Knowledge-based direction prediction to optimize the null-space parameter of a redundant robot in a telemanipulation scenario

This paper shows an approach of a knowledge-based path-guidance for minimally invasive robotic surgery. For an accurate path-guidance it is important to know or estimate the direction of the tool tip's motion. Therefor it is necessary to predict the future direction. In this work, the predicted direction is based on clustering applied to typical trajectory sets, combined with building first and second order Markov models which represent cluster transitions. A coarse prediction is obtained by cluster transitions. For improvement, this is refined by projecting the cluster points distribution on a (unit) sphere surrounding the current tool tip position. For the latter step, a discrete procedure is proposed that takes into account the most likely consecutive cluster(s). The resulting predicted direction can be used to guide the robot's movement, especially by controlling the joint angles of a redundant robot in order to avoid joint limits. The main focus of this work is on the actual prediction of direction which is evaluated using a synthetic test scenario.