Oxford Brookes Racing Autonomous

MPC Controller for Autonomous Vehicles

I am the leader of the Control Department, responsible for developing the MPC (Model Predictive Controller) for our Formula Student competition vehicle. Our team's task is to design a controller capable of managing throttle, braking, and steering using predictive algorithms.

The MPC is chosen for its ability to optimize vehicle behavior in real time while considering constraints like tyre grip and vehicle dynamics. Compared to traditional controllers like PID or PPC, MPC provides superior stability and performance by anticipating future states and adjusting accordingly. Our model uses a simplified bicycle dynamics base and includes realistic tyre behavior based on Formula Student data.

Technically, the MPC computes control actions by minimizing a cost function over a prediction horizon, taking into account both lateral and longitudinal dynamics. It integrates constraints from the Pacejka tyre model and uses data from the path planner to calculate curvature, velocity targets, and optimal steering angles at each timestep.