Using weights as tools in handling systems instability in model-reality optimal control algorithm

Abstract

The Dynamic Integrated Systems Optimization and Parameter Estimation (DISOPE) is an algorithm for solving nonlinear dynamic optimal control problems. It uses the linear quadratic models to approximate the real problem and the identities to model the weights Q and R of the performance index. The algorithm then reverts to using two mechanisms, namely the convexification terms and scalar gains to improve convergence. These schemes work well most of the time but when unstable model plant dynamics is involved, the mechanisms fail to perform. This paper shows that stabilization of the plant dynamics with proper weights before attempting the algorithm is a better way of forging convergence. Furthermore this procedure will supersede the use of the scalar convexification factors.