Multi-Objective Trajectory Planning for Flexible Spacecraft via Physics-Informed Neural Network
Xilin Zhong, Shujie Liu, Ti Chen, Haiyan Hu
Abstract: This study addresses the autonomous collision-free motion planning of a spacecraft with flexible appendages using a neural network embedded with the rigid-flexible dynamics. The study presents the trajectory diffusion network and input transition method for local motion planning with rigid-flexible dynamics taken into account. Furthermore, the study deals with the coupling among three Euler angles via the inter-Euler-angle embedding method, and a reconstruction loss function designed to incorporate the coupling dynamics into the network. The proposed neural network integrates with the rapidly-exploring random tree star framework for the overall collision-free motion planning, and a precise algorithm designed for collision detection with the consideration of the deformation of the appendages. Both simulation and experimental studies validate the effectiveness of the motion planning method.
文章链接：https://www.sciencedirect.com/science/article/pii/S1270963825007813
