Abstract
In general, it is very difficult to control motion of mechanical manipulators when their hand grips an unknown object. The reason is that the dynamic characteristics for manipulators vary according to the payload of the object. In this case, mass, center of mass, and inertia tensor of the end link become unknown parameters, while other system parameters do not change.
This paper presents a new approach to trajectory control by using adaptive controller. Only unknown parameters are identified by an on-line recursive algorithm based on the least squares error criterion, though this method requires angular acceleration observation. The controller is calculated on-line, using an accurate dynamic model with estimated values for unknown parameters. Simulation results for a six degrees of freedom manipulator show that trajectory errors become very small in spite of the fact that the least square estimate is biased when observation noise of joint angular acceleration exists. The effectiveness of the proposed control method was verified by experimental results obtained for a two degrees of freedom manipulator.
