Autonomous Racecar Stack

Vision-based lane following and RRT/PRM path planning for a 1/10-scale autonomous racecar.

For 6.141 Robotics: Science and Systems, I helped program our racecar, named Alfredo, to localize, plan, and navigate autonomously across a series of increasingly complex tasks.

By the end of the course, Alfredo drove laps autonomously on an indoor track at up to 5 m/s, staying in its lane with computer vision and a pure-pursuit controller. For navigation, it planned paths through a loaded map with bidirectional rapidly-exploring random trees (RRT) or probabilistic roadmaps (PRM).

Wall following

Cone parking

Line following on a zigzag course

Line following on a circular course

Stack

ROSC++