Manipulator Control using Federated Deep Reinforcement Learning

摘要

Manipulators or Robotic arms have revolutionized the manufacturing sector and have also been employed for scientific research in environments humans can’t access. Reinforcement learning has drawn a lot of interest as a means of improving self-improvement and generalisation skills. Reinforcement learning has also been actively used for manipulator control. Concerns about data sharing privacy are also becoming more widespread, especially in robotic environments. In this study, a deep reinforcement learning strategy is applied to the manipulator to move from a random initial point to a provided target point. Specifically, the Deep Deterministic Algorithm has been used for this purpose. This strategy is trained using a federated learning approach to overcome privacy issues and to set a foundation for further analysis of distributed learning. The results obtained indicate that both the traditional deep reinforcement learning and federated deep reinforcement learning frameworks are providing similar success rate of about 86%. However, federated deep reinforcement learning offers a promising alternative as it addresses privacy concerns and provides other benefits outlined in this paper. The models were also simulated using the Gazebo 3-D simulator.