mmWave Radar Based Approach for Pedestrian Identification in Autonomous Vehicles

Abstract

In this paper, a mmWave radar algorithm will be presented to detect moving pedestrians and differentiate them from other vehicles on the road. This is an important problem for advanced driver assistance systems (ADAS), and autonomous vehicles (AV) applications, especially for pedestrians safety. This approach will leverage radar ability to operate under different lighting and environmental conditions (e.g. day, night, fog, etc.) without being computationally extensive. To test the proposed approach, a setup that involves Texas Instruments (TI) AWR1642 77 GHz radar kit and DCA1000 FPGA board was used. In addition, a recently developed Python based testbed, developed by the authors, was used for control, data acquisition, and real-time processing of radar data (25 radar images per second without any dropped frames). Results show that the proposed approach can distinguish between moving pedestrians and vehicles with accuracy comparable to results obtained from cameras and lidars. This is especially important for hardware in the loop (HIL) type testing of high level algorithms in ADAS and AV systems.