Real-time Switched Model Predictive Control of a Twin Rotor System

In this paper, we present a Switched-Decentralized Model Predictive Control (S-DMPC) scheme for the real-time flight control of a Twin Rotor System (TRS). Due to highly nonlinear dynamics and unavailability of full-state information, the linearized TRS model contains modelling (linearization) errors which cause poor state-estimation and lead to controller performance degradation. In order to address this problem, we propose an MPC-based control scheme in which, the associated (linear) observer's gain is switched according to the level of linearization error present in the model. Through experimental results, we show that the proposed S-DMPC scheme significantly improves TRS tracking in comparison to the Centralized-MPC as well as the individual (non-switched) Decentralized-MPC schemes.