Automotive radar interference suppression using improved dual-path convolutional recurrent network

Mutual interference among co-frequency automotive radars severely degrades detection performance in autonomous driving. This paper proposes an improved Dual-Path Convolutional Recurrent Network (DPCRN) with a novel Dynamic Tanh (DyT) layer that replaces conventional normalization modules. The DyT layer emulates normalization via learnable parameters, eliminating real-time statistical computation and reducing latency by 38.5%. Trained on the ARIM-v2 dataset simulating multi-radar scenarios, the model achieves a 20.1 dB output MSNR with near-identical suppression capability to DPRNN while halving processing time. This end-to-end framework significantly enhances real-time interference mitigation for 76–79 GHz frequency-modulated continuous-wave (FMCW) radars.