Machine Learning Approach for Obstacle Classification in Automotive Radar
Sujith K Suresh, Bhavya Puttaswamappa, Karthikeyan Rajarathinam, Prashant V. Kamat · 2022
Machine Learning and Deep Learning have made a significant contribution in camera sensor perception, but not much to the Radar sensor, because of its time-series nature of the data and system complexity. Radar is mostly used for object detection and tracking, perception using Radar is not explored as camera sensor. Our previous work on the moving object classification [6] is another attempt for introducing machine learning in Radar systems. In this paper, we discuss the different machine learning/deep learning architecture-based obstacle classification in Radar data for automotive application. We propose novel architectures that can handle time series and instantaneous data simultaneously for classification. Neural Network, LSTM, 1D-CNN, and hybrid architecture are explored in this paper.