轨道追逃制导与控制
近年来，为确保航天器长期在轨安全运行，轨道追逃博弈问题逐渐引起学者们的关注。追逃问题自20世纪由Isaacs在研究微分对策理论时提出以来，已在各领域被广泛研究，但航天器的特殊动力学特性为策略求解带来了新的挑战，而人工智能技术的迅猛发展，为轨道追逃策略求解提供了新的可行方案。在“航天 智能”思路的牵引下，深度神经网络和强化学习技术不断地被应用于轨道追逃博弈问题，对提高航天器的自主化和智能化水平具有重要意义。
鉴于此，《国防科技大学学报》组织策划“轨道追逃博弈”专题，专题发表于《国防科技大学学报》2024年第3期，包含1篇综述性论文、3篇技术性论文，由国防科技大学空天科学学院太空安全与航天器智能博弈技术团队撰写，从动力学与控制视角探讨轨道追逃博弈问题，内容涉及追逃策略优化、追逃态势分析、多航天器协同拦截等方面。
关键词：暂无
2024, 46(3):1-11.
DOI: 10.11887/j.cn.202403001
Abstract:
With the rapid development of spacecraft rendezvous and proximity operation technology, the problem of orbital pursuit-evasion has gradually become a research hotspot in the aerospace field. From the perspective of dynamics and control, the research status of spacecraft orbital pursuit-evasion was reviewed. General form of the orbital pursuit-evasion problem model based on quantitative differential games was given, and various types of orbital pursuit-evasion problems were systematically sorted out. For the solution of pursuit and escape strategies, the advantages and disadvantages of various methods were analyzed for closed-loop strategy and open-loop strategy. Focusing on the combination of artificial intelligence algorithm and orbital pursuit and escape problem, the research status of orbital pursuit and escape strategy based on deep neural network and reinforcement learning was expounded. Regarding future prospects, development directions has been proposed, including the pursuit-evasion game situation analysis, the multi-spacecraft game control, the game dynamics and control under three-body problem.
With the rapid development of spacecraft rendezvous and proximity operation technology, the problem of orbital pursuit-evasion has gradually become a research hotspot in the aerospace field. From the perspective of dynamics and control, the research status of spacecraft orbital pursuit-evasion was reviewed. General form of the orbital pursuit-evasion problem model based on quantitative differential games was given, and various types of orbital pursuit-evasion problems were systematically sorted out. For the solution of pursuit and escape strategies, the advantages and disadvantages of various methods were analyzed for closed-loop strategy and open-loop strategy. Focusing on the combination of artificial intelligence algorithm and orbital pursuit and escape problem, the research status of orbital pursuit and escape strategy based on deep neural network and reinforcement learning was expounded. Regarding future prospects, development directions has been proposed, including the pursuit-evasion game situation analysis, the multi-spacecraft game control, the game dynamics and control under three-body problem.
2024, 46(3):12-20.
DOI: 10.11887/j.cn.202403002
Abstract:
Aiming at guarding high-orbit high-value targets, a multi-spacecraft cooperative guard strategy based on reachable domain coverage was proposed. The cooperative guard mission was described from the perspective of relative motion, and the multi-pulse reachable domain of the threat was modeled as a convex optimization problem. In the framework of receding optimization, the guard planes and points were designed based on the dynamically updated terminal reachable domain of threat, and a multi-spacecraft cooperative trajectory planning model was constructed with the guard points as terminal position constraints, the corresponding guard trajectories were generated. Simulation results show that the proposed method can quickly calculate the terminal reachable domain of the threat. The cooperative guard strategy can effectively prevent the threat in multiple scenarios, and the guard success rate increases with the enhancement of the maneuvering ability of the guard spacecraft.
Aiming at guarding high-orbit high-value targets, a multi-spacecraft cooperative guard strategy based on reachable domain coverage was proposed. The cooperative guard mission was described from the perspective of relative motion, and the multi-pulse reachable domain of the threat was modeled as a convex optimization problem. In the framework of receding optimization, the guard planes and points were designed based on the dynamically updated terminal reachable domain of threat, and a multi-spacecraft cooperative trajectory planning model was constructed with the guard points as terminal position constraints, the corresponding guard trajectories were generated. Simulation results show that the proposed method can quickly calculate the terminal reachable domain of the threat. The cooperative guard strategy can effectively prevent the threat in multiple scenarios, and the guard success rate increases with the enhancement of the maneuvering ability of the guard spacecraft.
2024, 46(3):21-29.
DOI: 10.11887/j.cn.202403003
Abstract:
Given the influence of uncertainty such as measurement errors in the process of spacecraft free-time orbital pursuit-evasion game for rendezvous, a high-efficiency strategy based on receding horizon optimization was proposed as a solution method. The saddle point control strategy of the game was derived according to differential games, and the equivalent transformation of the problem was carried out. By solving open-loop saddle point strategy off-line in advance, the initial states of the problem and the corresponding solutions were taken as samples for neural network training, and the trained network structure can quickly obtain the approximate solution of the corresponding problem. In order to better deal with the measurement noise in the game environment, a receding horizon optimization framework was designed based on the neural network structure. By periodically solving the problem, the rendezvous of the pursuer and evader was finally realized. Numerical simulation shows that the proposed strategy can effectively deal with the uncertainty of measurement noise, and compared with the existing strategy in the literature, the calculation time can be reduced from minutes to several seconds.
Given the influence of uncertainty such as measurement errors in the process of spacecraft free-time orbital pursuit-evasion game for rendezvous, a high-efficiency strategy based on receding horizon optimization was proposed as a solution method. The saddle point control strategy of the game was derived according to differential games, and the equivalent transformation of the problem was carried out. By solving open-loop saddle point strategy off-line in advance, the initial states of the problem and the corresponding solutions were taken as samples for neural network training, and the trained network structure can quickly obtain the approximate solution of the corresponding problem. In order to better deal with the measurement noise in the game environment, a receding horizon optimization framework was designed based on the neural network structure. By periodically solving the problem, the rendezvous of the pursuer and evader was finally realized. Numerical simulation shows that the proposed strategy can effectively deal with the uncertainty of measurement noise, and compared with the existing strategy in the literature, the calculation time can be reduced from minutes to several seconds.
2024, 46(3):30-38.
DOI: 10.11887/j.cn.202403004
Abstract:
Spacecraft pursuit-evasion game is currently a research hotspot in the field of aerospace dynamics and control. Qualitative spacecraft pursuit-evasion game was studied in order to provide feasibility support for strategy design, and a situation analysis method for scenarios in close range was proposed by comprehensively using dimension-reduction dynamics and backwards reachability set. A dimension-reduction dynamic model was derived in the line-of-sight rotation coordinate system, and the pursuit-evasion qualitative model of the game system was determined to reduce state space dimension. The backwards reachable set of the target set was used to divide the pursuit-evasion state space and describe the capture zone. A time-dependent HJI (Hamilton-Jacobi-Isaacs) PDE (partial differential equation) was established to describe the evolution of backwards reachable set in the dimension-reduction dynamics, based on level set method, and a WENO-TVD(weighted essentially non-oscillatory-total variation diminishing) solver was designed to numerically calculate the viscous solution of the final value problem of the HJI PDE. These measures achieve the accurate description of pursuit-evasion target set and avoid the possible terminal singularity. The effectiveness of the method was demonstrated by numerical simulations of several pursuit-evasion scenarios with different thrust configurations, and the function of batch processing of initial situations in a single calculation was demonstrated.
Spacecraft pursuit-evasion game is currently a research hotspot in the field of aerospace dynamics and control. Qualitative spacecraft pursuit-evasion game was studied in order to provide feasibility support for strategy design, and a situation analysis method for scenarios in close range was proposed by comprehensively using dimension-reduction dynamics and backwards reachability set. A dimension-reduction dynamic model was derived in the line-of-sight rotation coordinate system, and the pursuit-evasion qualitative model of the game system was determined to reduce state space dimension. The backwards reachable set of the target set was used to divide the pursuit-evasion state space and describe the capture zone. A time-dependent HJI (Hamilton-Jacobi-Isaacs) PDE (partial differential equation) was established to describe the evolution of backwards reachable set in the dimension-reduction dynamics, based on level set method, and a WENO-TVD(weighted essentially non-oscillatory-total variation diminishing) solver was designed to numerically calculate the viscous solution of the final value problem of the HJI PDE. These measures achieve the accurate description of pursuit-evasion target set and avoid the possible terminal singularity. The effectiveness of the method was demonstrated by numerical simulations of several pursuit-evasion scenarios with different thrust configurations, and the function of batch processing of initial situations in a single calculation was demonstrated.
