Explicit Nonlinear MPC for Fault Tolerance Using Interacting Multiple Models

This article presents a novel algorithm for adaptive explicit nonlinear model predictive control (eNMPC) with applications to fault tolerance. In order to account for plant-model mismatch, under which fault tolerance applies, the controller's explicit solution is designed with multiple dynamic models representing various operating modes as opposed to a single system model. Each model is weighted by a parameter variable to be evaluated online as mode probabilities produced by an interacting multiple model (IMM). Weighting each potential system model allows the controller to use a dynamic model that best matches the current operating mode, thus mitigating the degrading performance brought on by plant-model mismatch. The developed strategy is validated on attitude maneuvers for a nonlinear spacecraft system in the presence of disturbances and two actuator faults, which are indicative of the system mode. Average root mean squared values on the tracking error and control effort over Monte Carlo simulations are used to evaluate the effectiveness of the proposed techniques. Results indicate eNMPC benefits from access to weighted system models and manages similar levels of tracking error to standard spacecraft controllers at the same or minimal control effort.

Automated summary

This article presents a new algorithm for adaptive explicit nonlinear model predictive control (eNMPC) with applications to fault tolerance. The controller's explicit solution is designed with multiple dynamic models to address plant-model mismatch, with each model weighted by a parameter variable determined by mode probabilities. The developed strategy is validated on attitude maneuvers for a nonlinear spacecraft system, showing that eNMPC benefits from weighted system models and achieves similar levels of tracking error to standard controllers.