High Mobility Wireless Channel Modeling And Adaptive Equalizer Based On Learning Algorithm

By virtue of rapid growth of the high-speed railway network and the exloding develop-ment of self-driving techniques,people are going to spend much more time on fast moving vehicles.Passengers are demanding high-rate and reliable wireless communications over high speed vehicles desperately.The increasing of mobility not only causes more Doppler effect of the channel,it will change the properties of channel in other aspects.Traditional model cannot describe high mobility channels accurately,leading to the existing communication system not performing well in high mobility scenarios.Therefore,we need to look deeply into the impact on wireless channels caused by high mobility,and bring up with a more suitable channel modeling.Moreover,the way to combat fast time-varying fading channel is adaptive techniques.Adaptive equalizer is capable to cancel inter-symbol interference and track time-varying channel by a number of training data.The high mobility channels make serious challenge on the design of adaptive equalizers.In this dissertation,we focus on the study of channel modeling in high mobility scenarios and design of adaptive equalizer.First,we propose an fading channel model by characterizing the time-dependent evo-lution of ambient scattering under linear mobility,and analyze the spatial and temporal channel correlation as a function of both the distribution of surrounding objects and the moving speed.We provide a closed-form spatial-temporal signal correlation expressions.By computational simulations,we focus on the features of correlation in high mobility scenarios and reveal the impacts of the varying scattering in comparison with existing models.The the-oretical analysis of the novel channel correlation model may serve as the design reference and performance evaluation for future high mobility communication systems.And we also add practical implications through the evaluation of channel coherence time and instantaneous SNR Level-crosing rate.Then,we bring up with a novel wideband channel model for a moving carriage with multiple windows through which a point source casts its radio signal to the in-carriage receivers.The high speed train is just such a typical case.Since the carriage is moving.it is reasonable to believe that the channel between the outside base station and the in-carriage receiver is quite different from the conventional static indoor channel model.To the best of our knowledge,such a moving indoor model has not been studied yet.Based on ray tracing,we obtain a deterministic wideband chanuel model considering the outdoor and indoor propagation,reflection and diffraction.The results show some special characteristics of the high speed train communication channel,which shed some light on the system design The results are further validated by simulationAt last,we design adaptive equalizers based on recurrent neural networks(RXNs).Long short-term memory(LSTM)RNN networks have great ability to deal with non-linear clas-sification problem.We use LSTM network to build three kinds of equalizers,and test their performance by simulation under several linear and non-linear channels.The experiment results show that LSTM equalizers we designed have better performance compared to tra-ditional decision feedback equalizer,especially under non-linear channel or systems with higher order niodulation.Besides,the LSTM equalizers5 tracking capability of time-varying channels is also verified by simulation results.