Robot torque and position control using an electrorheological actuator

Abstract

An electrorheological (ER) clutch driven from a constant speed motor provides a steady torque independent of shaft angle and can be controlled by control of the applied field. Such an actuator avoids the ‘cogging’ variation in torque observed in d.c. servo-motors and is thus well suited to robot control applications, particularly in view of the very rapid time response of ER clutches (≍ 10⁻³ s). However, the relationship between applied field and torque is difficult to model, being both non-linear and time varying. Whereas the non-linearity can be shown to be relatively small, the time-varying characteristic has remained a problem.

In most controlled plants, a non-linear or time-varying characteristic can be mitigated by providing a closed control loop around the plant. A PID (proportional plus integral plus derivative)-based torque controller was developed and tested. This was shown to be stable with at least critical damping and to exhibit low steady state error. Design of the controller was facilitated by the identification of the open-loop transfer function of the ER actuator.

The ER actuator with torque feedback was used to position a small robot link. A second PID control loop responding to the error in the link position and tuned using the standard Ziegler and Nichols method was designed and tested. A steady state error of less than 0.75 mm was achieved with a 2 per cent settling time of 2.0 s. Finally, the link position was controlled using a single-loop controller with no torque feedback and a similar steady state error achieved with a 2 per cent settling time of 1.4 s. It is argued that the ER torque actuator is ideally suited to the actuation of robot joints where precise smooth movement is required.