End-to-End Training of Neural Networks for Automotive Radar Interference Mitigation

TL;DR

This paper introduces a novel neural network training approach for FMCW radar interference mitigation that directly optimizes object detection maps, significantly improving detection performance while reducing model complexity.

Contribution

The paper proposes training neural networks directly on object detection maps using a continuous relaxation of CA-CFAR, and introduces separable convolution kernels to enhance efficiency.

Findings

Enhanced object detection accuracy on real-world radar data

Reduced neural network complexity with separable convolutions

Outperformed traditional signal processing interference mitigation methods

Abstract

In this paper we propose a new method for training neural networks (NNs) for frequency modulated continuous wave (FMCW) radar mutual interference mitigation. Instead of training NNs to regress from interfered to clean radar signals as in previous work, we train NNs directly on object detection maps. We do so by performing a continuous relaxation of the cell-averaging constant false alarm rate (CA-CFAR) peak detector, which is a well-established algorithm for object detection using radar. With this new training objective we are able to increase object detection performance by a large margin. Furthermore, we introduce separable convolution kernels to strongly reduce the number of parameters and computational complexity of convolutional NN architectures for radar applications. We validate our contributions with experiments on real-world measurement data and compare them against signal processing interference mitigation methods.