feb logo

Autonomous

The Autonomous subteam's job is to take the electric race car that Formula Electric at Berkeley builds and make it drive itself. At competition that means the Formula Student driverless events: a timed acceleration run, the figure-eight skidpad, an autocross lap on a course the car has never seen, and trackdrive, ten laps of a circuit marked only by blue and yellow cones that the car must discover on its first lap and then race. We own the whole pipeline. Perception detects the cones in the lidar and camera data and classifies their colour. SLAM builds a map from them and localises the car on it. Mapping and cone ordering turn the scattered cones into left and right track boundaries. Raceline optimisation plans the fastest line through those boundaries, and a model-predictive controller plans the steering and throttle a few seconds ahead, using a measured model of the car, to follow that line as fast as the tyres allow. We work with the mechanical and electrical subteams to get our code onto the actual car's steering, throttle and brakes, and we develop and test everything in simulation and on a 1:10 scale RC platform before it ever runs on the full-size car.

The Pipeline

notion image

F1-Tenth

 

Simulation