Leveraging Physical Human-Robot Interaction for Surgical Robot Learning in Neuroendoscopy

Neuroendoscopy is a minimally invasive surgical approach with numerous benefits, but it presents technical challenges such as spatial constraints and the need for precise, repetitive movements. While surgical robots have the potential to address these challenges, their adoption in neuroendoscopy remains limited due to inadequate user interfaces and reliance on overly specific preprogrammed commands. This work introduces a robot learning algorithm that enables surgeons to kinesthetically train a robot arm to navigate and operate a neuroendoscope. Surgeons can manually guide the robot arm through task demonstrations, make real-time corrections, and teach the robot preferred surgical approaches, all within the surgical workspace constraints to ensure patient and surgeon safety. We aim to develop this machine learning-driven approach to create an intuitive shared control system between the surgeon and the robot, potentially reducing task completion times and enabling adaptation to anatomical variations.