Joint control for flexible‐joint robot with input‐estimation approach and LQG method

Abstract

In this work, the input-estimation (IE) algorithm and the linear quadratic Gaussian (LQG) controller are adopted to design a control system. The combined method can maintain higher control performance even when the system variation is unknown and under the influence of disturbance input. The IE algorithm is an on-line inverse estimation method involving the Kalman filter (KF) and the least-square method, which can estimate the system input without additional torque sensor, while the LQG control theory has the characteristic of low sensitivity of disturbance. The design and analysis processes of the controller will also be discussed in this paper. The joint control of the flexible-joint robot system is utilized to test and verify the effectiveness of the control performance. According to the simulation results, the IE algorithm is an effective observer for estimating the disturbance torque input, and the LQG controller can effectively cope with the situation that the disturbance exists. Finally, higher control performance of the combined method for joint control of the robotic system can be further verified.to the external force into account. However, the operating environments for many robot manipulator systems become more and more complicated, and the requirements for accurate control performance are getting higher. For example, the recoil acting on the weapon manipulator of the unmanned ground vehicles (UGVs) is dynamic when the weapon is in action. If the disturbance torque input cannot be adequately handled, the system will have difficulties to achieve higher control performance [1]. Moreover, when the auxiliary rehabilitation robot cannot sense the reaction force from the patients and provide feedback to its control system for adaptive adjustments, an adequate reaction will not be made to provide a comfortable rehabilitation process to the patients [2]. In view of the above issues, it is fairly important to improve the ability of the estimation of the disturbance torque and to design a robust control system for the robot manipulator system.In the design of the robotic manipulator system, two methods for solving the force feedback problems are commonly used. One is using the torque sensor to measure the disturbance torque [3], and the other is utilizing an inverse method to estimate the disturbance torque according to the observation of the system states (without torque sensor) [4][5][6][7][8]. Although the torque sensor can measure the disturbance torque accurately, it is very expensive and will extensively increase the cost of the overall system, especially when the machinery system is more complex. Various kinds of controllers without the torque sensors have been proposed in the literature. The disturbance observer [9-13], the adaptive robust control (ARC) [14,15], and the enhanced internal model control (EIMC) [16] are several main examples. Recently, the generalized disturbance compensating framework based on the Lyapunov redesign, named the robust internal loop compensator (RIC), was proposed to show th...

A Tracking Control of a Flexible-Robot Including the Dynamics of the Induction Motor as Actuator

“…[3] addresses the tracking control problem of flexible robot arms improving the damping of the system through robust control techniques where the whole state is available. [4] develops the linear quadratic Gaussian method from the optimal control theory, in combination with an input-estimation algorithm, to enhance the ability of disturbance torque input estimation in the joint control of a flexible-joint robot system.…”

Optimal zeroing dynamics with applications to control of serial and parallel manipulators

“…Optimal control theory, as a powerful method that has been widely applied to robot manipulators, can take different kinds of subtasks including disturbance tolerance into consideration. () Linear‐Quadratic‐Gaussian (LQG) controller, as an popular and effective controller derived from optimal control theory, incorporates the advantages of Kalman filter, which can improve the tolerance to Gaussian white noise . For linear systems, the optimal controller also becomes linear when the objective function is quadratic and the constraint can be expressed as a set of linear differential equations.…”

Application of the state‐dependent Riccati equation for flexible‐joint arms: Controller and estimator design

“…Thereafter, observers were explored for deterministic autonomous and nonautonomous systems . Ji et al employed the linear quadratic Gaussian method to control a robotic arm . Alvarez‐Ramirez utilized a class of PID controller for a robot with elastic joints and proved its semi global asymptotic stability.…”