Interval-Arithmetic-Based Trajectory Scaling and Collision Detection for Robots with Uncertain Dynamics

Abstract

We consider two fundamental problems in control of robot manipulators: dynamic scaling of trajectories and collision detection using proprioceptive sensors. While most existing methods approach these problems by assuming accurate knowledge of the robot dynamics, we relax this assumption and account for uncertain model parameters and external disturbances. Our approach is based on the use of a recently proposed interval-arithmetic-based recursive Newton-Euler algorithm. This algorithm enables the efficient numerical computation of over-approximative sets of torques/forces arising from uncertain model parameters. The over-approximative nature of these sets is exploited in this work in order to provide a formally robust trajectory scaling and collision detection strategy. The effectiveness of the proposed approaches has been verified by means of experiments on a 6 degrees-of-freedom robot manipulator with uncertain dynamics.