Learning Robot Force/Position Control for Repetitive High Speed Applications with Unknown Non-Linear Contact Stiffness
Sprache des Vortragstitels:
86th Annual Meeting of the International Association of Applied Mathematics and Mechanics (GAMM)
Sprache des Tagungstitel:
In many industrial applications manipulators are used to perform repetitive force controlled tasks. Typically such tasks are polishing, grinding, assembly as well as endurance testing of machine parts. The repetitive nature of such tasks allow for using iterative learning control (ILC) methods  or adaptive learning feed-forward control . For such a controller it is necessary that the feed-back controlled system is stable, so that a learning feed-forward control, for example, minimizes a resulting error from one repetition to the next. Considering a task, where a robot processes the same kind of workpiece in a recurring manner, the end-effector of the robot has to provide a predefined contact force while following a trajectory along the workpiece. To achieve this goal, a parallel force/position robot control, as suggested in , is best suited.