A new control method of flexible-joint manipulator with harmonic drive

Abstract

Considerable elasticity and nonlinear friction caused by harmonic transmission challenge the performance of flexible-joint manipulators. The uncertain dynamics of manipulator and the inadequate measurable states also limit the controller design. A new control method is proposed to address these problems, achieving the precise motion control of the flexible-joint manipulator. The method consists of three cascaded controllers: an adaptive controller, a torque-tracking controller, and a motor controller. The adaptive controller was adopted to generate the desired torque ensuring the robustness for uncertain dynamics. The torque-tracking controller derived the position compensation for motor control according to the torque error. As the elastic torque is under control, the vibration caused by harmonic drive can be eliminated. The motor was controlled based on poles-assignment method and friction compensation. The Kalman observer based on the Brownian motion model observed both velocity and the high-order derivatives of torque sensing. The stability of the control method was strictly proved. Calibration was performed on each joint to obtain the required joint stiffness and motor friction parameters. The control method was verified on a single joint and the frequency response of the system was obtained. The results show that the controller has good performance. The controller was realized on the self-developed seven-degree-of-freedom manipulator. The results reveal that the controller has high-precision tracking performance.

A model-based PID-like motion control method for flexible-joint manipulator with harmonic drives

“…As differential operation will cause strikes and noise, a Kalman filter based observer is adopted to estimate the high order derivatives of torque. 19 The control law for motors consists of three parts, as:…”

“…, Vf and Ĉf are the estimation of V f and C f respectively, which are obtained by parameter identification with accuracy confidence of 15%. 19 The state equation for ẑ is given by:…”

“…It can be seen that, with 5% of modeling error, the maximum tracking error is about 2 mm, which is still acceptable. Actually, dynamics parameters can be off-line identified, 19 based on which PID algorithm can be implemented instead of more complex adaptive algorithm to achieve higher efficiency.…”

“…Besides the joint elasticity, the nonlinear friction inside motors, including Coulomb friction and viscous friction also impair the high-speed motion performance. Although, the viscous friction can be identified during experiment with single joint, 19 it's difficult to compensate the Coulomb friction force, which is affected by payloads dramatically. H 1 control and slide-mode control methods on motors give solutions for rejecting the nonlinear friction.…”

Design of Auxiliary Feeding Device Based on Single Chip Microcomputer Control System

“…[3][4] In addition to traditional sensors such as speed, speed and acceleration, assembly and welding manipulators also use sensors such as vision and power, while remote control manipulators use many sensors synthesis technologies such as vision, sound, power, and sensation for environmental modeling and decision control. [5][6] It is designed as an auxiliary feeding manipulator, so it is necessary to realize automatic control and automatic detection in some positions. Electronically controlled hydraulic lever and sensor are used, both of which can realize automatic control and automatic detection.…”