Model-free Object Detection and Tracking

The main goal of this project is to implement solution for model-free detection and tracking of moving objects using 2D laser scanner on the F1Tenth 1/10 scale autonomous race car system. Specifically, object tracking is achieved using 2D LiDAR of the F1Tenth vehicle. This project builds on the work of Dominic Zeng Wang and his collaborators, who proposed a model-free LiDAR-based object tracking system that separates out the static background to help track dynamic objects in a principled and straightforward way. We discuss the implementation approach and decisions that were used in development of the car and present an experimental evaluation of our solution. Given the restrictions posed by the pandemic, this project will heavily leverage virtual environments of the F1Tenth team’s proprietary gym environment.