DBSCAN
Oct. 2022
Overview
Autonomous vehicles use many sensors to understand the environment around them. However, individual sensors will have noises, ultimately increasing the uncertainty for autonomous driving algorithms. By using sensor fusion, one can combine data from all sensors and decrease the uncertainty in measurements. To prepare for the ecoCar competition, I integrated the DBSCAN algorithm to cluster lidar, radar, and camera detections on an ego vehicle in MATLAB driving scenerios.
Key Highlights
• This implementation can cluster vehicles into correct tracks with up to 85% validity
• The algorithm was analyzed on different test cases to discover new ways to improve the implementation
• The DBSCAN algorithm provided a starting base for the McMaster ecoCar CAVS team to develop more advanced sensor fusion algorithms