Autonomous control method for knowledge reforming and intuitionistic function inspiration based on structure guidance

Abstract

The invention provides an autonomous control method based on structure-guided knowledge reforming and intuitionistic function inspiration, and relates to the technical field of collision avoidance control, and the method comprises the steps: carrying out the arrangement and classification of orbit evolution data under different orbit heights and different disturbance environments, and carrying out the structure guidance through an orbit dynamics model, constructing a state space, an action space and a reward function for reinforcement learning training, and training by adopting an offline reinforcement learning algorithm to obtain an avoidance control strategy library; acquiring a system state quantity and mapping by the avoidance control strategy library to obtain a reference control quantity; setting a prediction time domain and constructing a prediction model; and constructing a target function, and solving an optimization problem to obtain a control sequence. The method can improve the generalization and pertinence of a control strategy, reduces the optimization solution complexity, guarantees the satellite orbit control precision, supports the long-term autonomous operation of a spacecraft in a complex environment, and guarantees the high efficiency and reliability of the avoidance control of the spacecraft.

Description

Technical Field

[0001] This application relates to the field of collision avoidance control technology, specifically to an autonomous control method based on structure-guided knowledge reorganization and intuition function inspiration. Background Technology

[0002] During the operation of a satellite in orbit, the space environment is complex and ever-changing, and the risk of collision is ever-present. In order to ensure the safe and stable operation of the satellite in orbit, it is necessary to complete the accurate assessment of the space situation through core means such as dynamic modeling, optimized control, orbit prediction and environmental perception, and generate reliable collision avoidance control commands based on the assessment results. This is also the core design requirement and key implementation path of the autonomous collision avoidance control system for spacecraft.

[0003] In existing technologies, there are two main approaches to autonomous collision avoidance control for spacecraft. One approach uses Model Predictive Control (MPC) to solve for collision avoidance commands aimed at minimizing fuel consumption or minimizing maneuver time. Related research has also proposed an adaptive convexity method based on MPC, which optimizes the performance of MPC in handling collision avoidance constraints between spacecraft and rotating targets by introducing an ellipsoidal collision avoidance zone (KOZ) and an adaptive reference point method, thereby improving the computational efficiency of constraint handling. The other approach introduces Reinforcement Learning (RL) into the field of spacecraft collision avoidance control. Relying on the optimization characteristic of Reinforcement Learning, which does not rely entirely on the model, the optimal control strategy is learned autonomously through interaction with the environment. Related research models the collision avoidance problem as a partially observable Markov decision process (POMDP) ​​and uses a deep recurrent Q-network (DRQN) to train the decision agent, realizing autonomous collision avoidance decision planning for spacecraft under incomplete information.

[0004] However, both of the above-mentioned technical solutions have obvious defects in practical applications, making it difficult to meet the high real-time, high stability, and high safety requirements of spacecraft on-orbit collision avoidance control: Pure MPC-based methods require online solution of optimization problems. As the number of constraints and prediction step size increases, the computational load will expand dramatically. Even with the introduction of adaptive convexity strategies to optimize computational efficiency, there are still problems of large computational load and poor real-time performance when dealing with complex nonlinear constraints. This can easily lead to decision lag and failure to avoid collisions in time. Moreover, this method has poor adaptability to uncertain environments, and the solution process is prone to failure or divergence. Pure reinforcement learning methods, on the other hand, suffer from a lack of structural guidance and unstable policy output. The training process of this method is highly dependent on a large amount of effective data, which not only results in low training efficiency and long training time, but also, due to the lack of spacecraft dynamic structure guidance, can easily output control strategies that do not conform to the physical feasibility of motion. In complex or unknown space environments, the policy output is prone to inaccuracy or even divergence, which cannot guarantee the stability and safety of collision avoidance control and significantly increases the risk of spacecraft on-orbit operation. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this application provides an autonomous control method based on structure-guided knowledge reorganization and intuitionistic function heuristics. This application solves the problems of large computational load, poor real-time performance, lack of structure guidance, and unstable policies in existing autonomous collision avoidance control for spacecraft, which are common in Model Predictive Control (MPC) methods.

[0006] To achieve the above objectives, this application provides the following technical solution: In a first aspect, embodiments of this application provide an autonomous control method based on structure-guided knowledge reorganization and intuitionistic function heuristics. This method includes: selecting a corresponding orbital dynamics model to model and simulate orbital evolution based on the satellite's orbital altitude and known disturbance conditions, obtaining simulation data; combining the acquired historical dataset with the simulation data to organize and classify orbital evolution data under different orbital altitudes and disturbance environments, obtaining a basic orbital dataset covering multiple orbital regions and multiple disturbance scenarios; wherein the historical dataset includes the satellite's orbital parameters and orbital evolution records under different disturbance conditions; and based on the basic orbital dataset, performing structure-guided knowledge reorganization and intuition function heuristics through the orbital dynamics model. The process involves selecting, combining, and encoding parameters from the data to construct a state space, action space, and reward function for reinforcement learning training. An offline reinforcement learning algorithm is used to train and obtain an evasion control strategy library. System state variables containing the positions, velocities, safe distances, and altitudes of both friendly and target satellites are acquired and mapped from the evasion control strategy library to obtain reference control variables. Internal dynamic states and cost function states are distinguished, a prediction time domain is defined, and a prediction model is constructed. An objective function is constructed to penalize position deviations from the desired orbit, control variable deviations from the reference value, and predicted terminal position deviations. The optimization problem is solved to obtain a control sequence, which is then used to control the friendly satellite using a control law based on an intuitionistic function-inspired model predictive control (MPC).

[0007] Secondly, embodiments of this application provide an autonomous control system based on structure-guided knowledge reorganization and intuitionistic function inspiration. The system includes a modeling and simulation module, a data processing module, a strategy training module, a model building module, and an optimization control module.

[0008] Specifically, the modeling and simulation module is used to select appropriate orbital dynamics models to model and simulate orbital evolution based on the satellite's orbital altitude and known disturbance conditions, obtaining simulation data. The data processing module combines the acquired historical dataset with the simulation data, organizing and classifying orbital evolution data under different orbital altitudes and disturbance environments to obtain a basic orbital dataset covering multiple orbital regions and multiple disturbance scenarios. The historical dataset includes satellite orbital parameters and orbital evolution records under different disturbance conditions. The strategy training module, based on the basic orbital dataset, uses orbital dynamics models for structure guidance, selecting and combining parameters from the data. The system employs several modules: an encoding module to construct the state space, action space, and reward function for reinforcement learning training; an offline reinforcement learning algorithm to train an evasion control strategy library; a model building module to acquire system state variables including the position, velocity, safe distance, and altitude of both friendly and target satellites, mapping these variables to the evasion control strategy library to obtain reference control variables; a distinction between internal dynamic states and cost function states; a prediction time domain to define and construct a prediction model; and an optimization control module to construct an objective function to penalize position deviations from the desired orbit, control variable deviations from the reference value, and predicted terminal position deviations. Solving the optimization problem yields a control sequence, which is then used to control the friendly satellite using a model predictive control (MPC) law inspired by intuitionistic functions.

[0009] Thirdly, embodiments of this application provide an electronic device, which includes: a processor, a memory, and a program stored in the memory and executable on the processor. When the program is executed by the processor, it implements the autonomous control method based on structure-guided knowledge reorganization and intuitive function inspiration described in the first aspect.

[0010] Fourthly, embodiments of this application provide a computer-readable storage medium storing a program or instructions that, when executed by a processor, implement the structure-guided knowledge reorganization and intuitive function-inspired autonomous control method described in the first aspect above.

[0011] This application provides an autonomous control method based on structure-guided knowledge reorganization and intuitionistic function heuristics. Compared with existing technologies, it has the following advantages: This application constructs a basic orbital dataset covering multiple orbital regions and multiple disturbance scenarios by combining historical datasets and simulation data. It achieves structure guidance based on orbital dynamics models, standardizes the state space, action space, and reward function of reinforcement learning, and uses offline reinforcement learning training to obtain an evasion control strategy library. This library can quickly map reference control quantities based on system state variables. The evasion control strategy library provides reference control, avoiding the complex calculations of MPC from scratch and improving the real-time performance of spacecraft evasion control. Furthermore, this application distinguishes between internal dynamic states and cost function states, constructs a predictive model, and uses an intuitionistic function-inspired model to predict the control MPC, solving for the objective function that includes position deviation, control quantity deviation, and predicted end-position deviation penalties. Ultimately, it achieves autonomous control of the satellite, improving the generalization and specificity of the control strategy, reducing optimization complexity, ensuring satellite orbital control accuracy, supporting long-term autonomous operation of the spacecraft in complex environments, and ensuring the efficiency and reliability of spacecraft evasion control. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart illustrating an autonomous control method based on structure-guided knowledge reorganization and intuitive function inspiration provided in an embodiment of this application. Figure 2 This is an exemplary architecture diagram of an autonomous control method based on structure-guided knowledge reorganization and intuitive function inspiration provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an autonomous control system based on structure-guided knowledge reorganization and intuition function inspiration provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are described clearly and completely. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0016] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0017] The following section first introduces an autonomous control method based on structure-guided knowledge reorganization and intuitive function inspiration provided in the embodiments of this application.

[0018] This application provides a flowchart illustrating an autonomous control method based on structure-guided knowledge reorganization and intuitionistic function heuristics, as shown in the embodiments below. Figure 1 As shown, the structure-guided knowledge reorganization and intuitive function-inspired autonomous control method may include the following steps S110-S150.

[0019] S110. Based on the orbital altitude of our satellite and known disturbance conditions, select the appropriate orbital dynamics model to model and simulate the orbital evolution, and obtain simulation data.

[0020] S120. By combining the acquired historical dataset with simulation data, the orbital evolution data under different orbital altitudes and perturbation environments are organized and classified to obtain a basic orbital dataset covering multiple orbital regions and multiple perturbation scenarios. The historical dataset includes satellite orbital parameters and orbital evolution records under different perturbation conditions. Understandably, the basic orbital dataset provides realistic and diverse orbital samples for subsequent reinforcement learning training, thereby supporting the learning of control strategies that can operate stably under different environmental conditions.

[0021] S130. Based on the basic orbit dataset, the structure is guided by the orbit dynamics model. The parameters in the data are selected, combined and encoded to construct the state space, action space and reward function for reinforcement learning training. The avoidance control strategy library is obtained by training with offline reinforcement learning algorithm.

[0022] S140. Obtain the system state variables including the position, velocity, safe distance and altitude of the friendly satellite and the target satellite, and map the reference control variables from the avoidance control strategy library; distinguish between the internal dynamic state and the cost function state, set the prediction time domain and build the prediction model.

[0023] S150. Construct an objective function to penalize position deviation from the desired orbit, control deviation from the reference value, and predicted terminal position deviation. Solve the optimization problem to obtain the control sequence. Use the control law of the predictive control MPC through the intuitionistic function-inspired model to control the satellite.

[0024] The above describes the specific implementation of the autonomous control method based on structure-guided knowledge reorganization and intuitionistic function heuristics provided in this application. It should be noted that when a satellite is in orbit, it is subject to the combined effects of multiple factors, including the Earth's gravitational field, atmospheric drag, solar radiation pressure, and the gravitational forces of the Moon and the Sun. Regarding the orbital dynamics model, an ideal two-body model is selected under low-disturbance conditions. In the near-Earth orbit region, the J2 perturbation caused by the non-spherical effect of the Earth's gravitational field and atmospheric drag perturbations are considered. In the medium-to-high orbit region, perturbations such as the gravitational influence of the Moon and the Sun and solar radiation pressure are further introduced.

[0025] This application is primarily applied to the active avoidance control of space debris, other spacecraft, or target orbits during satellite operation. Understandably, this application constructs a basic orbital dataset covering multiple orbital regions and multiple disturbance scenarios by combining historical datasets and simulation data. It relies on orbital dynamics models for structure guidance, standardizes the state space, action space, and reward function of reinforcement learning, and uses offline reinforcement learning training to obtain an avoidance control strategy library. This library can quickly map reference control quantities based on system state variables. The avoidance control strategy library provides reference control, avoiding the complex calculations of MPC from scratch and improving the real-time performance of spacecraft avoidance control. This application generates an avoidance control strategy library from historical data through a knowledge reorganization framework combined with satellite dynamics models. Reinforcement learning uses this historical data for training, making the control strategies more stable and reliable, and avoiding the computational burden of retraining the model for each mission. This avoidance control strategy library can quickly select applicable strategies based on the current satellite state, providing reference control quantities and effectively reducing computational load.

[0026] Furthermore, this application simultaneously distinguishes between internal dynamic states and cost function states, constructs a predictive model, and employs an intuitive function-inspired model predictive control (MPC) to solve the objective function, which includes position deviation, control quantity deviation, and predicted end-position deviation penalties. Ultimately, it achieves autonomous control of its own satellite, which can improve the generalization and specificity of the control strategy, reduce the complexity of optimization solutions, ensure the accuracy of satellite orbit control, support the long-term autonomous operation of spacecraft in complex environments, and ensure the efficiency and reliability of spacecraft evasion control.

[0027] Based on this, this application innovatively combines reinforcement learning (RL) and model predictive control (MPC) to achieve efficient and stable autonomous evasive control. By generating an evasive control policy library through reinforcement learning training and combining it with the optimization control capabilities of intuitionistic function-inspired MPC, the high computational complexity of traditional MPC is avoided, while simultaneously improving the real-time performance and adaptability of the control policy.

[0028] This application addresses the shortcomings of existing autonomous collision avoidance control methods for spacecraft. Pure model predictive control methods suffer from high computational cost, poor real-time performance, and weak adaptability to uncertain environments, while pure reinforcement learning methods lack dynamic structure guidance, suffer from unstable strategies, and have low training efficiency. Both methods fail to meet the high real-time performance, high stability, and high safety requirements for on-orbit collision avoidance. Based on the method provided in this application, an autonomous control architecture capable of online inference and offline updates can be constructed, supporting long-term autonomous operation of spacecraft in complex environments and ensuring the efficiency and reliability of spacecraft collision avoidance control.

[0029] In one example, the state space design process includes: S210. Based on the orbit determination and environmental perception results represented by the basic orbit dataset, and in conjunction with the selected orbit dynamics model, the orbital and environmental quantities in the basic orbit dataset are selected and organized to obtain multiple parameters directly related to the avoidance task; these parameters include: the position of our own satellites. and speed The location of the target satellite or space debris that needs to be avoided. and speed Safe distance threshold and orbital height .

[0030] S220. Encode the state of each sample in a unified manner to obtain a comprehensive state vector. : Based on the need for comprehensive state vector The coordinates of each state component are unified, the scale is normalized, and the numerical stability is processed.

[0031] Among them, the location of our own satellite and speed It is used to characterize the orbital state of one's own satellite in a selected reference coordinate system; the location of target satellites or space debris that need to be avoided. and speed The reference frame is the same as that of our own satellites. Safety distance threshold. Pre-set according to mission requirements and collision risk control requirements, and used to determine whether there is a potential collision threat; orbital altitude Used to characterize the current orbital region and correspond to the corresponding orbital dynamics model.

[0032] In one example, the orbital dynamics model in an inertial frame satisfies the expression: In the formula, The position of the satellite in the inertial frame. for The second derivative with respect to time, for The modulus length; It is perturbation acceleration; To control the acceleration term; It is a constant describing the Earth's fundamental gravitational force.

[0033] Actions in the action space are used to characterize the reference control quantity within a control cycle. Reference control quantity Used to determine maneuver commands, the action form in the action space corresponds to the control acceleration term in the selected orbital dynamics model. .

[0034] In one example, the reward function includes safety metrics and resource consumption metrics to guide reinforcement learning in making trade-offs between avoidance effects and resource consumption.

[0035] The safety indicator is the relative distance between friendly satellites and target satellites. : If relative distance Greater than the preset safe distance threshold If the relative distance is [missing information], then a positive reward is given in the current step of the strategy; if the relative distance [missing information], then a positive reward is given in the current step of the strategy. Less than the safe distance threshold If the current step strategy applies a negative reward to punish the control sequence that failed to be effectively avoided, then a negative reward will be applied.

[0036] The resource consumption index is the fuel consumption value corresponding to each maneuver, which is added to the reward function as a penalty. The fuel consumption value is added to the reward function as a penalty and is positively correlated with the penalty to guide reinforcement learning to prioritize the avoidance maneuver strategy with the lowest fuel consumption under the premise of meeting safety requirements, so as to reduce the overall propellant consumption.

[0037] In the embodiments of this application, it is understood that the reward function composed of two types of indicators, safety and resource consumption, enables reinforcement learning to automatically converge to an avoidance strategy that is both safe and economical under different orbital altitudes and perturbation conditions during the training process, providing a reasonable reference control trend for subsequent online optimization.

[0038] In some embodiments, during the training of the evasion control policy library using an offline reinforcement learning algorithm, the offline reinforcement learning method employed can be a value function-based algorithm, a policy-based algorithm, or a combination of both. During training, multiple orbital altitude ranges and their corresponding two-body model or perturbation model samples can be included simultaneously. Furthermore, during the data generation phase, factors such as dynamic parameter uncertainties, perturbation variations, and measurement noise are considered to improve the robustness and generalization ability of the final policy under different environmental conditions.

[0039] After multiple rounds of iterative training and convergence testing, a reference control quantity (control policy) is obtained, which satisfies the expression: ;in, This is a nonlinear mapping obtained through offline reinforcement learning; the mapping is stored in the form of a policy library for real-time in-orbit retrieval.

[0040] In some embodiments, the aforementioned acquisition of system state variables including the positions, velocities, safe distances, and altitudes of the friendly and target satellites is mapped to a reference control variable from an evasion control strategy library; distinguishing between internal dynamic states and cost function states, setting a prediction time domain, and constructing a prediction model, i.e., the aforementioned S140 may specifically include the following steps: S310, Obtain System state variables at time 1 System state variables Includes the location of our own satellites Own satellite speed Target satellite position Target satellite velocity Safe distance and height .

[0041] S320. The reference control quantity is calculated based on the avoidance control strategy library. Reference control quantity From system state variables The mapping is obtained.

[0042] S330. Distinguish between the internal dynamic state and the state used in the cost function, whereby the internal dynamic state... Used to advance orbital evolution in predictive models, and satisfying: The cost function selects the location of the user's own satellite to participate in the quadratic penalty, and the state vector used in the cost function... satisfy .

[0043] S340, Set the prediction time domain length as... An internal prediction model is constructed based on a pre-defined orbital dynamics model. The orbital evolution is advanced by using the internal prediction model to predict the current and future orbital dynamics states and corresponding control variables.

[0044] In some alternative embodiments, the aforementioned objective function penalizes deviations from the desired orbit, deviations from the reference value in the control quantity, and deviations from the predicted terminal position. Solving the optimization problem yields a control sequence, which is then used to control the satellite by predicting the control law of the control MPC through an intuitionistic function-inspired model. Specifically, S150 may include the following steps: S410. Constructing the objective function objective function Penalties for deviations of the satellite's position from the desired orbit and deviations of control variables from the reference control variables. It also penalizes deviations in the position of one's own satellites at the end of the predicted time domain; S420. Solve the optimization problem corresponding to the above objective function to obtain the future... A control sequence of one prediction step; only the first step of the control sequence is executed, and the internal dynamic state is updated according to the actual satellite dynamics model; S430. Determine whether the relative distance between our satellite and the target satellite after the update has reached the preset safe distance. If the conditions are met, the current control operation ends and the next control step begins; otherwise, the above steps are repeated until orbit avoidance is completed.

[0045] In some embodiments, the aforementioned structure-guided knowledge reorganization and intuitive function-inspired autonomous control method may further include: constructing a data feedback mechanism to feed back the actual data obtained during the execution of each evasion task to the historical dataset for retraining or incremental updates of subsequent offline reinforcement learning.

[0046] Understandably, after each task is completed, this application feeds back the data from the current task to the reinforcement learning system to update the historical dataset. By continuously updating the dataset, the reinforcement learning model can continuously optimize the policy library, improve the intelligence and adaptability of the control policy, and enable the system to generate control commands more efficiently when facing new tasks. Through this data feedback mechanism, subsequent offline reinforcement learning training or retraining can simultaneously utilize the initially constructed simulated orbital data and the real task data accumulated during on-orbit operation. This allows the policy library to maintain its adaptability to different orbital altitudes and disturbance environments while gradually absorbing experiential information from actual tasks, thereby continuously improving the quality of the reference control and better supporting subsequent intuitionistic function-inspired MPC autonomous control.

[0047] In summary, the evasion control strategy library of this application takes the current state of the satellite as input and outputs reference control quantities. These reference control quantities provide initial guidance for the control system, avoiding the computational burden of searching for control strategies from scratch. The reference control quantities are passed as input to an intuitionistic function-inspired objective function, and the control quantities are further optimized using an optimization algorithm based on model predictive control (MPC) to obtain the optimal control quantity. The optimal control quantity is input into the satellite control system, and after execution, the satellite's state is updated, and it is determined whether the evasion task has been completed. If the task is not completed, it will start from scratch and continue reasoning based on the current state until successful evasion; if completed, the data will be fed back to the reinforcement learning system to update the historical dataset and retrain to generate a new strategy library. Through this mechanism combining reinforcement learning and model predictive control (MPC) inference, this application can provide efficient, stable, and real-time control strategies in complex environments.

[0048] In some embodiments, this application provides an autonomous control system 500 based on structure-guided knowledge reorganization and intuitionistic function inspiration, such as... Figure 3 As shown, the autonomous control system 500 may include the following modules: The modeling and simulation module 510 is used to select the appropriate orbital dynamics model to model and simulate the orbital evolution based on the orbital altitude of the satellite and the known disturbance conditions, and to obtain simulation data. The data processing module 520 is used to combine the acquired historical dataset with simulation data, organize and classify the orbit evolution data under different orbital altitudes and different disturbance environments, and obtain a basic orbit dataset covering multiple orbital regions and multiple disturbance scenarios; among which, the historical dataset includes the satellite's orbital parameters and orbit evolution records under different disturbance conditions; The policy training module 530 is used to select, combine and encode parameters in the data based on the basic orbit dataset and through the orbit dynamics model to construct the state space, action space and reward function for reinforcement learning training, and to obtain the avoidance control policy library by training with offline reinforcement learning algorithm. The model building module 540 is used to obtain system state variables including the position, velocity, safe distance and altitude of the friendly satellite and the target satellite, and to obtain reference control variables by mapping from the avoidance control strategy library; it distinguishes between internal dynamic state and cost function state, sets the prediction time domain and builds the prediction model; The optimization control module 550 is used to construct an objective function to penalize position deviations from the desired orbit, control deviations from the reference value, and predict terminal position deviations. The optimization problem is solved to obtain a control sequence, which is then used to control the satellite by predicting the control law of the control MPC through an intuitive function-inspired model.

[0049] According to embodiments of this application, any and multiple modules among the modeling and simulation module 510, data processing module 520, policy training module 530, model building module 540, and optimization control module 550 can be combined into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module.

[0050] Figure 3 Each module in the system shown has the function of implementing each step in the aforementioned autonomous control method based on structure-guided knowledge reorganization and intuition function inspiration, and can achieve its corresponding technical effect. For the sake of brevity, it will not be elaborated here.

[0051] In some embodiments, this application provides an electronic device, the structural schematic of which is shown below. Figure 4 As shown.

[0052] The electronic device may include a processor 610 and a memory 620 storing computer program instructions.

[0053] Specifically, the processor 610 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0054] Memory 620 may include mass storage for data or instructions. For example, and not limitingly, memory 620 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 620 may include removable or non-removable (or fixed) media. Where appropriate, memory 620 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 620 is non-volatile solid-state memory.

[0055] Memory 620 may include read-only memory (ROM), random access memory (RAM), disk storage media device, optical storage media device, flash memory device, electrical, optical, or other physical / tangible memory storage device. Therefore, typically, memory 620 includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it can perform the operations described in any of the structure-guided knowledge reorganization and intuitive function-inspired autonomous control methods in the above embodiments.

[0056] The processor 610 reads and executes computer program instructions stored in the memory 620 to implement any of the structure-guided knowledge reorganization and intuitive function-inspired autonomous control methods in the above embodiments.

[0057] In one example, the electronic device may also include a communication interface 630 and a bus 600. Wherein, such as Figure 4 As shown, the processor 610, memory 620, and communication interface 630 are connected via bus 600 and communicate with each other.

[0058] The communication interface 630 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0059] Bus 600 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 600 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0060] Furthermore, in conjunction with the structure-guided knowledge reorganization and intuitionistic function-inspired autonomous control methods described in the above embodiments, this application embodiment can provide a computer storage medium for implementation. This computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the structure-guided knowledge reorganization and intuitionistic function-inspired autonomous control methods described in the above embodiments.

[0061] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0062] The functional blocks shown in the above block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0063] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0064] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0065] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for autonomous control based on structure-guided knowledge reorganization and intuitionistic function heuristics, characterized in that, include: Based on the orbital altitude of our satellite and known disturbance conditions, we select an appropriate orbital dynamics model to model and simulate the orbital evolution, and obtain simulation data. By combining the acquired historical dataset with the simulation data, the orbital evolution data under different orbital altitudes and different disturbance environments are organized and classified to obtain a basic orbital dataset covering multiple orbital regions and multiple disturbance scenarios; wherein, the historical dataset includes the satellite's orbital parameters and orbital evolution records under different disturbance conditions; Based on the aforementioned basic orbit dataset, the orbit dynamics model is used for structural guidance. Parameters in the data are selected, combined, and encoded to construct a state space, action space, and reward function for reinforcement learning training. An evasion control strategy library is obtained by training with an offline reinforcement learning algorithm. The system state variables, including the positions, velocities, safe distances, and altitudes of both friendly and target satellites, are acquired and mapped to reference control variables from the evasion control strategy library. Internal dynamic states and cost function states are distinguished, a prediction time domain is defined, and a prediction model is constructed. An objective function is constructed to penalize position deviations from the desired orbit, control deviations from the reference value, and predicted terminal position deviations. The optimization problem is solved to obtain the control sequence, which is then used to control the satellite by predicting the control law of the control MPC through an intuitionistic function-inspired model.

2. The autonomous control method based on structure-guided knowledge reorganization and intuitionistic function heuristics as described in claim 1, characterized in that, The design process of the state space includes: Based on the orbit determination and environmental perception results represented by the basic orbit dataset, and in conjunction with the selected orbit dynamics model, the orbital and environmental quantities in the basic orbit dataset are selected and organized to obtain multiple parameters directly related to the avoidance task; these multiple parameters include: the position of the friendly satellite. and speed The location of the target satellite or space debris that needs to be avoided. and speed Safe distance threshold and orbital height ; The state of each sample is uniformly encoded to obtain a comprehensive state vector. : Based on the need for comprehensive state vector The coordinates of each state component in the process are unified, the scale is normalized, and the numerical stability is processed. Among them, the location of one's own satellite and speed This is used to characterize the orbital state of one's own satellite in a selected reference coordinate system; the location of the target satellite or space debris to be avoided. and speed The reference frame is the same as that of our own satellite; The safe distance threshold The track height is pre-set according to mission requirements and collision risk control requirements, and is used to determine whether there is a potential collision threat; Used to characterize the current orbital region and correspond to the corresponding orbital dynamics model.

3. The autonomous control method based on structure-guided knowledge reorganization and intuitionistic function heuristics as described in claim 2, characterized in that, The orbital dynamics model in the inertial frame satisfies the following expression: In the formula, The position of the satellite in the inertial frame. for The second derivative with respect to time, for The modulus length; It is perturbation acceleration; To control the acceleration term; A constant describing the Earth's fundamental gravitational force; The actions in the action space are used to characterize the reference control quantity within a control cycle. The reference control quantity Used to determine maneuver commands, the action form in the action space corresponds to the control acceleration term in the selected orbital dynamics model. .

4. The autonomous control method based on structure-guided knowledge reorganization and intuitionistic function heuristics as described in claim 3, characterized in that, The reward function includes safety metrics and resource consumption metrics to guide reinforcement learning in making a trade-off between avoidance effects and resource consumption. The security indicator is the relative distance between our satellite and the target satellite. : If the relative distance Greater than the preset safe distance threshold If the relative distance is such that a positive reward is given in the current step of the strategy, then a positive reward is given; if the relative distance is such that a positive reward is given in the current step of the strategy, then a positive reward is given in the current step of the strategy. Less than the safe distance threshold If the current step strategy applies a negative reward to punish the control sequence that was not effectively avoided; The resource consumption index is the fuel consumption value corresponding to each maneuver, which is added to the reward function as a penalty. The fuel consumption value is added to the reward function as a penalty and is positively correlated with the penalty to guide reinforcement learning to prioritize the avoidance maneuver strategy with the lowest fuel consumption while meeting safety requirements, so as to reduce overall propellant consumption.

5. The autonomous control method based on structure-guided knowledge reorganization and intuitionistic function heuristics as described in claim 2, characterized in that, The system state variables, including the positions, velocities, safe distances, and altitudes of the friendly and target satellites, are obtained by mapping reference control variables from the avoidance control strategy library. Distinguish between internal dynamic states and cost function states, define the prediction time domain, and construct a prediction model, including: Get System state variables at time 1 The system state variables Includes the location of our own satellites Own satellite speed Target satellite position Target satellite velocity Safe distance and height ; The reference control quantity is calculated based on the aforementioned evasion control strategy library. The reference control quantity From the system state variables Mapped to obtain; Distinguishing between internal dynamic states and the states used in the cost function, where internal dynamic states... Used to advance orbital evolution in predictive models, and satisfying: The cost function selects the location of one's own satellite to participate in the quadratic penalty, and the state vector used in the cost function... satisfy ; Set the prediction time domain length to An internal prediction model is constructed based on a preset orbital dynamics model. The orbital evolution is advanced by the internal dynamics state and corresponding control variables of the current and future prediction steps through the internal prediction model.

6. The autonomous control method based on structure-guided knowledge reorganization and intuitionistic function heuristics as described in claim 5, characterized in that, The objective function is constructed to penalize deviations from the desired orbit, deviations from the reference value in the control input, and deviations from the predicted terminal position. Solving the optimization problem yields a control sequence, which is then used to control the satellite via a control law based on an intuitionistic function-inspired model predictive control MPC. This includes: Construct the objective function The objective function Penalties for deviations of the satellite's position from the desired orbit and deviations of control variables from the reference control variables. It also penalizes deviations in the position of one's own satellites at the end of the predicted time domain; Solve the optimization problem corresponding to the above objective function to obtain the future A control sequence of prediction steps; only the first step of the control sequence is executed, and the internal dynamic state is updated according to the actual satellite dynamics model; Determine whether the relative distance between our satellite and the target satellite after the update reaches the preset safe distance. If the conditions are met, the current control operation ends and the next control procedure begins; otherwise, the above steps are repeated until orbit avoidance is completed.

7. The autonomous control method based on structure-guided knowledge reorganization and intuitionistic function heuristics as described in claim 1, characterized in that, Also includes: A data feedback mechanism is constructed so that after each evasion task is completed, the actual data obtained during the execution of this task is fed back to the historical dataset for retraining or incremental updates of subsequent offline reinforcement learning.

8. An autonomous control system based on structure-guided knowledge reorganization and intuitionistic function inspiration, characterized in that, include: The modeling and simulation module is used to select the appropriate orbital dynamics model to model and simulate the orbital evolution based on the orbital altitude of the satellite and known disturbance conditions, and to obtain simulation data. The data processing module is used to combine the acquired historical dataset with the simulation data, and to organize and classify the orbit evolution data under different orbital altitudes and different disturbance environments to obtain a basic orbit dataset covering multiple orbital regions and multiple disturbance scenarios; wherein, the historical dataset includes the satellite's orbital parameters and orbit evolution records under different disturbance conditions; The strategy training module is used to select, combine and encode parameters in the data based on the basic orbit dataset and the orbit dynamics model, construct the state space, action space and reward function for reinforcement learning training, and train the avoidance control strategy library using offline reinforcement learning algorithm. The model building module is used to obtain system state variables including the position, velocity, safe distance and altitude of the friendly satellite and the target satellite, and to map the reference control variables from the avoidance control strategy library; to distinguish between internal dynamic state and cost function state, to set the prediction time domain and to build a prediction model; The optimization control module is used to construct an objective function to penalize position deviations from the desired orbit, control deviations from the reference value, and predict terminal position deviations. Solving the optimization problem yields a control sequence, which is then used to control the satellite by predicting the control law of the control MPC through an intuitionistic function-inspired model.

9. An electronic device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the autonomous control method based on structure-guided knowledge reorganization and intuitive function inspiration as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program or instructions that, when executed by a processor, implement the autonomous control method based on structure-guided knowledge reorganization and intuitive function heuristics as described in any one of claims 1 to 7.

Images