Generalized nonlinear optimal predictive control using iterative state-space trajectories: Applications to autonomous flight of UAVs

Abstract

Model Predictive Control (MPC) is a modern technique that, nowadays, encapsulates different optimal control techniques. For the case of non-linear dynamics, many possible variants can be developed which can lead to new control algorithms. In this manuscript a novel generic control system method is presented. This method can be applied to control, in an optimal way, different systems having non-linear dynamics. Particularly, in this paper, the proposed technique is presented in the context of developing a control system for autonomous flight of UAVs. This technique can be used for different types of aerial vehicles having any type of generic non-linear dynamics. The presented method is based on the use of iteratively defined optimal candidate state-space trajectories in global state-space. The method uses a generalized linearization process which, opposite to standard methods, does not need to be predefined in a certain equilibrium state but instead it is performed along any arbitrary state. The technique allows the inclusion of constraints with ease. The presented technique is used as a centralized control system unit that is able to control the full aircraft dynamics without the need of decoupling the system in different reduced modes. The technique is tested by making a Cessna 172 airplane model to perform the following autonomous unmanned maneuvers: climbing at constant speed to a desired altitude, heading change to a desired flight direction, and, coordinate turn.