Research On Autonomous Control For Spacecraft In Non-cooperative Rendezvous And Proximity Mission

The rendezvous and proximity technology of non-cooperative target has been widely used in space monitoring and services such as on-orbit maintenance,on-orbit assembly,refueling,and debris removal.However,the manual control method of human or ground station in the loop has the disadvantages of high cost and low efficiency.The autonomous control method based on the relative guidance method generally lacks the flexibility of manual control,which severely restricts its application in efficient and increasingly complex space tasks.Therefore,this paper takes the autonomous control for spacecraft as the background to carry out the research on the rendezvous and proximity technology of non-cooperative.Aiming at the problem that non-cooperative target cannot actively provide pose information for autonomous control method,a pose estimation method based on deep learning is proposed.Firstly,YOLOv5 s is used to detect the non-cooperative target.Then,a lightweight HRNet with concurrent spatial and channel squeeze and excitation block is used to complete pose estimation.Experimental results on public dataset show that,compared with other advanced methods,the proposed pose estimation method greatly reduces the model complexity and computation,and achieves better pose estimation results.Aiming at the lack of flexibility of autonomous control method based on relative guidance,an autonomous controller based on astronaut control model is designed.Firstly,the control characteristics of manual rendezvous and proximity are analyzed.Then adaptive neuro-fuzzy inference system（ANFIS）is used to simulate the astronaut’s perception,decision-making and execution process,and the astronaut control model is established as an autonomous controller.The simulation results show that the prediction accuracy of ANFIS trained by subtractive clustering is high,and it can effectively simulate the astronaut control process.Aiming at the complex process and uncertain factors of manually selected subtractive clustering parameters,a Hybrid Aquila Optimizer（HAO）based on Henon map and crisscross optimization algorithm is proposed.HAO is used to automatically search the optimal subtractive clustering parameters.The simulation results show that,compared with other advanced algorithms,HAO has higher convergence accuracy,faster convergence speed and stronger stability,which can effectively search the optimal parameters of subtractive clustering and improve the prediction accuracy of astronaut control model.