Search

Active non-cooperative target tracking is an important ability for free-floating space manipulator (FFSM) systems. Aiming at the difficulty of pose and posture estimation of non-cooperative targets and the complexity of FFSM planning and control, the paper proposed an end-to-end target tracking algorithm using deep reinforcement learning (DRL) algorithm. The algorithm uses raw camera images as input and directly outputs the manipulator joint velocities. To deal with the partially observable Markov decision processes (POMDP) problem caused by single image input, we combine the soft actor–critic algorithm with recurrent neural networks. Therefore, the velocity information of the target is introduced to transform the task into a fully observable Markov decision-making process (MDP). Our approach is completely data-driven. It avoids the position and posture estimation of non-cooperative target and the complex modeling motion planning process of FFSM. We trained and tested the algorithm by building a simulation environment in CoppeliaSim. The results showed that our approach achieved good results in active non-cooperative target tracking of FFSM systems and demonstrated the great potential of DRL algorithms in solving space tasks. [ABSTRACT FROM AUTHOR]

Target capture is the basis of many space missions. With free-floating space manipulators(FFSM) are more and more widely used in various space tasks, using FFSM to catch space targets has become a research hotspot. In this research, we propose an end-to-end capture strategy under unknown dynamics based on deep reinforcement learning, which directly controls the manipulator with the raw camera image as the input. The algorithm take imitation learning for pre-training. The policy after pre-training process will explore high-reward regions in the state space with higher probability. This will improve the efficiency of DRL exploring in the high-dimensional state space composed of images. We built a simulation environment in CoppeliaSim to train and test our algorithm. The results show that our algorithm has achieved good results in end-to-end target capture task without dynamic modeling and is more efficient than pure model free reinforcement learning. [ABSTRACT FROM AUTHOR]

Free-floating space manipulators(FFSM) are more and more widely used in various space tasks, and active object tracking(AOT) is the basis of many missions in space. AOT of FFSM systems has two main difficulties: modeling and control of FFSM systems and tracking motion planning of space manipulators. To deal with these problems, the paper proposed an active object tracking proposal of FFSM systems using deep reinforcement learning(DRL) algorithm, Proximal Policy Optimization(PPO). Our approach is completely data-driven, which avoids the complex modeling process of FFSM and does not require motion planning for space manipulators, which is more concise than traditional algorithms. We trained and tested the algorithm by building a simulation environment in CoppeliaSim, and compared with the resolved motion rate control(RMRC). The results showed that our approach achieved good results in active object tracking of FFSM systems and demonstrated the great potential of DRL algorithms in solving space tasks. [ABSTRACT FROM AUTHOR]