Analyzing Unstated Goal Constraints On Reinforcement Learning Policies For Robotic Scrub Nurse Application

The main objective of this paper is to make an empirical analysis of the effect of various unstated spatial constraints on different reinforcement learning policies for the Robotic Scrub Nurse(RSN)applications,especially on the “reacher” task.This task is one of the most critical aspects of the robotic manipulation task.This paper provides our experimental results and the evaluation of the “reacher” task under different spatial constraints.These various unstated spatial constraints were tested using reinforcement learning algorithms: Deep Deterministic Policy Gradient(DDPG)and Soft-Actor Critic(SAC),which are combined with Hindsight Experience Replay(HER).The algorithm was evaluated on the 7-DOF robotic arm.The implementation of these state-of-the-art algorithms on deep reinforcement learning presented a robust performance,which is measured by reward values and success rate.The experiments were performed in a virtual environment Mu Jo Co where the robotic arm was trained to reach the random target points.The important aspect of the “reacher” task and the development of reinforcement learning applications in medical robotics is one of the main motivations behind this research objective.