Skill Transfer and Learning by Demonstration in a Realistic Scenario of Laparoscopic Surgery

Abstract

Commercially available systems for laparoscopic surgery usually come without operator side force feedback nor instrument-side force sensory.This often leads to increased trauma of tissue.Yet another reason why such systems are not widely accepted for every-day usage is a prolonged operation time, due to time consuming micro-manipulation tasks, e.g.knot tying.We present an approach which tackles both issues.On the one hand we provide a realistic surgical environment including haptic feedback, on the other hand we present basic techniques for partial autonomy.Autonomy will be implemented by combining two established methods of human-robot instruction: learning by demonstration and primitive instantiation.Once presented by the human instructor, subtasks are generalized and partitioned in reusable primitives, which can be recombined in order to solve a priori unknown tasks.