Model Predictive Control

As an area of concentration, a model predictive control strategy is proposed to solve nonholonomic vehicle regulation problem. The generated trajectory can consider the constraints from input saturation directly in the optimal control computation.

Model predictive control or receding horizon control is a kind of control algorithm suitable for the case in which pre-computation of a control law is not feasible. In this control strategy, at each sampling instant, the current control law is obtained by solving a finite horizon open loop optimal control problem. In our research, based on nonlinear discrete system model, a stable controller is designed in terminal region for each predictive length. Parameters of the controller are chosen by considering the control stability requirement. According to the terminal region constraints, the corresponding requirement on discrete step size is given. The input saturation and minimum turning radius are considered as constraints in the optimal problem so that the given trajectory is feasible for a real car. Obstacle avoidance is realized with the proposed control architecture.

---