Model Predictive Control for Trajectory Following with Actuator Degradation

It is often desirable to control an automated vehicle to follow a predetermined trajectory, be it a road, a path to avoid obstacles, or a flight plan to avoid interference with other aircraft. Traditional approaches to this problem have valued closed form solutions to keep computational time low. As a result, these controllers have required not only a desired trajectory, but also a set of control inputs that can achieve the trajectory. The controller then only needs to make small deviations from the designed control inputs to correct deviations from the desired trajectory. Although this approach often works very well, it does not handle actuator failure. In the case of failure the system is unable to recover because the control inputs and trajectory are coupled and the controller cannot find an alternative actuator input. In this project I propose to use model predictive control (MPC) to solve the trajectory problem and uncouple the trajectory from the requisite force inputs. This will create a more computational intensive controller. However, because convex solvers have become more sophisticated and embedded processors have become more powerful, implicity should no longer be a primary design parameter.