Estudo comparativo de pacotes computacionais empregados na resolução de problemas de Controle Ótimo

Abstract

The Optimal Control Problem (OCP) consists of the control vector determination for the objective function minimization, subject to algebraic-differential constraints. In general, this problem can be solved considering two types of approaches, Direct and Indirect. The first consists in transforming the original problem into a non-linear programming equivalent. The Indirect approach consists of the application of optimality conditions, transforming the original problem into an equivalent in the algebraic-differential contour. Due to the inherent difficulty in solving the problem of the contour, the scientific community has used, preferably, the Direct Methods for the resolution of OCPs. In this context, several numerical packages that are based on the Direct approach to the resolution of PCOs can be found in specialized literature. Despite the wide variety of packages available, few are those that can be easily used by inexperienced users in the computational resolution of PCOs, and also few are the works that evaluate the efficiency and characteristics of these packages comparatively. In this context, this work aims to develop the COPILOTS (Basic Optimal Control Solver) package, which implements the trapezoidal and Hermite-Simpson colocations as discretization approaches for solving OCPs. A comparative study involving this package, FALCON and PSOPT was developed and the results obtained demonstrate that COPILOTS was able to obtain good results in comparison with the other evaluated packages. Finally, it is important to note that COPILOTS is a package designed for users with little experience in optimal control, easy to implement and use.