Deep Learning-based Person Detection on a Moving Robot

Abstract

This paper presents a 60 GHz MIMO radar system on a moving robot for person presence detection. Due to the movement of the robot, many environmental influences are added. For this purpose, various scenarios were created and 8,000 data frames were recorded and processed. A convolutional neural network (CNN) was trained and evaluated using this data. Ultimately this low-cost system achieves 97% accuracy in classifying whether a person is present or absent.