RIMformer: An End-to-End Transformer for FMCW Radar Interference Mitigation

TL;DR

This work tackles mutual interference in FMCW radar by introducing RIMformer, an end-to-end Transformer-based interference mitigation method that operates directly on time-domain IF signals. The architecture combines dual multi-head self-attention with convolutional blocks in an encoder–decoder framework, uses relative positional encoding, and processes data via sliding-window subsegments to achieve robust interference suppression. A hybrid time–frequency loss is employed to better capture both time-domain accuracy and frequency-domain characteristics, with thorough ablation and measurement studies validating substantial SINR improvements and accurate target reconstruction over multiple baselines. The approach offers practical gains for reliable target detection in crowded radar spectra, enabling safer and more robust remote sensing and autonomous systems.

Abstract

Frequency-modulated continuous-wave (FMCW) radar plays a pivotal role in the field of remote sensing. The increasing degree of FMCW radar deployment has increased the mutual interference, which weakens the detection capabilities of radars and threatens reliability and safety of systems. In this paper, a novel FMCW radar interference mitigation (RIM) method, termed as RIMformer, is proposed by using an end-to-end Transformer-based structure. In the RIMformer, a dual multi-head self-attention mechanism is proposed to capture the correlations among the distinct distance elements of intermediate frequency (IF) signals. Additionally, an improved convolutional block is integrated to harness the power of convolution for extracting local features. The architecture is designed to process time-domain IF signals in an end-to-end manner, thereby avoiding the need for additional manual data processing steps. The improved decoder structure ensures the parallelization of the network to increase its computational efficiency. Simulation and measurement experiments are carried out to validate the accuracy and effectiveness of the proposed method. The results show that the proposed RIMformer can effectively mitigate interference and restore the target signals.

Figures (11)

• Figure 1: Demonstration of a scenario where FMCW radar system on an UAV faces potential interference from other radars.
• Figure 2: Chirp signals from other FMCW radars that are contained in the same frequency band are received, which cause the pulse-like interference in the time domain.
• Figure 3: The RIMformer adopts a typical encoder-decoder architecture. Within the decoders, the initial signal serves as the query, while the encoding outcome functions are the key and values. The results obtained from the decoders are amalgamated to form the ultimate output, which mirrors the shape of the input time-domain signal. Both the encoders and decoders exhibit identical structures, featuring key components such as dual multi-head self-attention mechanisms, feedforward mechanisms, and convolutional blocks.
• Figure 4: In the dual multi-head self-attention structure, the input data are split into different dimensions according to their inter-frame and intra-frame dimensions. Self-attention is then computed separately. The output is obtained by merging the two attention results and the residual.
• Figure 5: Convolutional block and details of the gated linear unit (GLU).