Optimizing Robotic Automatic Suturing Through VR-Enhanced Data Generation for Reinforcement Learning Algorithms

摘要

This paper explores the integration of Virtual Reality(VR) to a Surgical Robotic Simulation to enhance the quality of data used for training a ground-truth algorithm for surgical procedures performed by the DaVinci robot inside a simulated environment. As it is to be demonstrated by this paper, VR and Reinforcement Learning (RL) techniques can significantly improve the realism and effectiveness of the training data compared to traditional methods. It also investigates and deepens the study of incorporating Cognitive Vision theories to guide the learning process, following the premise that the full-immersion visual and haptic feedback models will result in better quality training data for the surgical robot to perform an autonomous surgery algorithm, leading to more accurate and adaptable minimally-invasive robotic surgery (MIRS) systems. This research was inspired and shaped by our active participation in the 2023-2024 AccelNet Surgical Robotics Challenge.