Autonomous High-speed Drift Beyond Stability Limits
Using reinforcement learning approach to optimize the race car performance under extreme conditions.
Highspeed sideslip cornering, also known as drifting, is a kind of technique to stabilize the vehicle beyond the handling limit with a large steering angle. Professional drivers use this technique in drift competitions performing accurate control over the vehicle’s gesture, position, as well as sideslip. But the researcher does not assume this is merely a technique used for stunt and fun. A similar approach could be applied to help the autonomous cars to extend the envelope of stable handling limits and as a result, improve the overall driving safety.
We present a methodology and plans to analyze the dynamics of an autonomous electric car during drifting and to develop a novel controller that can regulate the car along a target path. The researcher will build a computer simulator to simulate the dynamics of a rear-drive electric car by building a force-based single-track model with equations of motion. Besides, a controller to control the car drift along a target path will be developed. The task of this controller is composed of path tracking and sideslip stabilization.
|
|