Automatic keypoint generation in learning-by-demonstration trajectories for a surgical robot

Resumen

Surgical robotics is one of the most promising areas of robotics, however most robotic systems applied to surgery operate under a direct teleoperation scheme. In order to improve the performance of these systems, it would be desirable to increase the level of autonomy of surgical robots. The main objective of this paper is to develop a methodology for the automatic generation of trajectories learned by demonstrating surgical procedures. The detail of an algorithm for obtaining keypoints in the trajectories obtained from surgeons is presented. These keypoints contain not only geometric information, but also kinematics and the force field exerted by the surgeon. The coding presented allows the implementation of a demonstration learning scheme based on hidden Markov models particularized for surgical operations. The major contribution of our proposal approach is the use of keypoints obtained from all the demonstrations and the integration of speed and force information together with geometric information. For this, the experimental results obtained in several tests carried out in a setup specially designed to obtain the forces and trajectories carried out by neurosurgeons in a mastoid bone milling procedure are presented.