Learned Task Space Control to Reduce the Effort in Controlling Redundant Surgical Robots

Redundant robots allow multiple robot joint configurations for the same end-effector pose by moving only in null space. Robot’s motions in null space are not intuitive to predict in general and in particular for medical personnel. In this work, we present a control concept that allows the operator to focus on the correct end-effector pose during time-critical tasks, e.g. change of the endoscope pose during a surgical intervention, while the shape of the redundant robotic structure is handled autonomously based on previously learnt preferred shapes close to the actual end-effector pose. We investigated the benefit of the proposed learned task space control over naive task space control that required an operator to manually control a virtual robot in task space and null space independently. In a first user study, we found that learned task space control significantly reduced the effort – as measured by task duration and task load – for operators compared to naive task space control.