Implementação de controle preditivo para rastreamento de trajetórias de um bicóptero com empuxo vetorial

Abstract

Vectoring the thrust generated by engines can increase the maneuverability of aircrafts enabling, for example, vertical takeoffs and landings. By adopting such strategy, it becomes possible to control separately attitude and position dynamics by manipulating the magnitudes and directions of the thrust forces. Thus, with a proper control system, it is possible to track position references while maintaining a constant attitude or to change the vehicle's attitude without affecting its position. In particular, Model-based Predictive Control (MPC) can be used for this purpose, as this technique applies to systems with multiple inputs and outputs. In MPC, the control input is determined by solving an Optimal Control Problem (OCP), which often include system constraints. Consequently, it is obtained an optimal control input which enforce constraint satisfaction. As a drawback, conventional MPC strategies can require a high computational cost to solve the OCP. This can prevent the implementation of an MPC controller, especially for fast dynamic systems. An alternative to reduce the computational cost is employ an explicit MPC. In this technique, the state space is partitioned in regions in which the optimal control action is piecewise affine. Thus, during operation, the optimal control input is calculated determining which region contains the current state and then applying the associated control law. In this context, work presents the implementation of an explicit MPC approach to control a bicopter equipped with a thrust vectoring mechanism. Simulations and experimental results show that the control strategy was able to track reference without steady state error with proper enforcement of input constraints in both conventional and explicit implementations. For this purpose, the MPC was designed considering a constant output disturbance, which was estimated by means of an observer. This technique allows compensating for model mismatches that acts similar to constant output disturbance in a steady-state. To evaluate the associated computational costs, the time to compute the control action by solving the OPC numerically and using the explicit approach was measured. It was found that the average time for the explicit MPC was not reduced. On the other hand, such approach provided a more uniform computation time throughout the operation.