Machine Learning Applied to Blockage Classification in Automotive Radar

Abstract

Detection of radar blockage is a critical safety function for automotive radar. In this paper, we report on a machine-learning approach to classify the blockage condition in automotive radar, using detection data. We consider logistic regression, tree-bagging, and neural network approaches. We used pruning to reduce the size of the neural network to make it a viable option for embedded processors with limited memory. The results show that the classifiers, especially the neural network, can achieve high accuracy with a low false-alarm rate.