Reinforcement learning control of collision avoidance for ultra-close formation of spacecraft with input constraints
MENG Yi-zhen
HUANG Jing
ZHOU Shao-hui
ZHOU Bin
ZHU Kang-wu
Abstract：Considering the control problem of reconstructing the ultra-tight formation of near-Earth orbit spacecraft in the presence of external disturbances,collision avoidance constraints,and fixed-time constraints,this study presents a robust control method for spacecraft formation that accounts for the dead-zone effect of the actuator under multiple constraint conditions.Firstly,we establish the nonlinear dynamic equations governing the relative positions of the spacecraft in the complete near-Earth orbit formation,as well as the dynamic response model capturing the dead-zone effect of the actuator.Secondly,we design a constraint mechanism for the relative positions of the formation based on state constraints.Robust control laws,employing a combination of backstepping and a reinforcement learning actor-critic network,are proposed to address collision avoidance constraints and fixed-time constraints.Additionally,we approximate the dead-zone characteristics of the actuator's thrusters by leveraging a reinforcement learning actor network.To mitigate the impact of the dead-zone effect on control accuracy,we minimize the cost function of the actor network.The Lyapunov stability theorem is employed to demonstrate the uniformly boundedness of the closed-loop system.Finally,we conduct simulation verification on the MATLAB/Simulink platform,and the results substantiate the effectiveness of the proposed method.
Keywords：spacecraft formationcollision avoidancereinforcement learning controldead-zone effectfixed time constraint
Publication Date：2025-04-30
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：10( 659-668 )