Ahmad, Alghooneh (University of Waterloo), Panahandeh, Pouya (University of Waterloo), Shaker, George (University of Waterloo), Khajepour, Amir (University of Waterloo)

Robust Radar Object Detection Using HD Map Likelihoods and Information

Scheduled for presentation during the Invited Session "Trustworthy Diagnosis and Prognosis in Connected, Cooperative and Automated Mobility" (WeBT10), Wednesday, September 25, 2024, 14:50−15:10, Salon 18

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on December 26, 2024

Abstract

We present an innovative approach that integrates radar data with high-definition (HD) maps to generate a robust cost-map for navigation systems. While radars are essential for perception through instant velocity measurement, high range, and their robustness to weather conditions, they often introduce errors such as false alarms and clutters that could mislead the cost-map generation. Through a detailed analysis of these errors, including their typical spatial patterns on the HD map, we developed a precise classifier capable of removing these misdetections from actual objects. To aid navigation, we create a cost-map by combining HD map details with radar detection data. This cost-map serves a dual purpose: guiding the path planning process and offering a correction layer to the radar's interpretation of the surroundings. Experimental validation of our methodology is conducted on a dataset collected using our autonomous shuttle-bus, WATonoBus, under various weather conditions. Find a sample of our work's performance by visiting href{https://youtu.be/cNb_OR19BQk?si=7CIX-7IRJfFR1bm6}{MVS-lab channel}.