Autonomous Mapping and Navigation for Small-Scale Car Racing

Abstract

We consider an autonomous car racing setting with limited information about the racing environ- ment thus requiring real-time localization, mapping, and control of our car when racing against opponents. We demonstrate the use of efficient navigation algorithms for autonomous car racing and obstacle avoidance when limited to onboard sensing and computation. We build off the open-source F1 Tenth platform by heavily modifying a one-tenth scale remote control vehicle, thus allowing our Robot Operating System (ROS) to interact with driving controls and onboard sensors simultaneously. Using LiDAR (Light Detection and Ranging), we use a Simultaneous Localization and Mapping (SLAM) algorithm to generate a map of the surroundings in real-time. The vehicle is able to avoid obstacles autonomously and use the generated map to determine its optimal speed. We validate our algorithms in simulation as well as real-world opponent racing.