Finally,based on the software radio platform,the proposed LSTM-MLP channel estimation algorithm has beed tested and verified.We design a transmitter and receiver system based on IEEE 802.11 p standard.Then,we test some modulation and demodulation algorithms in the host and RF transmitting and receiving tests on a single device.Finally,we show the performance of the proposed LSTM-MLP algorithm in the actual vehicular communication.From the test results,within the range of statistical signal-to-noise ratio,the bit error rate performance of the LSTM-MLP algorithm is better than that of the STA algorithm and the CDP algorithm.Besides,when using the the high-order modulation schemes,the proposed scheme has greater performance gain.