Approximate Message Passing-Enhanced Graph Neural Network for OTFS Data Detection

TL;DR

This paper introduces an AMP-GNN detector for OTFS modulation that iteratively refines symbol estimates by coupling an AMP module with a GNN module. It addresses the inability of purely learned detectors to fully exploit symbol priors and tackles high inter-Doppler interference (IDI) complexity through a learning-based IDI approximation and a real-valued reformulation. By integrating AMP priors and a simplified pair-wise MRF within a GNN, the approach achieves notable BER gains over MP, AMP, and standard GNN detectors with substantially reduced computational load. The proposed method demonstrates strong performance under both perfect and imperfect CSI, highlighting its practical potential for robust, low-complexity OTFS data detection in high-mobility scenarios.

Abstract

Orthogonal time frequency space (OTFS) modulation has emerged as a promising solution to support high-mobility wireless communications, for which, cost-effective data detectors are critical. Although graph neural network (GNN)-based data detectors can achieve decent detection accuracy at reasonable computational cost, they fail to best harness prior information of transmitted data. To further minimize the data detection error of OTFS systems, this letter develops an AMP-GNN-based detector, leveraging the approximate message passing (AMP) algorithm to iteratively improve the symbol estimates of a GNN. Given the inter-Doppler interference (IDI) symbols incur substantial computational overhead to the constructed GNN, learning-based IDI approximation is implemented to sustain low detection complexity. Simulation results demonstrate a remarkable bit error rate (BER) performance achieved by the proposed AMP-GNN-based detector compared to existing baselines. Meanwhile, the proposed IDI approximation scheme avoids a large amount of computations with negligible BER degradation.

Figures (4)

• Figure 1: . AMP-GNN-based detector for OTFS systems.
• Figure 2: . BER vs. SNR. The solid and dashed curves represent cases with perfect ($\sigma_e^2=0$) and imperfect CSI ($\sigma_e^2=-20$ dB), respectively, where $\overline{h}_{p}$ is contaminated by additive Gaussian noise $n_{e,p} \sim \mathcal{CN}(0, \frac{1}{P}\sigma_e^2)$.
• Figure 3: . An example of overlapping node-pairs, where $l_p$ and $k_p, p=1, 2, 3$ denote the integer delay and Doppler shift indexes, respectively.
• Figure 4: . Illustrative examples of the signal model with and without IDI approximation in the delay-Doppler domain, where $l_p$ and $k_p, p=1, 2$ denote the integer delay and Doppler shift indexes, respectively.