Towards a supervised robust control of a manipulator

Abstract

We present in this paper the global frame of a supervised robust control for a robot manipulator devoted to contour tracking. As usual, the global problem is separated in path planning problem and robust control loop synthesis problem. The aim of the Multiple Reference Model Robust control is to guarantee robust performances, including robust stability, regulation and tracking. As in any other robust control method, the level of performance must be released as the uncertainties are growing, so that some sources of uncertainty as aging, changing of dimensions from one machine to the other, etc... must be avoided: periodical adjustment of the robust control law must be done either when the performance specification is no more reached, or when a more accurate model is identified: this adjustment can be devoted to a supervisor. On another hand, the robust control theory involved assumes the linearity of the controlled plant, so that the path planning solver must compute only admissible trajectories. Usually, the path planning, through global optimisation, uses conservative bounds on speed and acceleration, so that performances are lowered. As shown on simulation results, the introduction of supervised local path planning solver permits us to adjust periodically the constraints, so that linearity can be guaranted almost everytime, and performances can be increased.