Learning a Deep RL Policy for Automated Needle Manipulation on Surgical Robots

Jun 1, 2024ยท
Jiaying Fang
Jiaying Fang
,
Baiyu Shi
ยท 0 min read
Needle Manipulation in Simulation Environment
Abstract
We developed a deep reinforcement learning policy for needle reaching, tracking and picking in surgical RL environment. A two-stage vision-based needle manipulation RL policy was implemented, which converges within 50k steps, while other end-to-end policies struggle to converge even in 80k steps.