AI satellite constellation-oriented computing power distribution and scheduling method and system

Abstract

The invention is suitable for the technical field of satellite constellations, and provides an AI satellite constellation-oriented computing power distribution and scheduling method, which comprises the following steps of: distributing a seat node for each satellite based on the orbit characteristics of the satellites, and configuring a basic computing power resource for each seat node; constructing a computing power flow potential field based on the orbit characteristics of the satellite and the basic computing power resources; based on the coverage range parameters of the satellite, calculating power demand evaluation is carried out on tasks in each coverage area, and a task demand gradient is generated; in combination with the computing power flow potential field and the task demand gradient, redistributing computing power resources for each seat node; and performing performance evaluation and performance verification on the newly-added satellites to enable the newly-added satellites to be gradually fused to the constellation network, thereby effectively improving the computing power resource utilization rate of the AI satellite constellation, optimizing the task scheduling efficiency and guaranteeing the stable operation of the system.

Description

Technical Field

[0001] The present application belongs to the field of satellite constellation technology, and in particular relates to a computing power allocation and scheduling method and system for AI satellite constellations. Background Art

[0002] With the rapid development of satellite constellations, AI technology is being used increasingly widely in satellite missions. An AI satellite constellation is a networked system consisting of multiple AI-capable satellites. Through inter-satellite communication and collaborative computing, it can directly process data in space and transmit the processing results back to the ground, greatly improving timeliness and data processing efficiency.

[0003] In the existing technology, the coordination between satellite orbit characteristics, computing power requirements and task types is not fully considered in the computing power allocation and scheduling process of satellite constellations, resulting in unreasonable allocation of computing power resources and inefficient task processing. In addition, the problem of computing power scheduling between newly added satellites and the existing network has not been effectively solved, resulting in uneven resource allocation and affecting the overall performance of the constellation. Summary of the Invention

[0004] The embodiments of the present application provide a computing power allocation and scheduling method and system for AI satellite constellations, which can solve one of the above-mentioned existing technical problems.

[0005] In a first aspect, an embodiment of the present application provides a computing power allocation and scheduling method for an AI satellite constellation, including: Based on the orbital characteristics of the satellite, a seat node is assigned to each satellite, and each seat node is configured with basic computing resources; Based on the orbital characteristics of the satellite and the basic computing power resources, a computing power flow potential field is constructed; Based on the satellite coverage parameters, the computing power requirements of the tasks in each coverage area are evaluated to generate a task demand gradient; Combining the computing power flow potential field and the task demand gradient, reallocating computing power resources to each agent node; The performance of the newly added satellites is evaluated and verified, so that the newly added satellites are gradually integrated into the constellation network.

[0006] Furthermore, based on the orbital characteristics of the satellite, a seat node is allocated to each satellite, and each seat node is configured with basic computing resources, including: Each satellite is considered as an agent node, and each agent node is marked with a unique identifier, orbital altitude parameters, computing power performance data and real-time status information; Based on the orbital altitude parameters and the computing power performance data, the orbital characteristic differences between different satellites are obtained, and the basic seat weight of each satellite is determined; Obtain reliability measurement data of each satellite through the satellite's fault diagnosis system and historical operation data, quantify the availability of the agent node based on the reliability measurement data, and construct a reliability scoring matrix; Determine the spatiotemporal reachability index of each of the agent nodes based on the reliability scoring matrix and in combination with satellite orbit parameters and coverage parameters; Based on the basic seat weight, reliability scoring matrix and spatiotemporal accessibility index, the priority weight of the computing power resource allocation of each seat node is determined, and differentiated basic computing power resources are configured for different satellites.

[0007] Furthermore, the construction of a computing power flow potential field based on the satellite's orbital characteristics and the basic computing power resources includes: Based on the differences in the orbital heights of each satellite, a spatial hierarchical distribution matrix from low orbit to high orbit is constructed; Determine the remaining computing capacity of each satellite based on its basic computing resources, computing performance data, and real-time status information; For each satellite, the energy efficiency weight factor is calculated based on power consumption and remaining computing capacity. Combined with the reliability score matrix, the comprehensive potential energy value is obtained through weighted calculation method. Combining the spatial hierarchical distribution matrix and the comprehensive potential energy value, a gradient field distribution model is constructed, and the gradient strength and direction vector between adjacent satellites are calculated by the potential energy difference to form a computing power flow potential field.

[0008] Furthermore, determining the remaining computing capacity of each satellite based on the basic computing resources, computing performance data, and real-time status information of each satellite includes: Obtain the corresponding basic computing resources, computing performance data and real-time status information from the seat nodes of each satellite; Based on the real-time status information, obtaining a load distribution view; Based on the load distribution view, combined with the basic computing resources and the computing performance data, the resource utilization ratio of each seat node is calculated to obtain the remaining computing capacity.

[0009] Furthermore, based on the satellite coverage parameters, computing power requirements of tasks within each coverage area are evaluated to generate a task requirement gradient, including: Determine the coverage area of each satellite based on the satellite's orbital parameters and the Earth's rotation parameters; In each of the coverage areas, calculating the regional computing power demand density based on the nature of the task and computing requirements; Based on the regional computing power demand density of each coverage area, the task demand gradient is determined, where the task demand gradient is used to measure the spatial change rate of computing power demand.

[0010] Furthermore, within each of the coverage areas, it further includes: Acquire mission request data from a ground station, perform preprocessing operations on the mission request data, and obtain mission requirement characteristic values; Based on the task requirement characteristic values, combined with task priorities and resource requirements, a multidimensional task requirement vector is constructed to determine the task distribution characteristics of the tasks in the multidimensional space; According to the task distribution characteristics, the tasks are accurately located in the vector space to obtain dimensional coordinates; Based on the dimensional coordinates, combined with task priority and latency requirements, the priority order of task execution is determined.

[0011] Furthermore, the reallocation of computing power resources to each agent node in combination with the computing power flow potential field and the task demand gradient includes: Calculate the computing power allocation tendency coefficient for the areas covered by two adjacent agent nodes based on the task demand gradient and computing power flow potential field strength; For each agent node, the total tendency coefficient of computing power allocation to all adjacent coverage areas is calculated based on the computing power allocation tendency coefficient; Based on the total tendency coefficient, the computing power distribution of each seat node is determined.

[0012] Furthermore, the performance evaluation and performance verification of the newly added satellites are performed to gradually integrate the newly added satellites into the constellation network, including: Obtaining hardware configuration parameters and orbital position information of the newly added satellite, and performing a quantitative performance evaluation on the newly added satellite; Assigning an initial verification task to the newly added satellite to obtain performance baseline data; Using the performance baseline data as a reference standard, gradually increasing the number of tasks and computational complexity for the newly added satellite to determine the load adaptability of the newly added satellite; If the load adaptability meets the expected standards, the newly added satellite is added to the constellation network and basic computing resources are allocated based on priority weights.

[0013] In a second aspect, an embodiment of the present application provides a computing power allocation and scheduling system for an AI satellite constellation, including: The first processing module is used to allocate a seat node to each satellite based on the orbital characteristics of the satellite, and each seat node is configured with basic computing resources; The second processing module is used to construct a computing power flow potential field based on the orbital characteristics of the satellite and the basic computing power resources; The third processing module is used to evaluate the computing power requirements of tasks in each coverage area based on the satellite coverage parameters and generate a task demand gradient; A fourth processing module is configured to reallocate computing power resources to each agent node based on the computing power flow potential field and the task demand gradient; The fifth processing module is used to perform performance evaluation and performance verification on the newly added satellite, so that the newly added satellite can be gradually integrated into the constellation network.

[0014] In a third aspect, an embodiment of the present application provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above-mentioned computing power allocation and scheduling method for AI satellite constellations is implemented.

[0015] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, comprising a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned computing power allocation and scheduling method for AI satellite constellations.

[0016] Compared with the prior art, the embodiments of the present application have the following beneficial effects: The present invention discloses a computing power allocation and scheduling method for AI satellite constellations. Based on the satellite orbit characteristics, seat nodes are allocated to each satellite and basic computing power resources are configured to construct a computing power flow potential field. The method can dynamically adjust the allocation of computing power resources according to factors such as the satellite's position in orbit and its operating cycle. Furthermore, the computing power demand of tasks in each coverage area is evaluated based on satellite coverage range parameters, and a task demand gradient is generated. The computing power flow potential field is combined with the computing power flow potential field to reallocate computing power resources to the seat nodes, thereby achieving optimized scheduling of computing power resources and task requirements. In addition, the present invention also introduces a new satellite access and performance evaluation mechanism, gradually increases the load through a lightweight task scheduler, and evaluates adaptability to meet the computing power scheduling running-in between the newly added satellite and the existing constellation network. Therefore, the present invention can effectively improve the computing power resource utilization of the AI satellite constellation, optimize the task scheduling efficiency, and ensure stable operation of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0018] Figure 1This is a flowchart of a computing power allocation and scheduling method for an AI satellite constellation provided by one embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a computing power allocation and scheduling system for AI satellite constellations provided by one embodiment of the present invention; Figure 3 It is a structural diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0020] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0021] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0022] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0023] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0024] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0025] See also Figure 1 As shown, the present invention is a computing power allocation and scheduling method for AI satellite constellations, comprising the following steps: S100. Based on the orbital characteristics of the satellite, an agent node is allocated to each satellite, and each agent node is configured with basic computing resources; In some embodiments, step S100 includes: Each satellite is considered as an agent node, and each agent node is marked with a unique identifier, orbital altitude parameters, computing power performance data and real-time status information; Based on the orbital altitude parameters and the computing power performance data, the orbital characteristic differences between different satellites are obtained, and the basic seat weight of each satellite is determined; Obtain reliability measurement data of each satellite through the satellite's fault diagnosis system and historical operation data, quantify the availability of the agent node based on the reliability measurement data, and construct a reliability scoring matrix; Determine the spatiotemporal reachability index of each of the agent nodes based on the reliability scoring matrix and in combination with satellite orbit parameters and coverage parameters; Based on the basic seat weight, reliability score and spatiotemporal accessibility index, the priority weight of the computing power resource allocation of each seat node is determined, and differentiated basic computing power resources are configured for different satellites.

[0026] In this embodiment, a constellation network management system based on the agent concept is established to facilitate the management of satellites in different orbital layers in the constellation network. In this system, each satellite is regarded as an agent node, and corresponding node configuration information is configured for each agent node, including a unique identifier, orbital altitude parameters, computing power performance data and real-time status information.

[0027] Specifically, the unique identifier is a unique identity identifier for each satellite. In one possible embodiment, quantum entanglement is introduced during the process of generating the identity identifier using a hash algorithm. Based on the principle that quantum states cannot be cloned, the communication process is highly secure. Specifically, an entangled photon pair is generated by a quantum entanglement source, such as a polarized lithium niobate waveguide at a ground station to generate an entangled photon pair with a wavelength of 1550nm, specifically photon A and photon B. The generated photon A is then retained in the quantum memory of the ground station for subsequent verification. Photon B is transmitted to the corresponding satellite via a laser uplink, providing quantum state input for generating the unique identifier. Furthermore, the satellite randomly selects a measurement basis, such as the Z basis or the X basis, to measure the received photon B, and the measurement result is recorded as mB, which specifically has a value of 0 or 1. Furthermore, the satellite's computing power, power consumption, and orbit type are extracted, and an initial identity is generated using a hash algorithm. Finally, the quantum measurement result mB is concatenated with the initial identity to generate the final unique identifier. To be more specific, when performing identity authentication, the ground station performs the same basis vector measurement on the retained photon A as the satellite, obtains the measurement result mA, and calculates the quantum correlation to verify the entanglement state. Specifically, if the Z basis measurement is used, when mA=mB, the correlation is 1; otherwise it is 0. If the X basis measurement is used, the correlation will fluctuate around 0.5.

[0028] In some embodiments, a registration transaction including information such as a satellite unique identifier, entangled photon pairs, initial identity, entanglement correlation, etc. is constructed and represented in JSON format to ensure that the satellite registration information is credible and cannot be tampered with.

[0029] Specifically, the orbital altitude parameter refers to the vertical distance from the center of mass of a satellite to the surface of the earth when it orbits the earth. According to the difference in orbital altitude and inclination, satellite orbits are mainly divided into low earth orbit, medium earth orbit and geostationary orbit. Each orbit type has unique advantages and limitations and is suitable for different application fields. For example, low earth orbit is close to the earth, has low signal transmission delay (millisecond level), high earth observation resolution, fast satellite operation speed and short orbital period. It is usually used in earth observation (high-resolution imaging), remote sensing, space science experiments and emerging giant broadband communication constellations. Medium earth orbit has a moderate orbital altitude and a long orbital period, taking into account both coverage and communication delay. A single satellite The satellite coverage range is greater than that of low-Earth orbit, and the number of satellites required to achieve global coverage is less than that of low-Earth orbit. It is usually used in global navigation satellite systems. For geostationary orbit, since its orbital period is exactly 24 hours, it is synchronized with the rotation of the earth, so that the satellite remains stationary relative to a certain point on the ground, and a single satellite can cover one-third of the earth's surface except the poles. It is usually used in communication broadcasting (live TV, satellite phone), meteorological observation (providing continuous cloud map monitoring), etc. Therefore, different tasks choose the corresponding orbit type according to their mission requirements. For example, navigation systems generally choose medium-Earth orbit to balance coverage, stability and latency, while earth observations mostly choose low-Earth orbit to obtain high-resolution and consistent lighting images.

[0030] The computing power performance data includes specific indicator data describing the satellite computing power, such as the number of CPU cores, main frequency, and floating-point computing capability, while the real-time status information includes data describing the real-time operation status, such as availability and load conditions.

[0031] In this embodiment, the basic computing power resources are the computing power pre-allocated to each satellite, that is, the tasks sent by the ground station are pre-allocated to the corresponding agent nodes based on the priority weights generated by the node configuration information of each agent node, thereby providing a basic computing power resource benchmark for the computing power allocation in the constellation network, and subsequently making dynamic adjustments based on actual mission requirements and satellite status.

[0032] In this embodiment, the orbital altitude information and computing power performance data of each satellite are obtained to form an initial data set. Each row of the initial data set should contain the satellite's unique identifier, orbital altitude parameters and computing power performance data. According to the orbital altitude information, the satellites are divided into different orbit types. Furthermore, for satellites of different orbital types, the corresponding orbital characteristic parameters such as orbital period, average orbital velocity and coverage range are extracted, and the corresponding orbital characteristic parameters are associated with the unique identification point to form an orbital characteristic parameter data set. Then, the satellite's orbital type, each orbital characteristic parameter in the orbital characteristic parameter data set and each specific indicator data in the computing power performance data are standardized using the min-max normalization method, and the values of the above indicators are mapped to the [0,1] interval. Then, the expert scoring method is used to assign a corresponding weight coefficient to each indicator, and the basic seat weight is calculated based on the standardized indicator value and weight coefficient.

[0033] In this embodiment, the reliability measurement data of each satellite is obtained through the satellite's fault diagnosis system and historical operation data, specifically the number of faults, fault type, fault repair time, etc., and then the availability of the agent node is quantitatively analyzed, and the average failure interval evaluation algorithm is used to calculate the average failure interval of each satellite. The specific calculation formula is: average failure interval = total operation time / number of failures, and a reliability scoring matrix is constructed, in which the rows of the reliability scoring matrix represent different satellite agent nodes, and the columns represent different reliability indicators and the calculated reliability scores. The reliability indicators specifically include the average failure interval and fault recovery time. It can be understood that the reliability score is calculated by standardizing each reliability indicator, assigning a weight coefficient to each reliability indicator using the expert scoring method, and then using the weighted average method.

[0034] In this embodiment, the reliability score of each satellite in the reliability score matrix is obtained to reflect the reliability of different agent nodes. The higher the reliability score, the lower the probability of failure of the node during operation, providing a basic reference for subsequent spatiotemporal reachability calculations. Furthermore, the satellite orbit parameters are obtained, specifically the orbital period, orbital semi-major axis, orbital eccentricity, orbital inclination, ascending node right ascension, perigee angle and other parameters, and then combined with the satellite orbit type to determine the satellite coverage parameters. Based on the satellite orbit parameters, the Kepler orbit model is used to calculate the satellite position. It can be understood that the Kepler orbit model is a model based on Kepler's three laws, which is used to calculate the position of the satellite at any time through given orbital parameters. Furthermore, combined with the reliability score, the accessibility probability P of the satellite at a specific time and space position is calculated. Specifically, assuming that the basic accessibility probability of the satellite when the distance condition is met is P0, then based on the reliability score, its accessibility probability P=S×P0, where S represents the reliability score of the satellite. It is worth noting that P0 is set based on experience or experiments. In a preferred embodiment, P0=0.8, and then the position coordinates and corresponding accessibility probabilities of the seat nodes of each satellite at different time points are recorded to form a spatiotemporal accessibility index data set. For example, at time t At 1, the position coordinates of Node_001 are (x1, y1, z1), and the accessibility probability is P1=0.2264. Then the service window duration is calculated. Specifically, the service window duration refers to the time period when the satellite covers the target area and meets the communication conditions. Due to the rotation of the earth, the position of the target area relative to the satellite will continue to change. At the same time, the orbital motion of the satellite will also change its relative position with the target area. Specifically, the initialization time t=0, the satellite's position at time t and the distance d from the center point of the target area are calculated to determine whether the satellite meets the coverage of the target area and the communication conditions. If so, the current time t is recorded as the start time t1 of the service window. As time t increases, the satellite's position and its relative relationship with the target area are continuously calculated. When the satellite no longer meets the coverage of the target area or the communication conditions, the current time t is recorded as the end time t2 of the service window. At this time, the service window duration Ts=t2−t1 is calculated. Then, the spatiotemporal accessibility index of the agent node of each satellite is measured by the accessibility probability and the service window duration.

[0035] In some embodiments, the hierarchical analysis method is used to determine the weight coefficients of the basic seat weight, reliability score and spatiotemporal accessibility index in the calculation of the computing power resource allocation priority weight, and then the weighted average method is used to calculate the computing power resource allocation priority weight of each seat node based on the weight coefficient of each indicator and the index value of the above-mentioned basic seat weight, reliability score and spatiotemporal accessibility index.

[0036] S200: Constructing a computing power flow potential field based on the satellite's orbital characteristics and the basic computing power resources; This application allocates seat nodes to each satellite and configures basic computing resources based on the characteristics of the satellite orbit, constructs a computing flow potential field, and can dynamically adjust the allocation of computing resources according to factors such as the satellite's position in orbit and its operating cycle.

[0037] In some embodiments, step S200 includes: Based on the differences in the orbital heights of each satellite, a spatial hierarchical distribution matrix from low orbit to high orbit is constructed; Determine the remaining computing capacity of each satellite based on its basic computing resources, computing performance data, and real-time status information; For each satellite, the energy efficiency weight factor is calculated based on power consumption and remaining computing capacity. Combined with the reliability score, the comprehensive potential energy value is obtained through weighted calculation method. Combining the spatial hierarchical distribution matrix and the comprehensive potential energy value, a gradient field distribution model is constructed, and the gradient strength and direction vector between adjacent satellites are calculated by the potential energy difference to form a computing power flow potential field.

[0038] In this embodiment, corresponding orbital altitude information is obtained from the seat node of each satellite. The orbital altitudes are grouped using a hierarchical clustering method. The orbit types of each satellite are further stratified from low orbit to high orbit, a spatial hierarchical distribution matrix is constructed, and the satellite distribution range corresponding to each level is determined. Specifically, based on the differences in orbital altitude and inclination, the satellite orbits of each satellite are initially divided into orbit types such as low Earth orbit, medium Earth orbit, and geostationary orbit. Then, an appropriate similarity measurement method, such as Euclidean distance, is selected to calculate the similarity between the orbital altitudes of different satellites within the same orbital type. Starting with each data point as a separate cluster, the closest clusters are gradually merged until a certain stopping condition is met, such as the intra-cluster similarity reaching a certain threshold. Through this process, satellites with similar orbital altitudes are grouped together. Thus, the satellites are divided into multiple levels according to their orbital altitude differences. Finally, a spatial hierarchical distribution matrix is generated to describe the satellite distribution and satellite distribution range at different levels. The satellite distribution specifically refers to the functional type of each satellite, such as the number of communication satellites or the number of observation satellites, and the satellite distribution range specifically refers to the range of satellite orbital altitudes within the level.

[0039] In this embodiment, based on the basic computing resources allocated to each agent node in step S100, the remaining computing resources of the agent nodes are calculated to determine the node potential distribution in the constellation network. Specifically, the remaining computing resources are assigned to nodes with higher node potential, resulting in a computing flow potential field. Specifically, for each satellite, an energy efficiency weighting factor is calculated based on power consumption and remaining computing capacity: energy efficiency weighting factor = power consumption / remaining computing capacity. Power consumption refers to the electrical power consumed by the satellite during operation, reflecting the energy input required to maintain normal operation and perform its mission. For example, if satellite A consumes 100W of power, it means that satellite A consumes 10 joules of power per second. The higher the power consumption, the greater the satellite's energy consumption. Then, based on the reliability scoring matrix in step S100, a reliability score for the corresponding satellite is obtained, thereby generating a comprehensive potential value for the satellite: P = energy efficiency weighting factor × reliability score.

[0040] Furthermore, the comprehensive potential energy value is combined with the space level distribution matrix to construct a gradient field distribution model. Specifically, in the space level distribution matrix, each level has a corresponding potential energy base value, that is, the potential energy of each satellite is determined by the potential energy base value and the comprehensive potential energy value of its level, and then the gradient strength and direction vector between adjacent satellites are calculated by the potential energy difference, specifically from the satellite with low potential energy to the satellite with high potential energy, and finally a computing power flow potential field is formed. Specifically, the space level distribution matrix divides the space where the satellite is located into different levels, such as the low orbit level and the medium orbit level, and sets a potential energy base value for each level. The potential energy base value reflects the difference in potential energy characteristics of different space levels themselves. In a possible embodiment, The basic value of the low-orbit layer potential energy is 1, and the basic value of the medium-orbit layer potential energy is 0.8. The comprehensive potential energy value reflects the contribution of various factors other than the space layer to the potential energy of the satellite. In this application, it is specifically the electric power consumed by the satellite during operation. Based on this, when calculating the satellite potential energy, the comprehensive potential energy value is multiplied by the layer potential energy basic value, and finally the comprehensive potential energy value is combined with the space layer distribution matrix to construct a gradient distribution model. Through this gradient field distribution model, the distribution of satellite potential energy in the entire space can be intuitively displayed. At the same time, the potential energy of each satellite, the potential energy difference of different space layers, and the potential energy change of satellites at different positions in the same layer can be determined, and then the potential energy difference between adjacent satellites is calculated, and finally a computing power flow potential field is formed.

[0041] In some embodiments, determining the remaining computing capacity of each satellite based on the basic computing resources, computing performance data, and real-time status information of each satellite includes: Obtain the corresponding basic computing resources, computing performance data and real-time status information from the seat nodes of each satellite; Based on the real-time status information, obtaining a load distribution view; Based on the load distribution view, combined with the basic computing resources and the computing performance data, the resource utilization ratio of each seat node is calculated to obtain the remaining computing capacity.

[0042] In this embodiment, the load distribution view is a view based on the real-time status information of each satellite, which is used to display the current load distribution of each satellite. It specifically describes the resource usage during the current operation of the satellite, including computing resource usage, storage resource usage and communication resource usage. Through the load distribution view, you can intuitively see which satellites have higher loads and which satellites have lower loads.

[0043] Furthermore, the basic computing power resources pre-allocated by the constellation network management system to the seat nodes of each satellite are obtained, specifically the basic computing resources, basic storage resources and basic communication resources required for the execution of each task pre-allocated to the corresponding satellite, that is, the resources theoretically required for the satellite to execute the pre-allocated tasks.

[0044] Specifically, based on the satellite's computing power performance data, the total computing resources, total storage resources, and total communication resources available for mission execution on each satellite are obtained. Combined with the load distribution view, the remaining computing power capacity of the satellite is obtained. Specifically, the remaining computing resources = total computing resources - used computing resources, the remaining storage resources = total storage resources - used storage resources, and the remaining communication resources = total communication resources - used communication resources. The final remaining computing power capacity is then obtained through weighted calculation using the formula: remaining computing power capacity = w1 × remaining computing resources + w2 × remaining storage resources + w3 × remaining communication resources. Here, w1, w2, and w3 are the weights of computing resources, storage resources, and communication resources, respectively, and the sum of the weights is 1.

[0045] Furthermore, for the resource utilization ratio, the ratio between the use of computing resources in the actual operation process and the resources theoretically required for the satellite to perform the pre-assigned tasks is calculated, that is, the ratio of resource usage to basic computing resources. This is associated with the weight of each resource in the remaining computing capacity. That is, the weight of each resource is dynamically adjusted according to the resource utilization ratio, so as to more accurately reflect the remaining computing power of different satellites in the current state. Specifically, the calculation formula for the weight of each resource is as follows: ,in, Represents the original weight of each resource, a represents the adjustment coefficient, and its value range is [0, 1]. It is used to control the adjustment amplitude of resource utilization ratio on weight. When a is larger, the influence of resource utilization ratio on weight is greater. When a=0, the influence of resource utilization ratio on weight is not considered, and it degenerates to the original scheme. Indicates the resource utilization ratio. Specifically, for computing resources, the resource utilization ratio U1 = used computing resources / basic computing resources, for storage resources, the resource computing ratio U2 = used storage resources / basic storage resources, and for communication resources, the resource utilization ratio U3 = used communication resources / basic communication resources.

[0046] In addition, the resource utilization ratio can be used to discover waste or insufficiency in resource allocation and load balance during task execution, thereby optimizing the resource allocation strategy. In some embodiments, if the actual resource usage of a satellite when executing a task is far higher than the theoretical demand, it indicates that there may be a performance bottleneck in the task on the satellite. The task can be transferred to a satellite with better performance by adjusting the task allocation strategy.

[0047] S300: Based on the satellite coverage parameters, perform computing power demand assessment for tasks within each coverage area and generate a task demand gradient; In some embodiments, the computing power requirement assessment of tasks within each coverage area based on satellite coverage parameters to generate a task requirement gradient includes: Determine the coverage area of each satellite based on the satellite's orbital parameters and the Earth's rotation parameters; In each of the coverage areas, calculating the regional computing power demand density based on the nature of the task and computing requirements; Based on the regional computing power demand density of each coverage area, the task demand gradient is determined, where the task demand gradient is used to measure the spatial rate of change of computing power demand.

[0048] In this embodiment, the coverage area of each satellite is determined based on the satellite's orbital parameters and Earth's rotation parameters. Specifically, orbital parameters, including key data such as inclination, right ascension of the ascending node, argument of perigee, orbital altitude, and orbital period, are obtained through tracking and measurement at ground stations. These parameters are used to determine the satellite's coverage area. Orbital inclination, specifically the angle between the orbital plane and the equatorial plane, determines the latitude range covered by the satellite. For example, when the inclination is 0°, the satellite covers only the equatorial region; when the inclination is 90°, the satellite covers the polar regions. The right ascension of the ascending node, specifically the longitude of the ascending node between the orbital plane and the equatorial plane in the direction of the vernal equinox, determines the starting longitude of the coverage area. The argument of perigee, specifically the angle between the perigee and the ascending node, affects the satellite's phase in orbit. The orbital altitude determines the ground coverage area of a single satellite. Specifically, a polar-orbiting satellite (90° inclination) has an ascending node right ascension of 30°E and an orbital altitude of 800km. Its coverage area rotates around the poles and gradually covers the entire globe as the Earth rotates.

[0049] More specifically, the ground coverage of a single satellite is usually determined by calculating the trigonometric relationship between the Earth's radius and the satellite's orbital altitude. The specific calculation formula is: , where h represents the satellite's orbital altitude parameter, represents the radius of the Earth, then represents the distance from the satellite to the center of the earth, and r is the coverage radius, which specifically represents the radius of the satellite coverage area on the ground. Therefore, the ground coverage range of a single satellite is measured by the above coverage radius.

[0050] Specifically, the rotation of the earth causes the satellite coverage area to change over time, so the angular velocity of the earth's rotation is introduced to obtain the ground area dynamically covered by the satellite per orbital period. Specifically, , calculate the satellite's orbital period T, where h represents the satellite's orbital height parameter, represents the Earth's gravitational coefficient, approximately 3.986×10 14 m³ / s², while the Earth's rotational speed is It is about 15° / hour or 360° / 24 hours. Thus, the ground area dynamically covered by the satellite per orbital period is determined. Specifically, for a satellite with an orbital altitude parameter of 800km, its orbital period is calculated to be T≈100 minutes. At this time, the ground area covered by the satellite per orbital period will deviate 25° to the west due to the rotation of the earth. Specifically, the deflection angle .

[0051] Furthermore, in this embodiment, for the convenience of analysis and management, the coverage area is divided into multiple geographic grid area blocks. Specifically, the earth is divided into multiple geographic grid area blocks of longitude × latitude, such as 1°×1°, or 2°×1°, and as the latitude of the coverage area increases, the grid area of the corresponding geographic grid area block becomes smaller. Specifically, 1° longitude near the equator corresponds to approximately 111 km, which corresponds to the length of the longitude. Since the latitude is the longest at the equator, the distance corresponding to 1° latitude is also longer, and the grid area is large; while the latitude is short in high latitudes, the distance corresponding to 1° latitude is short, so the grid area is smaller. Further specifically, according to the coverage radius and satellite orbit The system dynamically adjusts the size of the geographic grid area to ensure that each geographic grid area is completely within the coverage area. Furthermore, each geographic grid area must be assigned a unique ID and its geographic boundaries, i.e., its longitude and latitude range, must be recorded. Specifically, the coverage area of a polar-orbiting satellite can be divided into a strip-shaped grid along the orbital direction. Based on the orbital altitude parameter, the coverage radius of the coverage area is calculated, and the longitude range corresponding to each grid width within the coverage radius is calculated. In one possible embodiment, if the coverage radius is 2800 km, the longitude range of one lap of the Earth's surface is 360°, and the ground distance corresponding to 1° of longitude near the equator is approximately 111 km. The longitude range corresponding to the coverage radius can then be calculated as follows: Δλ = d / r. It is worth noting that this calculation is approximate, as the ground distance corresponding to 1° of longitude at different latitudes varies with latitude. Therefore, in the strip-shaped grid along the satellite's orbital direction, each grid cell has a width of approximately 25° in the longitude direction, which facilitates clearer analysis of satellite coverage characteristics and mission planning.

[0052] Furthermore, due to the periodicity of satellite orbits and the Earth's rotation, the coverage area will repeat. Therefore, by recording the geographic grid area blocks covered by the satellite at each time point (such as every minute), a coverage timetable is formed to facilitate subsequent calculations and analysis.

[0053] In this embodiment, tasks are divided into different types based on their nature and computational requirements, such as data processing tasks, image analysis tasks, and communication relay tasks. Each task type has different computational complexity and resource requirements. Initial computing resources are set for each task type, and then the computing requirements of tasks within each coverage area are assessed. Specifically, computing requirements can be quantified based on factors such as the task's data volume, computational complexity (such as the number of floating-point operations), and processing time requirements. For example, an image analysis task may require high computing resources for pixel processing and feature extraction.

[0054] To be more specific, for each coverage area, calculate its computing power demand density. The computing power demand density is specifically the computing power demand per unit area. Specifically, the computing power demand density of a coverage area is The specific calculation formula is: ,in, Indicates the computing power requirement of the jth task in the area, is the area of the coverage area, and n represents the number of tasks in the coverage area.

[0055] Specifically, based on the computing power demand density of each coverage area, the task demand gradient is determined to represent the spatial change rate of computing power demand of each coverage area. For two adjacent areas i and j, the task demand gradient can be expressed as: ,in, and are the computing power demand densities of coverage area i and coverage area j respectively, It is the geographical distance between two regional centers. It can be understood that the direction of the gradient points from the area with low computing power demand density to the area with high computing power demand density, which is used to indicate that computing power resources should flow from low-demand areas to high-demand areas to meet the task processing needs of high-demand areas.

[0056] In some embodiments, because mission requirements and satellite coverage areas change over time, a dynamic adjustment mechanism is established to regularly recalculate the mission requirement gradient and adjust the computing resource allocation strategy. At the same time, a feedback mechanism is established to monitor and evaluate the effectiveness of computing resource allocation. If computing resources in a certain area are insufficient or excessive, the resource allocation strategy is adjusted in a timely manner to ensure continuous optimization of the system.

[0057] In some embodiments, within each of the coverage areas, the following further comprises: Acquire mission request data from a ground station, perform preprocessing operations on the mission request data, and obtain mission requirement characteristic values; Based on the task requirement characteristic values, combined with task priorities and resource requirements, a multidimensional task requirement vector is constructed to determine the task distribution characteristics of the tasks in the multidimensional space; According to the task distribution characteristics, the tasks are accurately located in the vector space to obtain dimensional coordinates; Based on the dimensional coordinates, combined with task priority and latency requirements, the priority order of task execution is determined.

[0058] In this embodiment, the task request data within the satellite coverage area is further processed to obtain the priority ranking of each task execution, which facilitates the subsequent efficient execution of tasks within the coverage area. In addition, when allocating computing power resources to each seat node, it can also be based on the priority ranking of the tasks. Specifically, the tasks with higher rankings are preferentially allocated to the computing power calculation task execution operations in the corresponding satellite.

[0059] In this embodiment, task request data is obtained from a ground site through a wired or wireless communication link. Specifically, the ground site can be a ground control center, a user terminal, etc. The task request data specifically includes various information in the task request process, including computational complexity, latency requirements, task priority, and resource requirements, etc. The computational complexity can be measured by indicators related to the time complexity or space complexity of the algorithm, such as the specific form in the big O notation, such as O(n²), etc. The latency requirement is specifically the maximum time interval allowed from the initiation of the task request to the completion of the processing. The task priority is preliminarily divided into multiple levels according to the importance or urgency of the task, such as high, medium, and low priority. The resource requirement specifically includes the number of CPU cores, memory size, storage space, etc. required for the task execution process.

[0060] In this embodiment, the task request data obtained above is checked to remove duplicate data, erroneous data, and data missing key information. Furthermore, data in different formats, such as text format and JSON format, are uniformly converted into a structured data format, such as a database table format or a structured task requirement data set with a specific data structure. A preset standardization method is used to normalize parameters such as computational complexity and latency requirements in the structured task requirement data set. Specifically, a Z-score normalization method is used to convert the data into a distribution with a mean of 0 and a standard deviation of 1. The normalized parameters such as computational complexity and latency requirements are converted into a unified dimension to obtain the task requirement characteristic value under the unified dimension, so that the task requirement characteristic value can participate in subsequent analysis and calculation more fairly.

[0061] More specifically, a multi-dimensional task requirement vector is constructed based on the task requirement characteristic value, combined with the task priority and resource requirement. For example, the multi-dimensional task requirement vector of a task can be expressed as v=(v1,v2,v3,v4), which specifically represents the four dimensions of the task, namely computational complexity, latency requirement, task priority, and resource requirement. It can be understood that v1 is the normalized computational complexity, v2 is the normalized latency requirement, and v3 is the priority value corresponding to the task priority. The corresponding priority values are pre-set for different task priorities, such as high priority is 3, medium priority is 2, and low priority is 1. v4 is the value corresponding to the resource requirement, which is specifically the value obtained after quantifying each parameter in the resource requirement.

[0062] Specifically, using vector computation methods, we can calculate the Euclidean distance or cosine similarity between the multidimensional demand vectors of different tasks to obtain preliminary distribution characteristics of each task in the multidimensional space. For example, we can calculate the Euclidean distance between the demand vectors of two tasks and use the size of this distance to preliminarily determine the distribution of tasks in the multidimensional space.

[0063] To be more specific, the task is accurately positioned in the vector space. Specifically, the coordinate mapping technology is used to map the task to the specific position coordinates in the vector space. That is, each task can correspond to a multi-dimensional coordinate (x, y, z, w). Through the coordinate mapping technology, the specific position coordinates of the task in the vector space can be accurately determined, and the position relationship of the task in the multi-dimensional space can be more accurately described.

[0064] Furthermore, a priority sorting method is used to determine the order in which tasks are executed, based on their specific location coordinates in vector space, combined with their priority and latency requirements. For example, tasks are sorted first based on priority, with higher-priority tasks placed first. For tasks of equal priority, they are then sorted based on latency requirements, with tasks with more urgent latency requirements placed first. This sorting method prioritizes task execution, ensuring that important and urgent tasks are processed first.

[0065] In this embodiment, the above-mentioned priority sorting is used to generate a corresponding resource allocation strategy for the coverage area corresponding to each satellite. Specifically, the constellation network management system receives the task requirements sent by the ground station, and the constellation network management system assigns them to the corresponding satellite orbit according to their task type to wait for task arrangement. Furthermore, according to the priority sorting of each task, each task is pre-allocated to the seat node of the corresponding satellite orbit, and each task is configured with initial computing power resources. It can be understood that the initial computing power resources are the computing power resources required to execute the task due to its computational complexity. Therefore, the initial computing power resources of multiple tasks assigned to the same seat node are integrated and finally pre-allocated to the corresponding seat node in the form of basic computing power resources. Specifically, each task is assigned to the corresponding satellite, and the corresponding task is pre-executed through the computing resources, storage resources and communication resources in the satellite, and then dynamically adjusted based on the satellite status.

[0066] S400: reallocate computing power resources to each agent node based on the computing power flow potential field and the task demand gradient; This application evaluates the computing power requirements of tasks in each coverage area based on satellite coverage parameters, generates a task demand gradient, and reallocates computing power resources to agent nodes in combination with the computing power flow potential field, thereby achieving optimal scheduling of computing power resources and task requirements.

[0067] In some embodiments, step S400 includes: Calculate the computing power allocation tendency coefficient for the areas covered by two adjacent agent nodes based on the task demand gradient and computing power flow potential field strength; For each agent node, the total tendency coefficient of computing power allocation to all adjacent coverage areas is calculated based on the computing power allocation tendency coefficient; Based on the total tendency coefficient, the computing power distribution of each seat node is determined.

[0068] In this embodiment, the calculation formula for the computing power allocation tendency coefficient is specifically: , where i represents a coverage area; j represents any adjacent coverage area of coverage area i, ; N(i) represents the set of areas adjacent to coverage area i, Represents the absolute value of the task demand gradient, ensuring that the tendency coefficient is positive; It represents the proportion of the computing power demand gap of adjacent coverage area j to the total demand gap of all adjacent coverage areas of coverage area i, which is used to make the allocation more inclined to areas with relatively large demand gaps. Indicates the computing power demand gap of region k. If If the value of is positive, it means that the computing power demand of area k is greater than the current computing power, and there is insufficient computing power; if If the value of is negative, it means that the computing power of region k is surplus; It represents the gradient strength between coverage area i and coverage area j, that is, the computing power flow potential field strength. Combined with step S200, it can be seen that it is specifically the potential energy difference between the satellite corresponding to coverage area i and the satellite corresponding to coverage area j.

[0069] More specifically, for each coverage area i, the total propensity coefficient for allocating computing power to all adjacent areas is calculated. The specific formula is: , further, for each adjacent area pair (i, j), calculate the proportion of computing power allocated from coverage area i to coverage area j , the specific formula is: On this basis, the computing power allocated from coverage area i to coverage area j is calculated, that is, the computing power distribution between seat nodes. The formula is: ,in, Indicates the distribution of computing power among agent nodes. Represents the total amount of computing resources available for reallocation, based on which the computing resource utilization efficiency of the entire constellation network is maximized while meeting the task processing needs of each region.

[0070] In this embodiment, the computing power distribution between the agent nodes is specifically the computing power distribution between different satellites. In the satellite constellation, inter-satellite communication is achieved through the inter-satellite link technology of the satellite constellation, thereby realizing the dynamic allocation of computing power resources. Specifically, the inter-satellite link technology is a wireless link for direct communication between satellites without the need for transit through a ground station.

[0071] In this embodiment, a constellation network management system is provided, and each satellite regularly broadcasts its own computing resource status to the constellation network management system through inter-satellite links. In the constellation network management system, each satellite is specifically a seat node, and each seat node is configured with corresponding node configuration information. Each satellite regularly updates the corresponding information in the node configuration information, specifically power consumption, reliability score, real-time status information, computing performance data, etc. Furthermore, through the inter-satellite links, the constellation network management system performs task control between satellites based on the computing power distribution between the seat nodes it obtains, thereby realizing the computing power distribution and scheduling of the satellites. Specifically, the inter-satellite communication network adopts a dynamic routing algorithm to select the optimal transmission path according to the real-time link status. Specifically, if there is a multi-hop path between satellite A and satellite B (such as A→C→B), the routing algorithm will select the path with the lowest delay.

[0072] In this embodiment, when any satellite receives computing power resources, it updates its own computing power status to the constellation network management system. The constellation network management system broadcasts it to other satellites through inter-satellite links and records and monitors it in real time, dynamically optimizing the allocation of computing power resources to meet the mission requirements of different regions.

[0073] S500: Perform performance evaluation and performance verification on the newly added satellite, so that the newly added satellite is gradually integrated into the constellation network.

[0074] This application also introduces a new satellite access and performance evaluation mechanism, which gradually increases the load through a lightweight task scheduler and evaluates adaptability to meet the computing power scheduling and running-in between the newly added satellites and the existing constellation network.

[0075] In some embodiments, step S500 includes: Obtaining hardware configuration parameters and orbital position information of the newly added satellite, and performing a quantitative performance evaluation on the newly added satellite; Assigning an initial verification task to the newly added satellite to obtain performance baseline data; Using the performance baseline data as a reference standard, gradually increasing the number of tasks and computational complexity for the newly added satellite to determine the load adaptability of the newly added satellite; If the load adaptability meets the expected standards, the newly added satellite is added to the constellation network and basic computing resources are allocated based on priority weights.

[0076] In this embodiment, the hardware configuration parameters and orbital position information of the newly added satellite are obtained, and the processor performance of the added satellite is quantitatively evaluated through a computing power benchmark test program to obtain standardized computing power performance data and orbital altitude parameters. Specifically, the hardware data and configuration parameters of the newly added satellite are extracted from the satellite system database to complete the initial information collection and obtain a preliminary hardware configuration file. Based on the preliminary hardware configuration file, the processor performance-related data of the satellite is obtained, and a preset computing power benchmark test program is used to run the program to determine the performance data of the processor under different loads. The processor performance data is quantified using a standardized indicator system to obtain standardized computing power performance data. Based on the standardized computing power performance data, a seat node of the new satellite is created in the constellation network management system, and a distributed registration protocol is used to synchronize the seat node and the corresponding node configuration information to the existing constellation network to obtain a unique identifier for the newly added satellite.

[0077] In this embodiment, an initial verification task is assigned to the new satellite through a lightweight task scheduler. If the task execution time exceeds a preset threshold, the task complexity is reduced. If the execution time is within the expected range, the performance baseline data is recorded. Specifically, an initial verification task is assigned to the new satellite through a lightweight task scheduler, and the execution time data after the task assignment is obtained to determine whether the task execution is within the preset threshold range. If the task execution time exceeds the preset threshold, the complexity adjustment mechanism is triggered, and the task complexity is dynamically reduced to obtain the adjusted task configuration data. According to the adjusted task configuration data, the task is reallocated to the new satellite, and new execution time data is obtained to determine whether it meets the expected range. If the new execution time data is within the expected range, the performance baseline data is stored to determine the stability index of the current task configuration.

[0078] In some embodiments, a support vector machine algorithm is used to perform time monitoring and analysis on the stored performance baseline data to obtain the fluctuation trend characteristics of the execution time. Based on the fluctuation trend characteristics, the allocation strategy of the task scheduler is adjusted to obtain an optimized task allocation plan to determine whether the task execution efficiency has been improved. Through the optimized task allocation plan, the matching degree between the execution time of the new satellite and the performance baseline data is continuously monitored to determine the stability parameters of long-term operation.

[0079] In this embodiment, performance baseline data is used as a reference standard to gradually increase the number of tasks assigned to new satellites and their computational complexity. By monitoring task completion rates and response delays in real time, the adaptability of the current load level is determined. Specifically, a reasonable increase in the number of tasks and computational complexity is determined based on the performance baseline data. For example, the number of tasks can be increased by 10% or the computational complexity can be increased by 5%. To avoid excessive impact on the satellite, the intervals between load increases need to be reasonably scheduled. For example, the load can be increased every 24 hours to provide sufficient time to observe the performance of the satellite system after the load increase. Subsequently, more tasks are gradually assigned to the satellite according to the load increase plan. For example, the number of tasks can be increased by 10% initially, increasing the original 100 tasks assigned to the satellite daily to 110. At the same time, the computational complexity of the tasks is gradually increased. This can be achieved by increasing the data processing capacity of the tasks and introducing more complex algorithms. For example, tasks that originally only required simple data statistics can be replaced by tasks requiring complex data analysis.

[0080] In some embodiments, the load adaptability of the satellite is comprehensively judged. Specifically, when judging the load adaptability, multiple indicators are comprehensively considered, such as judging based on multiple indicators such as task completion rate or response delay. Specifically, the task completion status of the satellite is collected in real time, and the start time, end time and completion status of each task are recorded. Based on the collected data, the task completion rate is calculated in real time, and the time interval between the issuance of each task and the start of the satellite to process the task, that is, the response delay, is recorded. Based on the recorded response time, the average response delay is calculated in real time.

[0081] To be more specific, the task completion rate obtained by real-time monitoring is compared with the average task completion rate and threshold in the performance baseline data. If the real-time task completion rate is within the threshold range and is not much different from the average task completion rate, it means that the satellite can complete the task well under the current load. If the real-time task completion rate is lower than the lower limit threshold, it means that the satellite may not be able to withstand the current load and needs to be adjusted. In addition, the average response delay obtained by real-time monitoring is compared with the average response delay and threshold in the performance baseline data. If the real-time average response delay is within the threshold range and is not much different from the average response delay, it means that the satellite's response speed is normal under the current load. If the real-time average response delay exceeds the upper limit threshold, it means that the satellite's response speed has slowed down and may not be able to process the task in time, and needs to be adjusted.

[0082] In this embodiment, based on the load level adaptability evaluation results, if the processing capacity of the new satellite meets the expected standards, it will be added to the constellation network and obtain the same task allocation weight as the existing satellites. That is, based on the basic seat weight, reliability score and spatiotemporal accessibility index of the satellite, the priority weight of the computing power resource allocation of the corresponding seat node is determined, the corresponding basic computing power resources are allocated to it, and the computing power is redistributed and scheduled in the subsequent task processing process.

[0083] See also Figure 2 As shown, the present invention also provides a computing power allocation and scheduling system for AI satellite constellations, the system comprising: The first processing module 201 is configured to allocate a seat node to each satellite based on the orbital characteristics of the satellite, and each seat node is configured with basic computing resources; The second processing module 202 is configured to construct a computing power flow potential field based on the orbital characteristics of the satellite and the basic computing power resources; The third processing module 203 is used to evaluate the computing power requirements of tasks in each coverage area based on the satellite coverage parameters and generate a task requirement gradient; The fourth processing module 204 is configured to reallocate computing resources to each agent node based on the computing power flow potential field and the task demand gradient; The fifth processing module 205 is configured to perform performance evaluation and performance verification on the newly added satellite, so that the newly added satellite can be gradually integrated into the constellation network.

[0084] It is understandable that if Figure 1 The contents of the embodiment of the computing power allocation and scheduling method for AI satellite constellations shown in the figure are applicable to the embodiment of the computing power allocation and scheduling system for AI satellite constellations. The functions specifically implemented by the embodiment of the computing power allocation and scheduling system for AI satellite constellations are similar to those in the embodiment of FIG. Figure 1 The computing power allocation and scheduling method embodiment for AI satellite constellations shown in FIG is the same as that shown in FIG. Figure 1 The beneficial effects achieved by the embodiment of the computing power allocation and scheduling method for AI satellite constellations shown are also the same.

[0085] It should be noted that the information interaction, execution process and other contents between the above-mentioned systems are based on the same concept as the embodiment of the method of the present invention. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0086] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0087] See also Figure 3 As shown, an embodiment of the present invention further provides a computer device 3, including: a memory 302 and a processor 301 and a computer program 303 stored on the memory 302. When the computer program 303 is executed on the processor 301, the computing power allocation and scheduling method for the AI satellite constellation as described in any one of the above methods is implemented.

[0088] The computer device 3 may be a desktop computer, a notebook computer, a PDA, a cloud server or other computing devices. The computer device 3 may include, but is not limited to, a processor 301 and a memory 302. Those skilled in the art will understand that Figure 3 This is merely an example of the computer device 3 and does not constitute a limitation on the computer device 3 . The computer device 3 may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the computer device 3 may also include input and output devices, network access devices, etc.

[0089] The processor 301 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor may be a microprocessor or any conventional processor.

[0090] In some embodiments, the memory 302 may be an internal storage unit of the computer device 3, such as a hard drive or memory of the computer device 3. In other embodiments, the memory 302 may also be an external storage device of the computer device 3, such as a plug-in hard drive, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc. equipped on the computer device 3. Furthermore, the memory 302 may include both an internal storage unit of the computer device 3 and an external storage device. The memory 302 is used to store an operating system, application programs, a boot loader, data, and other programs, such as the program code of the computer program. The memory 302 may also be used to temporarily store data that has been output or is about to be output.

[0091] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the computing power allocation and scheduling method for an AI satellite constellation as described in any one of the above methods is implemented.

[0092] In this embodiment, if the integrated unit is implemented as a software functional unit and sold or used as a standalone product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application can implement all or part of the process steps in the above-mentioned method embodiments by using a computer program to instruct the relevant hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a camera / computer device, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signals, telecommunication signals, and software distribution media. Examples include USB flash drives, removable hard drives, magnetic disks, or optical disks. In some jurisdictions, based on legislation and patent practice, computer-readable media cannot be electric carrier signals or telecommunication signals.

[0093] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A computing power allocation and scheduling method for AI satellite constellations, characterized in that: include: Based on the orbital characteristics of the satellite, a seat node is assigned to each satellite, and each seat node is configured with basic computing resources; Based on the orbital characteristics of the satellite and the basic computing power resources, a computing power flow potential field is constructed; Based on the satellite coverage parameters, the computing power requirements of the tasks in each coverage area are evaluated to generate a task demand gradient; Combining the computing power flow potential field and the task demand gradient, reallocating computing power resources to each agent node; The performance of the newly added satellites is evaluated and verified, so that the newly added satellites are gradually integrated into the constellation network.

2. The method according to claim 1, wherein Based on the satellite's orbital characteristics, a seat node is assigned to each satellite. Each seat node is configured with basic computing resources, including: Each satellite is considered as an agent node, and each agent node is marked with a unique identifier, orbital altitude parameters, computing power performance data and real-time status information; Based on the orbital altitude parameters and the computing power performance data, the orbital characteristic differences between different satellites are obtained, and the basic seat weight of each satellite is determined; Obtain reliability measurement data of each satellite through the satellite's fault diagnosis system and historical operation data, quantify the availability of the agent node based on the reliability measurement data, and construct a reliability scoring matrix; Determine the spatiotemporal reachability index of each of the agent nodes based on the reliability scoring matrix and in combination with satellite orbit parameters and coverage parameters; Based on the basic seat weight, reliability scoring matrix and spatiotemporal accessibility index, the priority weight of the computing power resource allocation of each seat node is determined, and differentiated basic computing power resources are configured for different satellites.

3. The method according to claim 2, wherein The construction of a computing power flow potential field based on the satellite's orbital characteristics and the basic computing power resources includes: Based on the differences in the orbital heights of each satellite, a spatial hierarchical distribution matrix from low orbit to high orbit is constructed; Determine the remaining computing capacity of each satellite based on its basic computing resources, computing performance data, and real-time status information; For each satellite, the energy efficiency weight factor is calculated based on power consumption and remaining computing capacity. Combined with the reliability score matrix, the comprehensive potential energy value is obtained through weighted calculation method. Combining the spatial hierarchical distribution matrix and the comprehensive potential energy value, a gradient field distribution model is constructed, and the gradient strength and direction vector between adjacent satellites are calculated through the potential energy difference to form a computing power flow potential field.

4. The method according to claim 3, wherein Determining the remaining computing capacity of each satellite based on the basic computing resources, computing performance data, and real-time status information of each satellite includes: Obtain the corresponding basic computing resources, computing performance data and real-time status information from the seat nodes of each satellite; Based on the real-time status information, obtaining a load distribution view; Based on the load distribution view, combined with the basic computing resources and the computing performance data, the resource utilization ratio of each seat node is calculated to obtain the remaining computing capacity.

5. The method according to claim 1, wherein The computing power requirement evaluation of the tasks within each coverage area is performed based on the satellite coverage parameters to generate a task requirement gradient, including: Determine the coverage area of each satellite based on the satellite's orbital parameters and the Earth's rotation parameters; In each of the coverage areas, calculating the regional computing power demand density based on the nature of the task and computing requirements; Based on the regional computing power demand density of each coverage area, the task demand gradient is determined, where the task demand gradient is used to measure the spatial rate of change of computing power demand.

6. The method according to claim 5, wherein Within each of the coverage areas, there is also: Acquire mission request data from a ground station, perform preprocessing operations on the mission request data, and obtain mission requirement characteristic values; Based on the task requirement characteristic values, combined with task priorities and resource requirements, a multidimensional task requirement vector is constructed to determine the task distribution characteristics of the tasks in the multidimensional space; According to the task distribution characteristics, the tasks are accurately located in the vector space to obtain dimensional coordinates; Based on the dimensional coordinates, combined with task priority and latency requirements, the priority order of task execution is determined.

7. The method according to claim 1, wherein The reallocation of computing power resources to each agent node in combination with the computing power flow potential field and the task demand gradient includes: Calculate the computing power allocation tendency coefficient for the areas covered by two adjacent agent nodes based on the task demand gradient and computing power flow potential field strength; For each agent node, the total tendency coefficient of computing power allocation to all adjacent coverage areas is calculated based on the computing power allocation tendency coefficient; Based on the total tendency coefficient, the computing power distribution of each seat node is determined.

8. The method according to claim 1, wherein The performing of performance evaluation and performance verification on the newly added satellite to gradually integrate the newly added satellite into the constellation network includes: Obtaining hardware configuration parameters and orbital position information of the newly added satellite, and performing a quantitative performance evaluation on the newly added satellite; Assigning an initial verification task to the newly added satellite to obtain performance baseline data; Using the performance baseline data as a reference standard, gradually increasing the number of tasks and computational complexity for the newly added satellite to determine the load adaptability of the newly added satellite; If the load adaptability meets the expected standards, the newly added satellite is added to the constellation network and basic computing resources are allocated based on priority weights.

9. A computing power allocation and scheduling system for AI satellite constellations, characterized by: include: The first processing module is used to allocate a seat node to each satellite based on the orbital characteristics of the satellite, and each seat node is configured with basic computing resources; The second processing module is used to construct a computing power flow potential field based on the orbital characteristics of the satellite and the basic computing power resources; The third processing module is used to evaluate the computing power requirements of tasks in each coverage area based on the satellite coverage parameters and generate a task demand gradient; A fourth processing module is configured to reallocate computing power resources to each agent node based on the computing power flow potential field and the task demand gradient; The fifth processing module is used to perform performance evaluation and performance verification on the newly added satellite, so that the newly added satellite can be gradually integrated into the constellation network.

Images