Robotic Limb Posture Control for Astronaut EVA

In order to achieve the robotic limb’s ability to maintain the astronaut’s posture during EVA, the variable damping control method based on the Q-learning algorithm was proposed. Prior to the reinforcement learning training, it was necessary to model and simplify the astronaut system with the robotic limbs, which could function faster in the simulation environment, as shown in Figure 2. While the astronaut works outside the ISS cabin, one robotic limb holds the handrail to maintain the position in the working area. Under this condition, the handrail was considered as a fixed end and the end of the robotic limb was simplified to connect to that fixed end. The astronaut and the other robotic limb were combined and simplified into an end-load system, where the second robotic limb mainly provides auxiliary functions, such as tool delivery and operational support. As shown in Figure 2, they were reduced to a green solid ball at the end of the robotic limb. The blue ellipses represent the links of the robotic limb, and these links are connected by rotating joints, which are represented by the solid blue points. Each robotic limb had six degrees of spatial freedom. The fixed end was equal to the handrail of the ISS. The Cartesian coordinate system, which is the absolute coordinate system, was attached to the fixed end. Combined with the forward kinematics of the robotic limb, the end-load movement information for Cartesian space could be obtained in real time.
In addition, this model could also be split into two systems. One was the load system and the other was the robotic limb system without the load. Based on the model, the variable virtual restoring force was introduced to control the load for impact resistance and maintenance of position. In combination with the Q-learning algorithm, the variable damping controller was formed. The virtual restoring force was taken as an external force of the robotic limb. Finally, based on its dynamics, the virtual restoring force could be transformed into the control torque of each joint. In this way, the robotic limb could realize its position-maintaining control to help the astronaut.