Learning-based Stochastic Model Predictive Control for Autonomous Driving

Autonomous driving in urban environments requires safe con- trol policies that account for the non-determinism of moving obstacles, for instance, the intention of other vehicles while crossing an uncontrolled intersection. This thesis addresses the aforementioned problem by proposing a stochastic model predictive control (SMPC) approach. In this approach, we consider robust collision avoidance as a constraint to guar- antee safety and a stochastic performance index that will in- crease the quality of the closed-loop tracking by ignoring the unlikely obstacle configurations that could occur. We com- pute the probabilities associated with different obstacle tra- jectories by training a classifier on a realistic dataset gener- ated by the microscopic traffic simulator SUMO and show the benefits of the proposed stochastic MPC formulation in a sim- ulated real intersection. This thesis is divided into two parts: first, discuss the formulation of the existing control algorithm and our proposed approach, and second, the scenario predic- tion of the obstacle vehicles.