Adaptive-optimal control of constrained nonlinear uncertain dynamical systems using concurrent learning model predictive control

A concurrent learning adaptive-optimal control architecture for constrained aerospace systems with fast dynamics is presented. Exponential convergence properties of concurrent learning adaptive controllers are leveraged to guarantee a verifiable learning rate while guaranteeing stability in presence of significant modeling uncertainty. Radial Basis Function based adaptive elements are incorporated to approximate the uncertainty. The architecture switches to online-learned model based Model Predictive Control after an online automatic switch gauges the confidence in parameter estimates. A new switching metric ensures that the control architecture only switches to the model-based optimal controller if the uncertainty is approximated over the whole neural network operating domain. To achieve this a novel point selection algorithm for concurrent learning is presented. Numerical simulations on a wing-rock problem establish the effectiveness of the architecture.