Extended Kalman Filter Based Nonlinear Probabilistic Model Predictive Control of an Autonomous Motorcycle

This work focuses on the control of a nonlinear autonomous motorcycle model using Probabilistic Model Predictive Control (MPC). The main purpose of this work is to control the speed, yaw angle and roll angle of the motorcycle while keeping its balance during a maneuver. The Probabilistic MPC takes into account the probabilistic disturbances acting on the control inputs or the state variables. The cost function of the optimization of this proposed nonlinear probabilistic MPC is different than the cost of traditional deterministic MPC, such that, it contains terms that represent the mean of the state variable which is calculated with the help of an Extended Kalman Filter (EKF) and also the terms as the functions of covariances of the state variables and control inputs. The optimization algorithm is fed with these mean and covariance values every time instant, then the controller calculates the optimal control sequence to apply it to the system model. The speed of the motorcycle, the yaw angle and, the roll angle of the motorcycle are controlled with good efficiency.