Research on autonomous decision‐making method for spacecraft in the mission of rendezvous and approaching to maneuvering target based on deep reinforcement learning

Abstract In this paper, an autonomous decision‐making method for close‐range rendezvous and approaching to noncooperative target with strong maneuvering is proposed based on the distributed distributional deep determined policy gradient (D4PG), and ground simulation system of rendezvous and approaching is constructed by the improved YOLOv5s and universal robotic arm to verify the effectiveness and feasibility. The contributions of this paper are as follows: (1) an improved nearest neighbor exploration mechanism including random constant value and logarithmic constant value is proposed to stabilize the convergence for the decision‐making under the uncertainty; (2) a discrete wavelet transform (DWT)‐based preprocessing, a fusion of generalized additive model (GAM), and an optimization scheme are proposed for the dark background and false detection of black objects; and (3) a comparison analysis of simulation result between the traditional noise and proposed state‐based exploration is presented for the superiority and some robot‐based ground experiments are done to verify feasibility of close‐range space rendezvous and approaching.