基于动态低秩张量分解PLS的信道估计算法研究OA

As 5G/6G deployment extends into high-speed mobile scenarios,severe channel estimation errors arise due to high Doppler shifts and pilot sparsification.Traditional methods like Least Square(LS)and Linear Minimum Mean Square Error(LMMSE),reliant on static statistical characteristics,struggle to adapt to time-varying properties,resulting in significantly limited performance.To this end,a Dynamic Low-rank Tensor Decomposition combined with Partial Least Squares(DLRTD-PLS)algorithm is proposed.This approach tracks time-varying channel characteristics via recursively updated tensor decomposition,suppresses noise interference through frequency-domain smoothing and dual regularization constraints exploiting multipath sparsity,and enhances estimation robustness by optimizing latent variable projection using partial least squares.Theoretically,the algorithm achieves lower computational complexity by avoiding the matrix inversion operations inherent in conventional methods.Simulations demonstrate that under conditions of 5 dB Signal to Noise Ratio(SNR)and a Doppler shift of 1 kHz,the proposed DLRTD-PLS algorithm achieves a 75%reduction in Mean Square Error(MSE)compared to LS,with memory usage merely 0.65%of that required by Kalman filtering.Furthermore,it delivers lower error than the Basis Pursuit(BP)algorithm under low pilot density.This algorithm offers a novel approach for real-time estimation of highly dynamic channels in space-air-ground integrated networks and holds strong potential for further application in edge lightweight deployment scenarios,signifying high engineering value.