Research On Key Technologies Of OTFS Receiver

Orthogonal time frequency space(OTFS)is a two-dimensional modulation technology.The basic idea of OTFS is to transform the fast time-varying channel in the time-frequency domain into the time-invariant channel in the delay-Doppler domain through two-dimensional Fourier transforms,so as to combat the dynamics of the timevarying channel.Compared with orthogonal frequency division multiplexing(OFDM)technology,OTFS is not sensitive to the time-varying channel,so it is more suitable for the high-speed mobile scene,and has attracted the attention of the researchers.In this paper,signal detection technology and channel estimation technology in the OTFS receiver are studied.This paper first introduces the research background of OTFS,and points out the importance of OTFS in solving the communication problems in the future high-speed mobile scene.Then,the fast time-varying channel model and the structure of transmitter and receiver of the OTFS system are introduced.Finally,the characteristics of the equivalent channel matrix,which are used in signal detection and channel estimation of the OTFS system,are introduced and analyzed.Secondly,due to the high complexity of the message passing(MP)algorithm which is an OTFS signal detection algorithm,a generalized approximate message passing(GAMP)algorithm based on deep learning is proposed.Specifically,combined with the idea of deep unfolding,each iteration of the low complexity GAMP algorithm is unfolded into a layer of the neural network,and the detection performance is improved by learning the damping factors in the network.The optimized damping factors can be directly used in the original GAMP algorithm without additional complexity.In order to verify the effectiveness of the algorithm,the GAMP algorithm based on deep learning is simulated.The simulation results show that the performance of the algorithm is better than that of the classical damped GAMP algorithm,MMSE algorithm and MP algorithm.Finally,this paper studies the channel estimation technology in OTFS multiple access.Firstly,the channel estimation algorithm based on impulse pilots and the channel algorithm based on compressed sensing are described,and the performance advantages of the latter are compared through simulation.Then,aiming at the problem of channel estimation performance loss caused by atomic index deviation in OMP(Orthogonal Matching Pursuit)algorithm which has the best performance among the channel estimation algorithms based on compressed sensing,inspired by the deep learning method,a algorithm based on deep learning is proposed,where a Res Net is embedded in the original OMP algorithm.The algorithm models the atom selection problem as a classification problem,and uses data-driven method to reduce the interference of noise to atom selection,so as to optimize the performance of channel estimation in OTFS receiver.Simulation results verify the effectiveness of the channel estimation algorithm based on deep learning.