Stochastic Predictive Control of Autonomous Vehicles in Uncertain Environments

Autonomous vehicles operating in dynamic urban environments must account for the uncertainty arising from the behavior of other objects in the environment. For this purpose, we develop an integrated environment modeling and stochastic Model Predictive Control (MPC) framework. The trade–off between risk and conservativeness is managed by a risk factor which is a parameter in the control design process. The environment model consists of an Interacting Multiple Model Kalman Filter to estimate and predict the positions of target vehicles. The uncertain predictions are used to formulate a chance–constrained MPC problem. The overall goal is to develop a framework for safe autonomous navigation in the presence of uncertainty and study the effect of the risk parameter on controller performance. Simulations of an autonomous vehicle driving in the presence of moving vehicles show the effectiveness of the proposed framework.