Inspiration
Drifting cars is cool. Control systems are also cool, so building a control system to drift a car is extra cool.
What it does
Uses Model Predictive Control to simulate and control a Porsche CS to drift on the F1 Hockenheim track.
How we built it
From Vdrift opensource software, isolated relevant car and environment data utilizing Replay function, calculated various measures and sampled inputs, set up a model predictive control system to determine next least costly action while prioritizing drift (constraints), rewired Vdrift AI agent to use our model's controls.
Challenges we ran into
Environment and car data were given in various units and formats that were inconvenient to work with. C++, in which Vdrift is written, does not have support for the necessary ML libraries so we had to pipe in Python.
Accomplishments that we're proud of
The car drifts.
Log in or sign up for Devpost to join the conversation.