Research On Detection Algorithms Of Orthogonal Time Frequency Space System

The main challenge of the 5^th Generation Mobile Communication（5G）system is how to achieve reliable and high speed transmission in high-speed mobile scenarios such as satellite,UAV and Internet of vehicles.The Orthogonal Time Frequency Space（OTFS）modulation,as a two-dimensional modulation technology,which transform the Time-Frequency（TF）doubly selective channel into Delay-Doppler（DD）channel with approximately time-invariant through two-dimensional transformation,significantly improve the performance of the communication system in high-speed mobile scenarios.This paper mainly studies the implementation method and signal detection algorithms of the OTFS system.The main work is as follows:Firstly,this paper introduces the transmission characteristics of fast time-varying channel in high-speed mobile scene,and deduces the linear time-varying channel in Delay-Doppler（DD）domain,Delay-Time（DT）domain,Time-Frequency domain expression forms in three signal domains and conversion relations among them.The realization mode of the OTFS system,the principle of transmitter and receiver,signal processing flow and signal conversion mode in Delay-Doppler domain and Time-Frequency domain are analyzed.The input-output relationship of OTFS system in the Delay-Doppler domain and the two different OTFS frame structures used in this paper are studied.Secondly,this paper realizes the OTFS system based on OFDM by taking advantage of the compatibility of the OTFS system with the existing Orthogonal Frequency Division Multiplexing（OFDM）system.After the vectorized representation of the signal in the Delay-Doppler domain of the OTFS system is given,this paper studies three classical detection algorithms of the OTFS system in the Delay-Doppler domain,including Linear Minimum Mean Square Error（LMMSE）algorithm,Maximum Ratio Combining（MRC）algorithm and Message Passing（MP）algorithm.LMMSE algorithm is implemented by linear transformation of channel matrix of Delay-Doppler domain.MRC algorithm groups Delay-Doppler grid symbols into vectors,and reconstructs the input-output relationship between transmitting and receiving frames according to the transmitted and received vectors.MP algorithm is implemented by using the sparsity of channel matrix of Delay-Doppler domain.In order to further reduce the complexity of the above three algorithms,this paper proposes an improved Generalized Approximate Message Passing（IGAMP）algorithm based on the idea of the Generalized Approximate Message Passing（GAMP）algorithm.The IGAMP algorithm can accelerates the convergence speed and solves the convergence problem of traditional GAMP algorithm in high-speed mobile scenarios because the channel non-zero elements in Delay-Doppler domain do not obey the independent and identical distribution.It is worth mentioning that the Bit Error Ratio（BER）performance of the converged IGAMP algorithm is better than that of LMMSE algorithm,MRC algorithm and MP algorithm.Finally,this paper uses the Discrete Zak Transform（DZT）to optimize the structure of the OTFS system,and uses Zero Padding（ZP）in the Delay-Doppler domain to obtain a simplified ZP-OTFS input-output relationship in Delay-Doppler domain,Delay-Time domain and time domain.At the same time,the research of detection algorithms in this paper extends from a single Delay-Doppler domain to time domain,Time-Frequency domain and Delay-Time domain.Since the traditional LMMSE algorithm in Delay-Doppler Domain needs to reverse the whole Delay-Doppler channel matrix,this paper uses the block diagonal structure of the time domain channel matrix to realize the low-complexity block LMMSE detection algorithm in time domain without losing BER performance.In this paper,a low complexity MRC detection algorithm in the Delay-Time domain is implemented by transforming a lot of matrix multiplication in the traditional MRC algorithm of Delay-Doppler domain into multiplication of element vectors in the Delay-Time domain.In addition,in order to further improve the performance of the OTFS system,this paper combined the detection algorithms of the ZP-OTFS system with Low-Density Parity-Check（LDPC）codes to improve the BER performance of the system,and two combinations of OTFS detectors and LDPC decoder are studied.One is the cascade scheme,which mainly through soft output processing of LMMSE algorithm,MP algorithm,IGAMP algorithm in Delay-Doppler domain and MRC algorithm in Delay-Time domain,to realize the cascade of signal detector and LDPC decoder in each signal domain.Another is the iterative scheme,in the iterative scheme,this paper studies the Iterative Decoding feedback-MRC（IDF-MRC）algorithm based on LDPC codes with hard decision and the IDF-MRC based on initial estimation（IE-IDF-MRC）algorithm.Both algorithms use the hard decision bit information output by decoder as the initial estimate of detector in the outer loop,but fail to use the soft information output by decoder,resulting in certain BER performance loss.Aiming at the problems of IDF-MRC algorithm and IE-IDF-MRC algorithm,this paper proposed IGAMP algorithm based on LDPC codes iterative decoding feedback（IDF-IGAMP）.IDF-IGAMP algorithm makes full use of the iteration between the symbol-level soft information output by IGAMP detector and the bit-level soft information output by LDPC decoder,which can effectively reduce the loss of useful information in the outer loop.IDF-IGAMP algorithm is superior to the detection algorithms of the cascade scheme,IDF-MRC algorithm and IE-IDF-MRC algorithm in BER performance.