A Lunar Robot Obstacle Avoidance Planning Method Using Deep Reinforcement Learning for Data Fusion

Research on Dual-Arm Control of Lunar Assisted Robot Based on Hierarchical Reinforcement Learning under Unstructured Environment

“…The lunar surface is far away, the environment is extreme, and human resources are scarce. Therefore, artificial-intelligence-based robotics will play an important role in manned lunar exploration activities [1,2]. Robot systems, which assist astronauts in fine operations and other tasks, play an important role in manned lunar landing missions.…”

“…The commonly used manipulator control requires sensing and modeling of the astronaut's environment and manipulator [5]. This method has two major challenges: (1) it is difficult to model the unstructured environment on the lunar surface, and the solution efficiency is low; (2) when the environmental parameters change, the original model cannot easily solve the control problems in the new environment. In recent years, data-driven manipulator control methods have been proposed [6].…”

“…Research on the dual-arm control of redundant-degree-of-freedom robots is a very challenging problem. In an unstructured environment, the RL method of reinforcement learning is more challenging than the traditional reinforcement learning decision-making problem [1,15]. First of all, the trajectory of the dual arms of a redundant-freedom robot is of an exponential order of magnitude in joint action space.…”