Evolving granular control with high-gain observers for feedback linearizable nonlinear systems

Abstract

Feedback linearization control is a simple and effective strategy whenever a faithful model of the system and its states are available. Feedback linearization may suffer from the mismatches between the model used in the design and the actual system due to e.g. uncertain parameter values, parasitic dynamics, or because of the impossibility to measure some states of the system. To aleviate such an issue, we suggest a novel robust adaptive control approach using the evolving participatory learning algorithm together with a high-gain observer. The robust evolving granular high-gain observers (RegHGO) controller is suitable to control nonlinear systems that can be input-output linearized by feedback. The approach is robust to modeling mismatches and does not require full state availability because, once the system is in a suitable canonical form, a high gain observer can be constructed to supply the state information required for control. The usefulness and efficacy of the approach is shown using a fan and plate system, and an DC motor driven angular arm-position control. The fan and plate evaluates the controller in a regulation process, and the angular arm position control evaluates reference tracking perfromance. In both cases, time-varying parameter uncertainties disturb the closed-loop control system. Both, qualitative and quantitative performance evaluation of the RegHGO controller are done. Additionally, we compare the performance of the RegHGO controller with well-established methods such as exact feedback linearization with high-gain state observer and extensions. The results show that robust evolving granular control with high-gain observers achieves better performance than its counterparts.