Constrained control of flexible-joint lever arm based on uncertainty estimation with data fusion for correcting measurement errors

This study presents a novel uncertainty estimation algorithm that aims to enhance the accuracy of a reduced-order dynamic model for a flexible-joint lever arm (FJLA). The model uncertainties are compensated by adding a complementary term to the reduced-order model, upgrading it to the real model. The gyroscope bias in measuring the angular velocity is corrected by fusing the information of encoder/gyroscope. The effective data fusion is provided by adaptive adjustment of the estimator coefficient. The constructed model, with low order but rich content, is reliable for use in the position controller of FJLA, developed by a continuous predictive method. The control system adapts itself to real conditions and is cost-effective due to the use of fewer sensors. The saturation of motor torque, as the control input, is modeled within the structure of a constrained optimization problem solved by Karush–Kuhn–Tucker Theorem. The boundedness of the mean and covariance of tracking error and its derivative are demonstrated by stochastic analysis. The results of simulations and experimental implementations demonstrate the high efficiency of the proposed system in controlling the position of the FJLA under different trajectories. Moreover, the comparative results with the other methods show a great performance for the suggested control system in rejecting disturbances and other uncertainties under saturated control input.