Grasp Planning in Manufacturing with NAO Robot Using Reinforcement Learning

Abstract

Robotic manipulation plays a pivotal role in modern manufacturing, where efficient grasp planning is crucial for automation tasks. This paper presents a novel approach to grasp planning in manufacturing using the NAO robot and reinforcement learning (RL). Leveraging RL algorithms, the NAO robot learns optimal grasp strategies for manipulating various objects commonly encountered in manufacturing scenarios. The methodology involves simulating diverse manufacturing tasks, training the robot in a virtual environment, and evaluating its performance in both simulated and real-world settings. Results demonstrate the effectiveness of the proposed approach in achieving high grasp success rates and efficient manipulation. The proposed method was successfully tested on a real NAO robot within the FESTO MPS Station, further validating its applicability to real-world manufacturing environments. This research contributes to advancing robotic manipulation capabilities in manufacturing and opens avenues for further exploration in the integration of RL techniques with robotic systems for industrial automation.