LPV-MPC on the Berkeley Autonomous Race Car

In this video we show our last work on autonomous vehicles. In this case, we present the result of controling a racing scaled car at the MPC-Lab in UC Berkeley, California.

The trajectory planning and automatic control are based on the Model Predictive Control (MPC) strategy. Particularly, to solve this algorithm in real time (30 Hz) we propose a LPV formulation of the prediction car model. This allow us to represent the non-linear model as a set of future simulated instanciations (without using linearization) and, hence, to solve a convex optimization problem. The technique is called LPV-MPC.