Control-gain estimation and disturbance rejection for nonlinear mechatronic systems: Design and experimental validation

Abstract

Accurate control-gain estimation and disturbance rejection are important for practical nonlinear mechatronic systems operating under uncertain actuation and limited sensing conditions. In many engineering applications, the effective control gain may vary due to actuator aging, load changes, efficiency degradation, or abrupt operating-condition variations, while repeated nominal-gain tuning increases implementation cost and limits system adaptability. This paper proposes a position-measurement-based adaptive extended state observer framework for online nominal control-gain estimation and disturbance rejection for a class of uncertain nonlinear mechatronic systems. The proposed method synchronously estimates the system states, unknown nominal control gain, and lumped disturbance without requiring velocity measurement, explicit system models, direct disturbance information, or prior nominal-gain calibration. To improve robustness against measurement noise and differentiation-induced fluctuations, a first-order filtering operation and a sliding-window integral mechanism are incorporated into the adaptive law. The estimated nominal control gain and reconstructed disturbance are further embedded into a continuous sliding mode control law to improve tracking performance, disturbance rejection capability, and control input smoothness. Boundedness and input-to-state stability of the closed-loop system are analyzed. Real-time experiments on a series elastic actuator platform are conducted under different gain settings, reference trajectories, external disturbances, and time-varying gain perturbations. The experimental results demonstrate that the proposed method reduces the dependence on manual gain tuning and improves tracking accuracy, robustness, and practical control performance under limited measurement conditions.