Design And Implementation Of ADS-B Channel Adaptive Equalization System Based On TCN-LSTM

With the booming development of civil aviation,the communication quality of aviation communication transmission system is required more and more.However,in the complex aviation channel environment,the transmission signal of Automatic Dependent SurveillanceBroadcast(ADS-B)system is susceptible to multipath effect and doppler shift,resulting in signal time delay and Inter-Symbol Interference(ISI)and other problems.In order to solve these problems,adaptive equalization techniques are often used to process ADS-B channel.However,the traditional adaptive equalization technique suffers from slow convergence speed and poor steady-state BER performance in harsh channel environments,so it is important to study adaptive equalizers with better performance for the development of aviation communication technology.This article applies neural networks to adaptive equalizer and designs an adaptive equalization system for ADS-B channels based on TCN-LSTM neural networks,which can effectively improve the transmission performance of ADS-B signals in harsh air channel environment with fast training speed.The main research contents of this article are as follows:1)The effects of multipath effect and doppler shift in the aviation channel on ADS-B signals are analyzed,and the transmission environment of ADS-B signals is divided into four different channel scenarios,namely,cruise,takeoff and landing,cruising and parking,according to the flight status of the aircraft,and the corresponding channel models are established.The performance of the conventional Least Mean Square(LMS)algorithm and Recursive Least Square(RLS)algorithm in channel equalization is investigated.2)To address the shortcomings of convergence performance and BER performance of traditional adaptive equalization algorithms in ADS-B communication,this article proposes an adaptive equalization scheme for ADS-B channels based on Long-Short Term Memory(LSTM)network by using neural network structure instead of lateral filters in traditional adaptive equalization.The simulation is compared with traditional LMS and RLS equalization algorithms in four different channel scenarios.The simulation results show that the LSTM equalization algorithm can not only accelerate the convergence speed of the algorithm,but also improve the BER performance.3)Meanwhile,in order to further improve the feature extraction capability of the neural network equalizer,an adaptive equalization scheme for ADS-B channels with improved TCNLSTM neural network is proposed.The scheme uses the inflated causal convolution of Temporal Convolutional Network(TCN)to enhance the feature extraction capability and filters the key features of the neural network output by the self-attentiveness mechanism,while introducing batch normalization and improving the Re LU activation function in TCN to Leaky Re LU activation function to improve the equilibrium performance.The TCN-LSTM neural network adaptive equalization algorithm is compared with the LSTM neural network adaptive equalization algorithm in four different scenarios.The simulation results show that the improved TCN-LSTM neural network adaptive equalization algorithm improves the convergence speed by about 2% and the BER performance by about 3d B compared with the LSTM equalization algorithm.4)The TCN-LSTM neural network adaptive equalization system is designed and built based on the software radio platform,and the equalization performance of the neural network equalizer is verified by collecting ADS-B signals in the real environment.The experimental results show that the TCN-LSTM neural network equalization system designed in this article has a good equalization performance for the actual ADS-B channel,and its amplitude mean square error reaches 0.04.