Stochastic Model Predictive Control for Autonomous Emergency Integrated Braking and Steering System

Abstract: Controlling both longitudinal and lateral dynamics of an autonomous vehicle confronting possible collision requires ecient perception of surrounding environment and optimal use of its maneuverability. In emergency, separate steering or braking may not be able to avoid collision or ensure stability of the vehicle. This paper presents an integrated controller that combines classical Autonomous Emergency Braking (AEB), front steering and differential braking, is capable of deciding either to steer, brake or both to avoid collision with stability ensured. Considering imperfect modelling of the controlled vehicle and uncertainty of motion of surrounding vehicles, Stochastic Model Predictive Control (SMPC) method is used to formulate the proposed controller with an Interacting Multiple Model Kalman Filter (IMMKF) that estimates maneuver of other vehicles and a pair of fused camera and radar that measures current states of them. The effectiveness of the proposed controller is evaluated via simulation on Matlab/Simulink and hardware in the loop (HiL). The HiL platform is constructed with a real controller running on Nvidia Jetson Nano and a virtual controlled object (a high-delity vehicle model in CarSim). The communication between controller and controlled object is empowered by Robot Operating System (ROS) and Matlab/Simulink.