The Robotics Institute

• Adobe portable document format (pdf) (719KB)

Due to the modeling error or environmental uncertainties, robot motion may present a significant positioning error by using a conventional Computer-Torque Method. To improve tracking capability of robot manipulators, sliding mode control and nonlinear control algorithms have been introduced, but computation is costly, and thus a fast motion execution using simple computer sources is impossible.

To solve this problem, we present a composite control algorithm to control robot motion combining a discrete feedforward component and a continuous feedback component. The discrete feedforward component provides a nominal torque computed using the robot dynamics and compensates for dynamic coupling between the links. This part can be updated in a large sampling time, and can be computed off-line generally, thus real time computation is decreased. The continuous feedback control component uses a structure of Variable Structure System and provides a robust control to disturbances during the sliding mode. This part can be digitally implemented using a short sampling time, and thus a fast motion of a multi-degree freedom robot manipulator can be executed by using a simple computer, or even a single board computer with an 8-bit CPU.

The stability of the proposed multiple-rate control scheme is proven in the paper and efficiency of the control scheme has been demonstrated by simulations of a three-link robot subject to parameter and payload uncertainties.