Theoretical Aspects of Lightweight Robot Control

Abstract

Today’s industrial robot controllers only use sensors, (mainly encoders and tachometers), which are collocated with the actuators. To position the robot end effector, the desired end effector position and orientation are converted by inverse kinematic computations into joint angles. The joints are then driven to these angles, each joint using a proportional, integral, and derivative (PID) controller based on collocated encoder and tachometer feedback. The robot is presumed stiff enough such that the end effector is in the intended position and orientation. The two main limitations of collocated controllers are: 1) the robot links and joints must be very stiff, and therefore very heavy, in order to obtain some degree of precision. This limits the robot to slow speeds and/or requires high levels of drive power. 2) The inherent flexibility in the structure (link and/or joint flexibility) makes it impossible to achieve truly high precision. Moreover, due to the ever increasing specifications in terms of speed, acceleration, and terminal accuracy, these robot structures and controllers do not satisfy anymore. New control strategies are needed, and the weight and inertia of the moving parts have to be decreased. This decrease leads to lower structural resonance frequencies and larger link and joint deformations, and therefore urges the need for new control algorithms. They prevent or damp out undesired oscillations, resulting from excitation of structural dynamics. But even for the present-generation industrial robots there is need for such control algorithms. Tests on industrial robots have shown that traditional robot controllers excite the structural resonance frequencies, and that inaccuracies due to static and dynamic deformations can be significant [1, 2, 3].