High-orbit high-maneuverability satellite deduction system and method based on distributed architecture

Abstract

The invention discloses a high-orbit high-maneuverability satellite deduction system and method based on a distributed architecture, and belongs to the technical field of a satellite pursuit game strategy, and the method comprises the steps: setting a test orbit and a virtual reference satellite according to the positions of a tracking satellite and an escape satellite; according to the motion state of the virtual reference satellite, motion states of spacecrafts of the tracking satellite and the escape satellite are determined respectively; determining motion equations of the tracking satellite and the escape satellite in the circular reference orbit according to the CW equation and in combination with the motion states of the tracking satellite and the escape satellite; a pursuit game model is established according to motion equations of a tracking satellite and an escape satellite, reward functions of the tracking satellite and the escape satellite are constructed based on the principle that game purposes of a pursuit party and an escape party of the pursuit game are opposite, and a reinforcement learning optimization algorithm is adopted to carry out iterative training on pursuit strategies of the tracking satellite and the escape satellite to obtain a pursuit game model. The pursuit strategies of the tracking satellites and the escape satellites are optimized with the purpose of maximizing the respective rewards, and the trained tracking satellites and the escape satellites output the respective pursuit strategies; the method has the capability of quickly, intelligently and efficiently generating the orbital maneuvering strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite pursuit and escape game strategies, and specifically to a high-orbit and high-maneuverability satellite deduction system and method based on a distributed architecture. Background Art

[0002] Satellites can provide real-time intelligence, conduct reconnaissance and surveillance, provide precise navigation and guidance, and have high military value. Satellites can be used to complete military tasks such as enemy reconnaissance, friendly communications, and long-range precision strikes. Satellites can affect battlefield situations in the air, at sea, and on the ground. Based on the security of future space dominance, higher requirements are placed on the capabilities of on-orbit spacecraft. While completing their own tasks, they must have the ability to quickly avoid danger and the ability to actively interfere with target aircraft.

[0003] Space attack and defense technology is a very important research field at present, which can ensure the safety of spacecraft when encountering danger. And spacecraft pursuit is the key research object in space attack and defense technology. The pursuit game essentially makes both parties seek the optimal control of the spacecraft. Solving this problem is actually calculating the solution of a high-dimensional nonlinear equation group. In the past, differential game methods were often used to solve it, but due to the high dimension, the solution is very complicated. The computing power of the onboard computer is poor, and the control instructions need to be injected into the spacecraft after the ground calculation is completed. The situation of the spacecraft pursuit game changes quickly, and the control instructions need to be solved in real time. This means that it is difficult to complete the actual pursuit task by solving the high-dimensional nonlinear equation group through traditional methods. It is necessary to further study the artificial intelligence method that can quickly and real-time generate spacecraft control instructions. Summary of the invention

[0004] In view of the problems that the differential game method is very complicated and computationally intensive to solve the spacecraft orbital pursuit game strategy, and it is difficult to meet the real-time requirements of the on-orbit pursuit game, and the high coordination requirements of the ground control due to the complexity and autonomy of the on-orbit pursuit game of the spacecraft, the present invention provides a high-orbit and high-maneuverability satellite simulation system and method based on a distributed architecture, which uses a neural network to guide the spacecraft to conduct game pursuit, so that it has the ability to quickly, intelligently and efficiently generate orbital maneuvering strategies.

[0005] The present invention is achieved through the following technical solutions:

[0006] A high-orbit high-mobility satellite deduction method based on a distributed architecture comprises the following steps:

[0007] Step 1: According to the positions of the tracking satellite and the escape satellite, set the test orbit and the virtual reference satellite;

[0008] Step 2, according to the motion state of the virtual reference satellite, the motion states of the spacecraft of the tracking satellite and the escaping satellite are determined respectively;

[0009] Step 3: According to the CW equation and in combination with the motion states of the tracking satellite and the escaping satellite, the motion equations of the tracking satellite and the escaping satellite in the circular reference orbit are determined;

[0010] Step 4. Establish a pursuit-escape game model based on the motion equations of the tracking satellite and the escaping satellite. Based on the principle that the game objectives of the pursuer and the escaping party in the pursuit-escape game are opposite, construct the reward function of the tracking satellite and the escaping satellite, and use the reinforcement learning optimization algorithm to iteratively train the pursuit-escape strategies of the tracking satellite and the escaping satellite. With the goal of maximizing their respective rewards, optimize the pursuit-escape strategies of the tracking satellite and the escaping satellite. After training, the tracking satellite and the escaping satellite output their respective pursuit-escape strategies.

[0011] Preferably, the reference orbit is a circular reference orbit.

[0012] Preferably, the method for determining the motion state of the spacecraft of the tracking satellite and the escaping satellite is as follows:

[0013] constructing a reference spacecraft local orbital coordinate system in which the motion states of the spacecraft for the tracking satellite and the escape satellite are determined;

[0014] The local orbital coordinate system of the reference spacecraft takes the center of mass of the virtual spacecraft as the origin of the coordinate system, the x-axis coincides with the geocentric radius of the reference spacecraft, and the positive direction is from the center of the earth to the spacecraft; the y-axis is perpendicular to the x-axis in the orbital plane, and the positive direction points to the direction of motion; the z-axis is determined by the right-hand rule, that is, along the direction of the orbital angular momentum.

[0015] Preferably, the motion state of the tracking satellite is

[0016] The motion state of the escaped satellite

[0017] Preferably, the motion equations of the tracking satellite and the escaping satellite are expressed as follows:

[0018]

[0019]

[0020] In the formula is the average angular velocity of the reference orbit, μ is the Earth's gravitational constant, R r is the reference orbit radius, a x 、a y 、a z They are the thrust acceleration components of the spacecraft in the x, y, and z directions respectively.

[0021] Preferably, in step 4, during the iterative training of the pursuit and escape strategies of the tracking satellite and the escaping satellite using the PPO algorithm in reinforcement learning, the observation information of the tracking satellite and the escaping satellite is the motion state information of the confrontation target;

[0022]

[0023] The actions of the tracking and escaping satellites are set to their velocity increments:

[0024]

[0025] The control quantity of the tracking satellite and the escape satellite is the thrust acceleration vector, which is expressed as follows:

[0026] u A =(Δv x ,Δv y ,Δv z ) T

[0027] u B =(Δv x ,Δv y ,Δv z ) T .

[0028] Among them, T is the terminal moment of the pursuit and escape game, and t is the current moment.

[0029] Preferably, the reward function is expressed as follows:

[0030] r B =λ1r tf -0.95λ2||Δv B ||1+0.95λ3||Δv A ||1-λ t (R min -R pe ).

[0031] Preferably, the reinforcement learning optimization algorithm is a proximal strategy optimization algorithm.

[0032] A system for a high-orbit and high-mobility satellite deduction method based on a distributed architecture, comprising:

[0033] Initialization module, used to set the test orbit and virtual reference satellite according to the positions of the tracking satellite and the escape satellite;

[0034] A motion state module, for determining the motion states of the spacecraft of the tracking satellite and the escaping satellite respectively according to the motion state of the virtual reference satellite;

[0035] The motion equation module is used to determine the motion equations of the tracking satellite and the escape satellite in the circular reference orbit according to the CW equation and the motion states of the tracking satellite and the escape satellite;

[0036] The pursuit and escape strategy module is used to establish a pursuit and escape game model based on the motion equations of the tracking satellite and the escaping satellite. Based on the principle that the game objectives of the pursuer and the escaping party in the pursuit and escape game are opposite, the reward function of the tracking satellite and the escaping satellite is constructed, and the PPO algorithm in reinforcement learning is used to iteratively train the pursuit and escape strategies of the tracking satellite and the escaping satellite. The pursuit and escape strategies of the tracking satellite and the escaping satellite are optimized with the goal of maximizing their respective rewards. The trained tracking satellite and the escaping satellite output their respective pursuit and escape strategies.

[0037] A distributed satellite simulation system comprises a simulation platform and a simulation control system;

[0038] The simulation platform is used to simulate multiple high-orbit satellite entity models;

[0039] The simulation control system is used to execute the steps of the high-orbit and high-maneuverability satellite simulation method based on a distributed architecture, and control the pursuit process of the high-orbit satellite entity model according to the output pursuit strategy.

[0040] Compared with the prior art, the present invention has the following beneficial technical effects:

[0041] The high-orbit high-mobility satellite deduction method based on a distributed architecture provided by the present application can ensure that the pursuit strategy can adapt to the ever-changing confrontation environment by dynamically setting the reference orbit and the virtual reference satellite according to the real-time position of the tracking satellite and the escaping satellite. The introduction of the virtual reference satellite as a benchmark for the motion state helps to simplify the problem, making the description of the relative motion state of the two pursuit satellites more intuitive and easy to handle. Determining the motion state of the tracking satellite and the escaping satellite based on the motion state of the virtual reference satellite can ensure the accuracy of the state description and provide a reliable basis for subsequent model establishment and strategy optimization. The use of the CW equation (Clohessy-Wiltshire equation) to describe the relative motion of the two pursuit satellites on the circular reference orbit is a scientific method based on the principles of orbital mechanics, which can accurately reflect the relative motion relationship between the two satellites. Constructing the reward function based on the principle that the game objectives of the pursuit party and the escape party in the pursuit game are opposite helps to deeply understand the essence of the pursuit confrontation from the perspective of game theory and provide new ideas for strategy optimization. The use of reinforcement learning optimization algorithm to iteratively train the pursuit strategy can make full use of the self-learning and optimization capabilities of reinforcement learning, continuously adapt to changes in the confrontation environment, and improve the intelligence and adaptability of the pursuit strategy. Optimizing the pursuit and escape strategy with the goal of maximizing their respective rewards helps ensure the effectiveness and practicality of the strategy, allowing tracking satellites and escaping satellites to perform better in high-orbit confrontations.

[0042] The distributed satellite simulation system of the present application dynamically constructs a "virtual satellite game environment" that runs in parallel with the real system, analyzes the changes in the satellite game situation in the simulation environment, and dynamically generates various possible situation diagrams of enemy and friendly satellites in the future, providing an effective solution for the dynamic analysis of the real-time situation of satellite games. The technical solution of the present invention can significantly improve the simulation efficiency in the distributed satellite simulation process. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of the high-orbit and high-mobility satellite deduction method based on a distributed architecture of the present invention;

[0044] Figure 2 It is a cross-sectional view of the high-orbit and high-maneuverability satellite pursuit and escape simulation system of the present invention.

[0045] Figure 3 is a flow chart of the simulation method of the present invention;

[0046] Figure 4 It is a display interface of the simulation scene of the present invention;

[0047] Figure 5 It is a parameter configuration diagram of the simulation system of the present invention;

[0048] Figure 6 It is the deduction and image display interface of the present invention. DETAILED DESCRIPTION

[0049] The present invention will be further described in detail below in conjunction with the accompanying drawings, which are intended to explain the present invention rather than to limit it.

[0050] Example 1

[0051] See also Figure 1 A high-orbit high-mobility satellite deduction method based on a distributed architecture comprises the following steps:

[0052] Step 1: According to the positions of the tracking satellite and the escape satellite, a circular reference orbit and a virtual reference satellite are set;

[0053] Because the high-orbit satellites used for high-orbit confrontation are not too far apart, a circular reference orbit and a virtual reference satellite are selected near the tracking satellite and the escaping satellite, and the motion states of the two spacecraft are described relative to the reference satellite.

[0054] Step 2: According to the motion state of the virtual reference satellite, the motion states of the spacecraft of the tracking satellite and the escaping satellite are determined respectively.

[0055] The center of mass of the virtual reference satellite's spacecraft is the origin of the coordinate system. The x-axis coincides with the reference spacecraft's geocentric radius vector, and the positive direction is from the center of the earth to the spacecraft. The y-axis is perpendicular to the x-axis in the orbital plane, and the positive direction points to the direction of motion. The z-axis is determined by the right-hand rule, that is, in the direction of the orbital angular momentum.

[0056] According to the Earth orbit coordinate system of the spacecraft of the virtual reference satellite, the motion states of the tracking satellite and the escape satellite are as follows:

[0057] Tracking the motion of satellites,

[0058] The motion state of the escaped satellite

[0059] Step 3: According to the CW equation and in combination with the motion states of the tracking satellite and the escape satellite, determine the motion equations of the tracking satellite and the escape satellite in the circular reference orbit.

[0060] The reference spacecraft is selected near the two spacecrafts (the distance between the two spacecraft and the reference spacecraft is small relative to the reference orbit radius), and the reference orbit is assumed to be a circular orbit. Therefore, under the assumption of the two-body dynamics model, the motion equations of the tracking satellite and the escaping satellite are expressed as follows:

[0061]

[0062]

[0063] In the formula is the average angular velocity of the reference orbit (the reference orbit is a circular orbit), μ is the Earth's gravitational constant, R r is the reference orbit radius, a x 、a y 、a z They are the thrust acceleration components of the spacecraft in the x, y, and z directions respectively. This equation is a linear differential equation with constant coefficients, and it has an analytical solution when the right side is zero.

[0064] Step 4. Establish a pursuit-escape game model based on the motion equations of the tracking satellite and the escaping satellite. Based on the principle that the game objectives of the pursuer and the escaping party in the pursuit-escape game are opposite, construct the reward function of the tracking satellite and the escaping satellite, and use the PPO algorithm in reinforcement learning to iteratively train the pursuit-escape strategies of the tracking satellite and the escaping satellite. With the goal of maximizing their respective rewards, optimize the pursuit-escape strategies of the tracking satellite and the escaping satellite. After training, the tracking satellite and the escaping satellite output their respective pursuit-escape strategies.

[0065] The high-orbit confrontation time limit of the tracking satellite and the escaping satellite is set to T (T is a finite value), and the control quantities of the two satellites are thrust acceleration vectors, which are respectively denoted as u A =(Δvx ,Δv y ,Δv z ) T and u B =(Δv x ,Δv y ,Δv z ) T .

[0066] The PPO in reinforcement learning is used to solve the spacecraft pursuit game problem. The environmental information is the relative state information of the two satellites and the remaining time of the pursuit game, s = (X A -X B ,Tt) where t is the current time and T is the terminal time of the pursuit-escape game.

[0067] The observation information of the tracking satellite and the escaping satellite is the motion state information relative to the confrontation target.

[0068]

[0069] The actions of the pursuit and escape satellites are set to their velocity increments,

[0070]

[0071] The reward functions for tracking satellites and escaping satellites are set as follows:

[0072] r B =λ1r tf -0.95λ2||Δv B ||1+0.95λ3||Δv A ||1-λ t (R min -R pe )

[0073] Among them, R min is the relative distance between the two satellites when they are closest to each other, R pe It is the minimum relative distance between two satellites calculated without control, representing the reward after the game ends. If satellite A is captured, the reward for satellite B will be increased, otherwise the opposite is true.

[0074]

[0075] Where Δv B represents the velocity increment of satellite B’s maneuver, Δv A represents the speed increment of satellite A’s maneuver. If the speed increment is larger, the fuel consumption will be greater. Therefore, the reward function of satellite B hopes that its own maneuver is smaller and the enemy’s maneuver is larger. f =T.

[0076] Since the pursuit and escape goals of the pursuit and escape game are opposite, the reward function of satellite A is set to: R A =-R B .

[0077] The PPO algorithm is used to train satellite A and satellite B. Different reward functions are set for the two satellites. The same neural network training is used, and both use their own decision-making neural networks to make decisions.

[0078] Example 2

[0079] Based on the above-mentioned high-orbit and high-mobility satellite deduction method based on a distributed architecture, the present application also provides a high-orbit and high-mobility satellite deduction system based on a distributed architecture, which may include:

[0080] An initialization module is used to set the circular reference orbit and the virtual reference satellite according to the positions of the tracking satellite and the escape satellite;

[0081] The motion state module is used to determine the motion states of the spacecraft of the tracking satellite and the escaping satellite respectively according to the motion state of the virtual reference satellite.

[0082] The motion equation module is used to determine the motion equations of the tracking satellite and the escape satellite in the circular reference orbit based on the CW equation and combined with the motion states of the tracking satellite and the escape satellite.

[0083] The pursuit and escape strategy module is used to establish a pursuit and escape game model based on the motion equations of the tracking satellite and the escaping satellite. Based on the principle that the game objectives of the pursuer and the escaping party in the pursuit and escape game are opposite, the reward function of the tracking satellite and the escaping satellite is constructed, and the PPO algorithm in reinforcement learning is used to iteratively train the pursuit and escape strategies of the tracking satellite and the escaping satellite. With the goal of maximizing their respective rewards, the pursuit and escape strategies of the tracking satellite and the escaping satellite are optimized, and the trained tracking satellite and the escaping satellite are output.

[0084] Example 3

[0085] The above-mentioned high-orbit and high-maneuverability satellite simulation method based on distributed architecture can be applied in a distributed satellite simulation system to simulate the pursuit and escape status of hostile satellites in the satellite simulation system.

[0086] A distributed satellite simulation system comprises a simulation platform, a simulation management process unit and a simulation unit.

[0087] A simulation platform for simulating a plurality of entity models; and a plurality of computers for running the simulation platform in parallel, wherein the simulation platform installed on each of the plurality of computers is provided with at least one entity model among the plurality of entity models, and is used for simulating the plurality of entity models in parallel, wherein: each of the computers is used for monitoring the simulation status of the entity models distributed on the computers themselves, obtaining the resource usage of the computers themselves, and migrating the entity models between the computers according to the simulation status of the entity models and / or the resource usage of the computers themselves;

[0088] The simulation management process unit is used to determine a first target computer with the highest resource utilization rate among the computers according to the resource utilization rate of each computer sent by the simulation process of each computer; when the resource utilization rate of the first target computer is higher than the first utilization rate, the simulation management process migrates the entity model on the first target computer to a second target computer among the computers except the first target computer;

[0089] The simulation unit is used to execute the steps of the high-orbit and high-maneuverability satellite simulation method based on a distributed architecture. The pursuing spacecraft are all equipped with intelligent decision-making capabilities, realizing the autonomous pursuit and escape game of the two spacecraft in the continuous maneuvering space. The proximal policy optimization algorithm (PPO) is used to train the two spacecraft simultaneously to alternately improve their game capabilities. In order to make the neural network adapt to different game scenarios, random initial points are set during the training process for the pursuing spacecraft to explore. Finally, the trained neural network is used to guide the spacecraft to play the game of pursuit, so that it has the ability to quickly, intelligently and efficiently generate orbital maneuvering strategies.

[0090] Example 5

[0091] See also Figure 2-6 ,A simulation method for a distributed satellite simulation system is as follows:

[0092] Step 1: The system establishes the participating satellites of the red and blue parties according to the mission assumptions, and simulates the working models of the satellite models, command and control systems, battlefield environment and other systems of both parties. In terms of configuration, the red and blue bilateral simulation forces can be distributed on one or more computers according to the model simulation granularity, and the modules such as combat assumptions, simulation control, recording and playback, and efficiency evaluation run on independent computers. To ensure the flexibility and scalability of the system, each platform realizes data interaction through the data layer residing in the terminal.

[0093] Step 2. The high-orbit satellite game visualization simulation and deduction system drives the model according to the simulation data to achieve a comprehensive perception of the game scenario. Carry out data-driven game simulation design, and coordinate the simulation data with the comprehensive situation display model data through the intermediate components to achieve good versatility and scalability. The data-driven intermediate component consists of a data parsing interface and a data configuration file. The data configuration file is responsible for configuring the driving relationship between the simulation data and the comprehensive situation display model data; the data parsing interface parses the data configuration file to realize the simulation data driving the correct situation model. According to the simulation driving requirements of different satellites, the simulation drive is realized by configuring the corresponding relationship between the simulation data and the satellite model data. The system data structure is shown in the following table:

[0094] Table 1. System interface data structure

[0095]

[0096] Step 3. Combined with the relevant needs and typical characteristics of satellite gaming, satellite gaming simulation is conducted in a turn-based card system. For a new scenario simulation, the basic process is as follows: before the simulation begins, a designated user (usually the person who plays the leading role) first creates a new scenario. The new scenario requires setting the scenario name and describing the scenario. Then the user needs to set the scenario for each simulation party based on the scenario information.

[0097] After the new scenario is established, the users participating in the game enter the scenario and select the role of the game player or the director. After each user determines the role, they need to preset the concept and game strategy according to the scenario description. The preset concept requires the game player to explain their strategic goals in this game and preset the goals of each round according to their own strategic goals. After completing the concept preset, the game player needs to preset the strategy for each round according to the goals and strategic goals of the round. The preset strategy can define the strategy and its component means by itself, or select the strategy means from the strategy means database for modification. The preset concept and preset strategy are unified into the game preparation stage.

[0098] After all the parties have completed their preparations, the leader will start to make assumptions and enter the simulation phase. Before the simulation begins, the leader needs to announce the simulation rules to all the parties, and each party will conduct simulations according to the simulation rules. During the simulation phase, each party will conduct simulations according to the simulation rules and implement several strategies in each round. The leader will evaluate the visibility, feasibility and effects of the strategies in various fields in real time, and control the progress of the rounds according to the simulation situation.

[0099] The review and evaluation stage is an important stage for evaluating the implementation of the game. After the game is completed, the game data elements can reflect the game-related data including the gain / loss effect of the means, the strategic objectives and round goal achievement between the game parties, the strategy statistics dashboard, the game line, etc., to achieve a refined evaluation of the entire game process.

[0100] During the review and evaluation phase, the satellite game simulation system can provide the game lines of the entire simulation process, allowing the evaluators to conduct a detailed analysis of the progress of the round and the game strategy and their relationship. For key game strategies, users can choose to re-simulate at this strategy point, that is, retain all the simulation information before this strategy, start from this strategy, and re-simulate subsequent simulations based on this assumption. The same assumption generates multiple game lines after multiple simulation iterations, forming the game chain of this assumption.

[0101] Step 4: Construct a high-orbit and high-mobility satellite simulation method based on a distributed architecture.

[0102] The study of the complex two-way confrontation problem of high-orbit target pursuit and escape game in space provides a plan for autonomous planning of spacecraft maneuvering trajectory, and explores the use of artificial intelligence to achieve intelligent, autonomous, and rapid online planning of maneuvering trajectory in pursuit and escape game. Based on the generated autonomous planning plan of satellite maneuvering trajectory, a complete dynamic process simulation analysis is performed.

[0103] For details of the high-orbit and high-mobility satellite simulation method based on a distributed architecture, please refer to Example 1.

[0104] Step 5: Simulate the system operation mode.

[0105] The operation process of the simulation system is as follows: Figure 2 As shown in the figure, the visualization deduction program is responsible for operation control in the system, and the satellite game simulation program is responsible for the operation of the game pursuit algorithm. Its operation and stop are uniformly scheduled by the visualization deduction program. During the operation of the satellite ultra-real-time simulation deduction operation system, after the game simulation program is started and initialized, a start command is sent to the simulation deduction visualization program, and then the cycle cycle mode of game simulation model solution is entered. At the same time, during the solution process, the solution data of the satellite entity model of both the enemy and our side is released to the visualization deduction program. After the model solution is completed, a stop command is sent to the visualization deduction program to end the satellite game simulation program; after the simulation deduction visualization program is started and initialized, it waits to receive the start command sent by the satellite game simulation program. After receiving the command, it cyclically receives the data released by the satellite game simulation program to perform satellite game deduction display. When the stop command sent by the satellite game simulation program is received, the program ends.

[0106] After the simulation system is initialized, a communication check is performed. After the communication connection is successful, the user performs scenario configuration and generates a scenario file. The user can choose a classic game scenario or build a new game scenario. The motion state of the two spacecraft is initialized in the scenario setting. After the scenario is built, the user can configure the spacecraft game pursuit algorithm parameters, spacecraft maneuverability, etc. After the above information is configured, it is saved to the scenario file. Before the simulation runs, the parallel simulation system sends the scenario file to the game simulation program through network communication. The game simulation program completes the initialization after reading the scenario file and waits for the parallel simulation system to send an instruction to start the simulation, such as Figure 4 and Figure 5 shown.

[0107] After the game simulation program is initialized, the user sends a start simulation command through the parallel simulation system, and then the game simulation program starts running, and the simulation data is sent to the parallel simulation system. The user controls the simulation process through the process deduction control module, such as simulation start, pause, end, etc., and can manually set the spacecraft maneuver strategy. After the deduction is completed, the user can backtrack according to the data file saved by the simulation system. During the backtracking process, for the key game strategy, you can choose to re-do it in the key round, that is, keep all the game data before the round, and start the subsequent game from this round. This process is like Figure 6 shown.

[0108] The simulation process is animated through the 3D display module, which can demonstrate the game situation in all directions and from multiple angles, allowing users to experience the most realistic spacecraft pursuit and escape game process. At the same time, users can use the historical game data in the sample library for animation playback, which can realize functions such as acceleration, deceleration and reverse playback. The dynamic, highly realistic game situation screen that simulates the real environment allows users to conduct more detailed analysis of each game round, helping users to further optimize the spacecraft game pursuit and escape algorithm.

[0109] The simulation method is based on the on-orbit pursuit and escape maneuver autonomous decision-making method of the spacecraft based on the proximal policy optimization algorithm (PPO). It adopts the method of training two spacecraft at the same time to alternately improve their autonomous decision-making ability. It constructs a suitable reward function considering the game task requirements and constraints. After completing the training, it has the ability to quickly, intelligently and efficiently generate orbital maneuvering strategies. A simulation platform is used to simulate multiple high-orbit satellite entity models; and multiple computers are used to run the simulation platform in parallel. The simulation platform installed on each of the multiple computers is distributed with at least one entity model of the multiple satellite entity models, which is used to simulate the multiple satellite models in parallel, wherein: each computer is used to monitor the simulation status of the satellite models distributed on each computer, obtain its own resource usage, and migrate the satellite models between computers according to the simulation status of the entity model and / or its own simulation data.

[0110] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A high-orbit and high-mobility satellite deduction method based on a distributed architecture, characterized in that: The following steps are involved: Step 1: According to the positions of the tracking satellite and the escape satellite, set the test orbit and the virtual reference satellite; Step 2, according to the motion state of the virtual reference satellite, the motion states of the spacecraft of the tracking satellite and the escaping satellite are determined respectively; Step 3: According to the CW equation and in combination with the motion states of the tracking satellite and the escaping satellite, the motion equations of the tracking satellite and the escaping satellite in the circular reference orbit are determined; Step 4. Establish a pursuit-escape game model based on the motion equations of the tracking satellite and the escaping satellite. Based on the principle that the game objectives of the pursuer and the escaping party in the pursuit-escape game are opposite, construct the reward function of the tracking satellite and the escaping satellite, and use the reinforcement learning optimization algorithm to iteratively train the pursuit-escape strategies of the tracking satellite and the escaping satellite. With the goal of maximizing their respective rewards, optimize the pursuit-escape strategies of the tracking satellite and the escaping satellite. After training, the tracking satellite and the escaping satellite output their respective pursuit-escape strategies.

2. According to the method of claim 1, the high-orbit and high-mobility satellite deduction method based on distributed architecture is characterized in that: The reference orbit is a circular reference orbit.

3. According to the method of claim 1, the high-orbit and high-mobility satellite deduction method based on distributed architecture is characterized in that: The method for determining the motion state of the spacecraft for the pursuit satellite and the escape satellite is as follows: constructing a reference spacecraft local orbital coordinate system in which the motion states of the spacecraft for the tracking satellite and the escape satellite are determined; The reference spacecraft local orbit coordinate system takes the center of mass of the virtual spacecraft as the origin of the coordinate system. The axis coincides with the reference spacecraft's geocentric radial vector, and the positive direction is from the center of the earth to the spacecraft; The axis is in the orbital plane The axes are perpendicular to each other, with the positive direction pointing in the direction of motion; The axis is determined by the right-hand rule, i.e. along the direction of the orbital momentum.

4. The high-orbit and high-mobility satellite deduction method based on a distributed architecture according to claim 3 is characterized in that: The motion state of the tracking satellite, ; The motion state of the escaped satellite .

5. The high-orbit and high-mobility satellite deduction method based on a distributed architecture according to claim 1 is characterized in that: The motion equations of the tracking satellite and the escaping satellite are expressed as follows: In the formula is the average angular velocity of the reference orbit, is the Earth's gravitational constant, is the reference orbit radius, , , are the thrust acceleration components of the spacecraft in the x, y, and z directions respectively.

6. The high-orbit and high-mobility satellite deduction method based on a distributed architecture according to claim 1 is characterized in that: In step 4, the PPO algorithm in reinforcement learning is used to iteratively train the pursuit and escape strategies of the tracking satellite and the escaping satellite. The observation information of the tracking satellite and the escaping satellite is the motion state information of the confrontation target. The actions of the tracking and escaping satellites are set to their velocity increments: The control quantity of the tracking satellite and the escape satellite is the thrust acceleration vector, which is expressed as follows: in, To reach the end of the pursuit game, For the current moment.

7. The high-orbit and high-mobility satellite deduction method based on a distributed architecture according to claim 1 is characterized in that: The expression of the reward function is as follows: 。 8. The high-orbit and high-mobility satellite deduction method based on a distributed architecture according to claim 1 is characterized in that: The reinforcement learning optimization algorithm is a proximal strategy optimization algorithm.

9. A system for executing the high-orbit and high-mobility satellite deduction method based on a distributed architecture as described in any one of claims 1 to 8, characterized in that: include: Initialization module, used to set the test orbit and virtual reference satellite according to the positions of the tracking satellite and the escape satellite; A motion state module is used to determine the motion states of the spacecraft of the tracking satellite and the escaping satellite respectively according to the motion state of the virtual reference satellite; The motion equation module is used to determine the motion equations of the tracking satellite and the escape satellite in the circular reference orbit according to the CW equation and the motion states of the tracking satellite and the escape satellite; The pursuit and escape strategy module is used to establish a pursuit and escape game model based on the motion equations of the tracking satellite and the escaping satellite. Based on the principle that the game objectives of the pursuer and the escaping party in the pursuit and escape game are opposite, the reward function of the tracking satellite and the escaping satellite is constructed, and the PPO algorithm in reinforcement learning is used to iteratively train the pursuit and escape strategies of the tracking satellite and the escaping satellite. The pursuit and escape strategies of the tracking satellite and the escaping satellite are optimized with the goal of maximizing their respective rewards. The trained tracking satellite and the escaping satellite output their respective pursuit and escape strategies.

10. A distributed satellite simulation system, characterized in that: Including simulation platform and simulation control system; The simulation platform is used to simulate multiple high-orbit satellite entity models; The simulation control system is used to execute the steps of the high-orbit and high-maneuverability satellite simulation method based on a distributed architecture as described in any one of claims 1 to 8, and control the pursuit process of the high-orbit satellite physical model according to the output pursuit strategy.

Images