A method and device for predicting edge computing offloading based on multi-constellation collaboration

Abstract

This application discloses a multi-constellation collaborative edge computing offloading prediction method and apparatus, comprising: determining the task computation amount of each satellite in the satellite constellation for the same region; constructing a first graph structure with each satellite as a first node and the first task offloading relationship of each satellite as a first edge; iteratively updating the first graph structure using a graph neural network model and according to a pre-set computation rule to generate a second graph structure; constructing a first target feature vector and a second target feature vector in the second graph structure corresponding to the virtual task offloading relationship, determining the second task offloading relationship between each satellite constellation in the second graph structure based on the first target feature vector and the second target feature vector; determining a third target feature vector corresponding to the associated node in the second graph structure that is associated with the second task offloading relationship, and inputting the third target feature vector into a pre-trained first prediction model to predict the task offloading amount between each satellite constellation.

Description

Technical Field

[0001] This application relates to the field of satellite computing technology, and in particular to an edge computing offloading prediction method and apparatus based on multi-constellation cooperation. Background Technology

[0002] In a multi-satellite constellation, when any one of the satellites is overloaded with computing power, it can offload some of its tasks to other satellites with available computing resources, thereby ensuring computing efficiency and improving the utilization of computing resources. This operation is called inter-satellite computing task offloading and is one of the core advantages of modern satellite network design.

[0003] Furthermore, to achieve global coverage, optimize resources, and enhance communication reliability, multiple satellite constellations can be used for collaborative communication. For example, a single satellite constellation may have limited computing resources, preventing satellites from completing tasks independently. Therefore, by utilizing the idle computing resources of satellites in other constellations, large-scale tasks can be completed collaboratively.

[0004] Figure 1 This is a schematic diagram illustrating the offloading of existing multi-constellation computing tasks. (Reference) Figure 1 As shown, the system includes satellite constellation 1 and satellite constellation 2. Satellite constellation 1 includes multiple satellites 1-3, and satellite constellation 2 includes multiple satellites 4-6. Satellites 1 and 2 are connected via inter-satellite links, and satellite 1 is also connected to satellite 3 via an inter-satellite link. Satellites 4 and 5 are connected via inter-satellite links, and satellite 4 is also connected to satellite 6 via an inter-satellite link. Furthermore, between constellations 1 and 2, satellites 1 and 5 communicate with each other, and satellites 1 and 4 communicate with each other. Therefore, if satellite 1 has a large number of computational tasks, and even if satellite 1 offloads some computational tasks to satellites 2 and 3, it still cannot complete the computational tasks (i.e., the computational tasks cannot be completed using only the satellites in constellation 1), satellite 1 can offload some computational tasks to satellites 5 and / or 4.

[0005] On the other hand, satellites 5 and / or 6 in satellite constellation 2, in addition to their own computational tasks, also need to compute some tasks offloaded from satellite 1 in satellite constellation 1. However, due to the limited computational resources of satellites 5 and / or 4 in satellite constellation 2, and the potential impact of other objective factors, predicting the offloaded task relationships and workload between satellite constellations 1 and 2 to ensure the successful completion of computational tasks is a pressing issue that needs to be addressed.

[0006] There is currently no effective solution to the technical problem in the existing technology of how to predict the task offloading relationship and task offloading amount between multiple constellations in order to ensure the smooth completion of computing tasks and improve the utilization rate of computing resources, due to the limited computing resources of a single constellation for processing tasks. Summary of the Invention

[0007] The embodiments of this disclosure provide an edge computing offloading prediction method and apparatus based on multi-constellation collaboration, which at least solves the technical problem in the prior art of how to predict the task offloading relationship and task offloading amount between multiple constellations due to the limited computing resources of a single constellation processing task, so as to ensure the smooth completion of computing tasks and improve the utilization rate of computing resources.

[0008] According to one aspect of the present disclosure, a multi-constellation collaborative edge computing offloading prediction method is provided, comprising: determining multiple satellite constellations and the task computing volume of each satellite in the constellation for the same region; constructing a first graph structure with each satellite in the constellation as a first node and a first task offloading relationship of each satellite as a first edge, wherein the first task offloading relationship includes the actual task offloading relationship between two adjacent satellites within the constellation and the virtual task offloading relationship between two adjacent satellites corresponding to different satellite constellations, the virtual task offloading relationship including offloading task relationship and receiving offloading task relationship; iteratively updating the first graph structure using a pre-trained graph neural network model and according to a pre-set calculation rule to generate a second graph structure; constructing a first target feature vector and a second target feature vector in the second graph structure corresponding to the virtual task offloading relationship, and determining a second task offloading relationship between each satellite constellation in the second graph structure based on the first target feature vector and the second target feature vector, wherein the second task offloading relationship represents an offloading task relationship or a receiving offloading task relationship; and determining a third target feature vector corresponding to the associated node of the second task offloading relationship in the second graph structure, and inputting the third target feature vector into a pre-trained first prediction model to predict the task offloading volume between each satellite constellation.

[0009] According to another aspect of the present disclosure, a storage medium is also provided, the storage medium including a stored program, wherein, when the program is executed, a processor performs any of the methods described above.

[0010] According to another aspect of the present disclosure, a multi-constellation collaborative edge computing offloading prediction apparatus is also provided, comprising: a task computation volume determination module, configured to determine the task computation volume of multiple satellite constellations and each satellite in a satellite constellation for the same region; a first graph structure construction module, configured to construct a first graph structure with each satellite of a satellite constellation as a first node and the first task offloading relationship of each satellite as a first edge, wherein the first task offloading relationship includes the real task offloading relationship between two adjacent satellites within a satellite constellation and the virtual task offloading relationship between two adjacent satellites corresponding to different satellite constellations, the virtual task offloading relationship including offloading task relationship and receiving offloading task relationship; and an iterative update module, configured to utilize a pre-trained graph neural network. The network model iteratively updates the first graph structure according to pre-set calculation rules to generate the second graph structure; the task unloading relationship determination module is used to construct the first target feature vector and the second target feature vector corresponding to the virtual task unloading relationship in the second graph structure, and determine the second task unloading relationship between each satellite constellation in the second graph structure based on the first target feature vector and the second target feature vector, wherein the second task unloading relationship represents the unloading task relationship or the receiving unloading task relationship; and the task unloading amount prediction module is used to determine the third target feature vector corresponding to the associated node of the second task unloading relationship in the second graph structure, and input the third target feature vector into the pre-trained first prediction model to predict the task unloading amount between each satellite constellation.

[0011] According to another aspect of the present disclosure, a multi-constellation collaborative edge computing offloading prediction apparatus is also provided, comprising: a processor; and a memory connected to the processor, configured to provide the processor with instructions for processing the following steps: determining multiple satellite constellations and the task computation volume of each satellite in the satellite constellation for the same region; constructing a first graph structure with each satellite of the satellite constellation as a first node and a first task offloading relationship of each satellite as a first edge, wherein the first task offloading relationship includes the actual task offloading relationship between two adjacent satellites within the satellite constellation and the virtual task offloading relationship between two adjacent satellites corresponding to different satellite constellations, and the virtual task offloading relationship includes an offloading task relationship and a receiving offloading task relationship. The process involves: establishing a task unloading relationship; iteratively updating the first graph structure using a pre-trained graph neural network model according to pre-defined calculation rules to generate a second graph structure; constructing a first target feature vector and a second target feature vector in the second graph structure corresponding to the virtual task unloading relationship; determining the second task unloading relationship between each satellite constellation in the second graph structure based on the first and second target feature vectors, where the second task unloading relationship represents an unloading task relationship or a receiving unloading task relationship; and determining the third target feature vector corresponding to the associated node in the second graph structure that corresponds to the second task unloading relationship, and inputting the third target feature vector into a pre-trained first prediction model to predict the task unloading amount between each satellite constellation.

[0012] This application provides a multi-constellation collaborative edge computing offloading prediction method. First, the processor determines multiple satellite constellations and the computational load of each satellite within a constellation for the same region. Then, the processor constructs a first graph structure using each satellite in the constellation as a first node and the first task offloading relationship between the satellites as a first edge. Further, the processor iteratively updates the first graph structure using a pre-trained graph neural network model according to pre-defined computation rules, generating a second graph structure. The processor then constructs a first target feature vector and a second target feature vector in the second graph structure corresponding to the virtual task offloading relationship, and determines the second task offloading relationship between the satellite constellations in the second graph structure based on these vectors. Finally, the processor determines a third target feature vector corresponding to the associated node in the second graph structure that corresponds to the second task offloading relationship, and inputs this third target feature vector into a pre-trained first prediction model to predict the task offloading load between the satellite constellations.

[0013] As described above, this application first constructs a first graph structure corresponding to multiple satellite constellations, using each satellite in a satellite constellation as the first node and the first task offloading relationship of each satellite as the first edge. The first task offloading relationship includes the virtual task offloading relationship between two adjacent satellites in different constellations. That is, even when no actual offloading tasks are sent between different satellite constellations, it can be assumed that there are offloading task relationships and receiving offloading task relationships between different satellite constellations. Therefore, by iteratively updating the first graph structure and generating a second graph structure, the second task offloading relationship between each satellite constellation can be determined to identify whether each satellite constellation needs to offload tasks to other satellite constellations or receive tasks offloaded by other satellite constellations.

[0014] Furthermore, by determining the second task offloading relationship between each satellite constellation, this application can further predict the task offloading amount between each satellite constellation. Therefore, when a single satellite constellation cannot independently complete a certain computing task, some tasks can be offloaded to other satellite constellations through the second task offloading relationship with them; or, when a satellite constellation has ample idle computing resources and other satellite constellations have heavy computing tasks, some tasks can be accepted to ensure the successful completion of the task.

[0015] This ensures the successful completion of computational tasks and improves the utilization rate of computing resources across satellite constellations. Furthermore, it solves the technical problem in existing technologies where, due to the limited computing resources available for a single constellation, it is crucial to predict the task offloading relationship and amount between multiple constellations to guarantee successful completion of computational tasks and improve the utilization rate of computing resources. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of this disclosure and form part of this application, illustrate exemplary embodiments of the present disclosure and are used to explain the disclosure, but do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the offloading of existing multi-constellation computing tasks; Figure 2 This is a schematic diagram of the multi-constellation collaborative edge computing offloading prediction system according to Embodiment 1 of this application; Figure 3A This is a schematic diagram of the hardware architecture of each satellite according to Embodiment 1 of this application; Figure 3B This is a schematic diagram of the hardware architecture of the ground system according to Embodiment 1 of this application; Figure 4 This is a modular schematic diagram of the multi-constellation collaborative edge computing offloading prediction system according to Embodiment 1 of this application; Figure 5 This is a flowchart of the multi-constellation collaborative edge computing offloading prediction method according to Embodiment 1 of this application; Figure 6 This is a schematic diagram of the structure of the first figure according to Embodiment 1 of this application; Figure 7 This is a model architecture diagram of the graph neural network model according to Embodiment 1 of this application; Figure 8 This is a schematic diagram of the first prediction model according to Embodiment 1 of this application; Figure 9 This is a schematic diagram of the multi-constellation collaborative edge computing offloading prediction device according to Embodiment 2 of this application; Figure 10 This is a schematic diagram of the multi-constellation collaborative edge computing offloading prediction device according to Embodiment 3 of this application. Detailed Implementation

[0017] To enable those skilled in the art to better understand the technical solutions of this disclosure, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this disclosure.

[0018] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0019] Example 1 According to this embodiment, a method embodiment for predicting edge computing offloading through multi-constellation collaboration is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0020] Figure 2 This is a schematic diagram of a multi-constellation collaborative edge computing offloading prediction system according to Embodiment 1 of this application. (Reference) Figure 2 As shown, the edge computing offloading prediction system comprises satellite constellations 1 to m, with each constellation containing n satellites. Satellites within constellations 1 to m can establish communication links to offload or receive offloading tasks. Adjacent satellites within constellations 1 to m can also establish communication links to offload or receive offloading tasks.

[0021] In addition, each satellite in the satellite constellation 1 to m can establish a communication connection with the ground system 20, thereby ensuring that two adjacent satellites in the satellite constellation 1 to m can send prediction requests to the ground system 20, and that the ground system 20 can send the determined second task unloading relationship and task unloading amount to two adjacent satellites in the satellite constellation 1 to m.

[0022] Figure 3A Further shown Figure 1 A schematic diagram of the hardware architecture of each of the 10 satellites. (Reference) Figure 3A As shown, each satellite 10 includes an integrated electronic system, which includes a processor, a memory, a bus management module, and a communication interface. The memory is connected to the processor, allowing the processor to access the memory, read program instructions stored in the memory, read data from the memory, or write data to the memory. The bus management module is connected to the processor and also to a bus such as a CAN bus. Thus, the processor can communicate with onboard peripherals connected to the bus through the bus managed by the bus management module. Furthermore, the processor also communicates with devices such as cameras, star sensors, telemetry and command transponders, and data transmission equipment via the communication interface. Those skilled in the art will understand that… Figure 3A The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, a satellite system may also include... Figure 3A The more or fewer components shown, or having the same Figure 3A The different configurations shown.

[0023] Figure 3B Further shown Figure 1 A schematic diagram of the hardware architecture of the ground system 20. (Reference) Figure 3BAs shown, the ground system 20 may include one or more processors (processors may include, but are not limited to, microprocessors such as MCUs or programmable logic devices such as FPGAs), a memory for storing data, a transmission device for communication functions, and an input / output interface. The memory, transmission device, and input / output interface are connected to the processor via a bus. In addition, it may also include a display, keyboard, and cursor control device connected to the input / output interface. Those skilled in the art will understand that... Figure 3B The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, the ground system may also include... Figure 3B The more or fewer components shown, or having the same Figure 3B The different configurations shown.

[0024] It should be noted that, Figure 3A and Figure 3B One or more processors and / or other data processing circuits shown herein may generally be referred to as "data processing circuitry". This data processing circuitry may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuitry may be a single, independent processing module, or may be integrated, in whole or in part, into any other element in a computing device. As involved in embodiments of this disclosure, the data processing circuitry serves as processor control (e.g., selection of a variable resistor termination path connected to an interface).

[0025] Figure 3A and Figure 3B The memory shown can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the multi-constellation collaborative edge computing offloading prediction method in this embodiment of the present disclosure. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the multi-constellation collaborative edge computing offloading prediction method of the above-mentioned application. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory.

[0026] It should be noted here that, in some optional embodiments, the above... Figure 3A and Figure 3B The device shown may include hardware elements (including circuitry), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware and software elements. It should be noted that... Figure 3A and Figure 3B This is only one instance of a specific particular instance, and is intended to illustrate the types of components that may exist in the aforementioned devices.

[0027] Figure 4 This is a modular schematic diagram of a multi-constellation collaborative edge computing offloading prediction system according to an embodiment of this application. (Reference) Figure 4 As shown, the edge computing offloading system includes a data acquisition module, a task volume prediction module, a graph structure construction module, an iterative update module, an offloading relationship determination module, an offloading volume prediction module, and a storage module.

[0028] The system comprises the following modules: a data acquisition module and a storage module, used to acquire the computational workload of each satellite in multiple satellite constellations for the same region during various historical time periods. A task workload prediction module, also connected to the data acquisition module, calculates the computational workload of each satellite for the same region during the current time period based on the computational workload of each satellite during various historical time periods. A graph structure construction module, connected to both the task workload prediction module and the storage module, retrieves data corresponding to the node attributes of the first node and the edge attributes of the first edge from the storage module, and constructs a first graph structure. An iterative update module, connected to the graph structure construction module, iteratively updates the first graph structure using a pre-trained graph neural network model, and generates a corresponding second graph structure. An unloading relationship determination module, connected to the iterative update module, determines the second unloading relationship between satellite constellations based on the first and second target feature vectors corresponding to the second graph structure. An unloading amount prediction module, connected to the unloading relationship determination module, predicts the task unloading amount between satellite constellations based on the third target feature vector corresponding to the second task unloading relationship and using a first prediction model.

[0029] Under the aforementioned operating environment, according to the first aspect of this embodiment, a multi-constellation collaborative edge computing offloading prediction method is provided, which consists of... Figure 4 The multi-constellation collaborative edge computing offloading prediction system shown is implemented. Figure 5 A flowchart illustrating the method is shown below. (Refer to...) Figure 5 As shown, the method includes: S502: Determine the computational workload of multiple satellite constellations and the mission workload of each satellite in the constellation for the same region; S504: Take each satellite of the satellite constellation as the first node and the first task unloading relationship of each satellite as the first edge to construct the first graph structure. The first task unloading relationship includes the real task unloading relationship between two adjacent satellites in the satellite constellation and the virtual task unloading relationship between two adjacent satellites corresponding to different satellite constellations. The virtual task unloading relationship includes the unloading task relationship and the receiving unloading task relationship. S506: Use a pre-trained graph neural network model and follow pre-defined calculation rules to iteratively update the first graph structure to generate the second graph structure; S508: Construct the first target feature vector and the second target feature vector in the second graph structure corresponding to the virtual task unloading relationship, and determine the second task unloading relationship between each satellite constellation in the second graph structure based on the first target feature vector and the second target feature vector, wherein the second task unloading relationship represents the unloading task relationship or the receiving unloading task relationship; S510: Determine the third target feature vector corresponding to the associated node in the second graph structure that is related to the second task unloading relationship, and input the third target feature vector into the pre-trained first prediction model to predict the task unloading amount between each satellite constellation.

[0030] Specifically, first, in response to a mission offload prediction request sent by any satellite in the satellite constellation, ground system 20 determines the satellite constellation corresponding to the satellite sending the prediction request. Then, based on the ephemeris information of each satellite, ground system 20 determines several other satellite constellations capable of establishing communication links with the determined satellite constellation. For example, when determining the satellite sending the prediction request... In this case, it can be determined that there is a connection with the satellite. Corresponding satellite constellation Furthermore, the ground system 20 can also determine its relationship with the satellite constellation based on the ephemeris information of multiple satellites. Satellite constellation establishing communication connections .

[0031] Then, with multiple satellite constellations identified in the ground system 20, the data acquisition module retrieves historical time periods from the storage module, showing the computational workload of each satellite in different constellations for the same area. The computational workload indicates the total number of tasks calculated when a satellite covers that area. For example, the data acquisition module retrieves historical time periods from the storage module. Internal satellite constellation Satellites within Regarding the computational load of the task in region A ;satellite Regarding the computational load of the task in region A ;...;satellite Regarding the computational load of the task in region A .

[0032] The data acquisition module retrieves historical time periods from the storage module. Internal satellite constellation Satellites within Regarding the computational load of the task in region A ;satellite Regarding the computational load of the task in region A ;...;satellite Regarding the computational load of the task in region A .

[0033] For example, the data acquisition module retrieves historical time periods from the storage module. Internal satellite constellation Satellites within Regarding the computational load of the task in region A ;satellite Regarding the computational load of the task in region A ;...;satellite Regarding the computational load of the task in region A .

[0034] The data acquisition module retrieves historical time periods from the storage module. Internal satellite constellation Satellites within Regarding the computational load of the task in region A ;satellite Regarding the computational load of the task in region A ;...;satellite Regarding the computational load of the task in region A .

[0035] And so on.

[0036] For example, the data acquisition module retrieves historical time periods from the storage module. Internal satellite constellation Satellites within Regarding the computational load of the task in region A ;satellite Regarding the computational load of the task in region A ;...;satellite Regarding the computational load of the task in region A .

[0037] The data acquisition module retrieves historical time periods from the storage module. Internal satellite constellation Satellites within Regarding the computational load of the task in region A ;satellite Regarding the computational load of the task in region A ;...;satellite Regarding the computational load of the task in region A .

[0038] The data acquisition module then sends the acquired data to the workload prediction module, which in turn predicts the workload based on the historical data of each satellite in multiple satellite constellations. ~ The computational workload of the task is calculated, and a task computation vector corresponding to each satellite is constructed. This vector is related to the satellite constellation. Each satellite in ~ Corresponding task computation vector ~ , , ,..., And among them, satellite constellations... Each satellite in ~ Corresponding task computation vector ~ , , ,..., .

[0039] Furthermore, the task calculation module inputs the task calculation vectors corresponding to each satellite in multiple satellite constellations into the pre-trained second prediction model, thereby outputting the task calculation amount of each satellite in multiple satellite constellations for region A in the current time period (S502).

[0040] Then, the task load prediction module sends the task calculation load for the same region for each satellite in the determined multiple satellite constellations to the graph structure construction module. The graph structure construction module constructs a first graph structure with each satellite in the constellation as the first node and the first task offloading relationship of each satellite as the first edge (S504). Since there are no offloading tasks between satellite constellations before the task offloading load of each constellation is determined, when constructing the first graph structure, it is necessary to pre-set that there are both offloading task relationships and receiving offloading task relationships between adjacent satellites in each constellation. These pre-set offloading task relationships and receiving offloading task relationships between adjacent satellites in each constellation are virtual task offloading relationships.

[0041] Furthermore, since mission unloading has occurred among the satellites in the constellation during various historical periods, the actual mission unloading relationships among the satellites in the constellation can be determined based on the actual mission unloading situation of each satellite in the constellation during each historical period.

[0042] Therefore, given the actual mission offloading relationships between satellites within a satellite constellation and the virtual mission offloading relationships between adjacent satellites within the constellation, the graph structure construction module can construct the first graph structure. This will be described in detail later, and therefore will not be repeated here.

[0043] Furthermore, when constructing a first graph structure, the graph structure construction module sends the first graph structure to the iterative update module. The iterative update module then inputs the first graph structure into a pre-trained graph neural network model, uses the graph neural network model and follows pre-set calculation rules to iteratively update the first graph structure, generating a second graph structure (S506). In this embodiment, the graph neural network model includes multiple first MLP models for iteratively updating the edge attributes of the first edge of the first graph structure and multiple second MLP models for iteratively updating the node attributes of the first node of the first graph structure.

[0044] Thus, the graph neural network model can utilize multiple first MLP models and construct edge attribute vectors according to the first calculation rules corresponding to the first edge, and use multiple first MLP models to iteratively update the edge attribute vectors of the first edge, thereby generating edge features corresponding to the second graph structure.

[0045] Similarly, the graph neural network model can utilize multiple second MLP models and construct node attribute vectors according to the second computation rules corresponding to the first node. Iterative updates of the node attribute vectors of the first node are then performed using multiple second MLP models to generate node features corresponding to the second graph structure. The above will be described in detail later, and therefore will not be repeated here.

[0046] The iterative update module then sends the second graph structure to the unloading relationship determination module. The unloading relationship determination module determines the edge features of the second edge corresponding to the unloading task relationship in the virtual task unloading relationship, as well as the node features of the two second nodes corresponding to the second edge. Based on the corresponding edge features and node features, it constructs the first target feature vector corresponding to the unloading task relationship.

[0047] At the same time, the unloading relationship determination module also determines the edge features of the second side corresponding to the receiving unloading task relationship in the virtual task unloading relationship, as well as the node features of the two second nodes corresponding to the second side, and constructs the second target feature vector corresponding to the receiving unloading task based on the corresponding edge features and node features.

[0048] Finally, the unloading relationship determination module calculates the difference vector between the first target feature vector and the second target feature vector, and determines the second task unloading relationship based on the difference vector. The above will be described in detail later, so it will not be repeated here.

[0049] Finally, the unloading relationship determination module sends the determined second task unloading relationship to the unloading quantity prediction module. The unloading quantity prediction module determines the third target feature vector corresponding to the associated node in the second graph structure that is related to the second task unloading relationship, and inputs the third target feature vector into the pre-trained first prediction model to predict the task unloading quantity between each satellite constellation (S510). Specifically, firstly, the task quantity prediction module determines the third target feature vector corresponding to the associated node in the second graph structure. Then, the task quantity prediction module inputs the third target feature vector into the first prediction model and determines the task unloading quantity between each satellite constellation. The above will be described in detail later, so it will not be repeated here.

[0050] As described in the background section, from another perspective, satellites 5 and / or 4 in constellation 2, in addition to their own computational tasks, also need to compute some tasks offloaded by satellite 1 in constellation 1. However, due to the limited computational resources of satellites 5 and / or 4 in constellation 2, and the potential impact of other objective factors, predicting the offloaded task relationship and workload between constellations 1 and 2 to ensure the successful completion of computational tasks is a pressing problem that needs to be solved.

[0051] In view of this, this application provides a multi-constellation collaborative edge computing offloading prediction method. As described above, this application first constructs a first graph structure corresponding to multiple satellite constellations, using each satellite in a satellite constellation as the first node and the first task offloading relationship of each satellite as the first edge. The first task offloading relationship includes the virtual task offloading relationship between two adjacent satellites in different constellations. That is, even when no actual offloading tasks are sent between different satellite constellations, it can be assumed that there are offloading task relationships and receiving offloading task relationships between different satellite constellations. Therefore, by iteratively updating the first graph structure and generating a second graph structure, the second task offloading relationship between each satellite constellation can be determined to determine whether each satellite constellation needs to offload tasks to other satellite constellations or receive tasks offloaded by other satellite constellations.

[0052] Furthermore, by determining the second task offloading relationship between each satellite constellation, this application can further predict the task offloading amount between each satellite constellation. Therefore, when a single satellite constellation cannot independently complete a certain computing task, some tasks can be offloaded to other satellite constellations through the second task offloading relationship with them; or, when a satellite constellation has ample idle computing resources and other satellite constellations have heavy computing tasks, some tasks can be accepted to ensure the successful completion of the task.

[0053] This ensures the successful completion of computational tasks and improves the utilization rate of computing resources across satellite constellations. Furthermore, it solves the technical problem in existing technologies where, due to the limited computing resources available for a single constellation, it is crucial to predict the task offloading relationship and amount between multiple constellations to guarantee successful completion of computational tasks and improve the utilization rate of computing resources.

[0054] Optionally, the operation of determining the task computation volume of multiple satellite constellations and each satellite in the constellation for the same region includes: determining the task computation volume of each satellite in the multiple satellite constellations for the region under historical time periods; constructing a task computation vector corresponding to each satellite based on the task computation volume of each satellite under each historical time period; and inputting each task computation vector into a pre-trained second prediction model and outputting the task computation volume of each satellite for the region under the current time period.

[0055] Specifically, the data acquisition module retrieves historical time periods from the storage module. Internal satellite constellation Satellites within Regarding the computational load of the task in region A ;satellite Regarding the computational load of the task in region A ;...;satellite Regarding the computational load of the task in region A .

[0056] The data acquisition module retrieves historical time periods from the storage module. Internal satellite constellation Satellites within Regarding the computational load of the task in region A ;satellite Regarding the computational load of the task in region A ;...;satellite Regarding the computational load of the task in region A .

[0057] For example, the data acquisition module retrieves historical time periods from the storage module. Internal satellite constellation Satellites within Regarding the computational load of the task in region A ;satellite Regarding the computational load of the task in region A ;...;satellite Regarding the computational load of the task in region A .

[0058] The data acquisition module retrieves historical time periods from the storage module. Internal satellite constellation Satellites within Regarding the computational load of the task in region A ;satellite Regarding the computational load of the task in region A ;...;satellite Regarding the computational load of the task in region A .

[0059] And so on.

[0060] For example, the data acquisition module retrieves historical time periods from the storage module. Internal satellite constellation Satellites within Regarding the computational load of the task in region A ;satellite Regarding the computational load of the task in region A ;...;satellite Regarding the computational load of the task in region A .

[0061] The data acquisition module retrieves historical time periods from the storage module. Internal satellite constellation Satellites within Regarding the computational load of the task in region A ;satellite Regarding the computational load of the task in region A ;...;satellite Regarding the computational load of the task in region A .

[0062] The data acquisition module then sends the acquired data to the workload prediction module, which in turn predicts the workload based on the historical data of each satellite in multiple satellite constellations. ~ The computational workload of the task is calculated, and a task computation vector corresponding to each satellite is constructed. This vector is related to the satellite constellation. Each satellite in ~ Corresponding task computation vector ~ , , ,..., And among them, satellite constellations... Each satellite in ~ Corresponding task computation vector ~ , , ,..., .

[0063] Furthermore, the task load prediction module inputs the task computation vectors corresponding to each satellite within multiple satellite constellations into a pre-trained second prediction model. This second prediction model is a neural network model, comprising an input layer, a fully connected layer, and an output layer. Thus, the second prediction model can output the task computation load for the same region for each satellite within multiple satellite constellations during the current time period. For example, with satellite constellations... Each satellite in ~ Corresponding task computation vector ~ , , ,..., And among them, satellite constellations... Each satellite in ~ Corresponding task computation vector ~ , , ,..., .

[0064] Optionally, the operation of constructing a first graph structure with each satellite of the satellite constellation as the first node and the first task offloading relationship of each satellite as the first edge includes: determining the node attributes corresponding to the first node based on the task computation amount of each satellite; determining the first edge attributes corresponding to the first edge based on the task offloading weight, distance weight, and attenuation coefficient weight corresponding to the real task offloading relationship; determining the second edge attributes corresponding to the first edge based on the distance weight and attenuation coefficient weight corresponding to the virtual task offloading relationship; and constructing the first graph structure based on the first node, the first edge attributes, and the second edge attributes.

[0065] Specifically, Figure 6 This is a schematic diagram of the structure described in the first figure according to an embodiment of this application. (Reference) Figure 6 As shown, the first node of the first graph structure ~ as well as ~ Including satellite constellations Each satellite in ~ and satellite constellations Each satellite in ~ And with the first node. The corresponding node attributes are , with the first node The corresponding node attributes are ... and the first node The corresponding node attributes are With the first node The corresponding node attributes are , with the first node The corresponding node attributes are ... and the first node The corresponding node attributes are .

[0066] Furthermore, since there are real mission unloading relationships between the satellites in the constellation at different historical periods, the attributes of the first side corresponding to the first side can be determined based on the mission unloading weight, distance weight, and attenuation coefficient weight between adjacent satellites. The mission unloading weight indicates the weighted relationship between the amount of unloaded missions received by the satellites in the constellation and the total amount of missions unloaded by the constellation over multiple historical periods.

[0067] For example, in historical periods ~ Internal satellite constellation satellites in To satellite The amount of work to be uninstalled is Satellite constellation satellites in To satellite The amount of work to be uninstalled is Satellite constellation satellites in To satellite The amount of work to be uninstalled is Satellite constellation satellites in To satellite The amount of work to be uninstalled is And so on. Satellite constellations. satellites in To satellite The amount of work to be uninstalled is ,constellation satellites in To satellite The amount of work to be uninstalled is Among them, in historical periods ~ Inside, multiple satellites ~ The total number of tasks uninstalled is .

[0068] Thus in historical periods ~ Internal satellite constellation satellites in To satellite The task uninstallation weight of the uninstallation task is Satellite constellation satellites in To satellite The task uninstallation weight of the uninstallation task is Satellite constellation satellites in To satellite The task uninstallation weight of the uninstallation task is Satellite constellation satellites in To satellite The task uninstallation weight of the uninstallation task is And so on. Satellite constellations. satellites in To satellite The task uninstallation weight of the uninstallation task is Satellite constellation satellites in To satellite The task uninstallation weight of the uninstallation task is .

[0069] Among them, determining the satellite constellation The method of offloading mission weights among the various satellites is the same as described above, so it will not be repeated here.

[0070] Furthermore, distance weights are used to indicate the weighted relationship between the distance between two adjacent satellites, using the two closest adjacent satellites as a reference, and the aforementioned reference distance. The first side attributes include the satellite constellation. satellites in With satellite Distance weights between ,satellite With satellite Distance weights between ,...,satellite With satellite Distance weights between Among these, determining the satellite constellation. The method of weighting the distances between the satellites is the same as described above, so it will not be repeated here.

[0071] Furthermore, the attributes of the first side also include satellite constellations. satellites in With satellite The weight of the attenuation coefficient between ,satellite With satellite The weight of the attenuation coefficient between ,...,satellite With satellite The weight of the attenuation coefficient between The attenuation system weights between adjacent satellites are influenced by factors such as rainfall, cloud cover, and fog. Furthermore, this involves determining the satellite constellation. The method of weighting the attenuation coefficients among the various satellites is the same as described above, so it will not be repeated here.

[0072] Since there is no real mission unloading relationship between adjacent satellites in a satellite constellation at any historical time period, only a virtual mission unloading relationship, there is no mission unloading weight between adjacent satellites in a satellite constellation; only distance weight and attenuation coefficient weight exist. Therefore, the attributes of the second side corresponding to the first side can be determined based on the distance weight and attenuation coefficient weight.

[0073] Among them, the second side attribute of the first side includes the satellite constellation. satellite With satellite constellations satellite Distance weights between Satellite constellation satellite With satellite constellations satellite Distance weights between Satellite constellation satellite With satellite constellations satellite Distance weights between .

[0074] Furthermore, the second side attributes of the first side also include satellite constellations. satellite With satellite constellations satellite The weight of the attenuation coefficient between Satellite constellation satellite With satellite constellations satellite The weight of the attenuation coefficient between Satellite constellation satellite With satellite constellations satellite The weight of the attenuation coefficient between .

[0075] Thus, given the determination of each first node, the first edge attributes corresponding to each first node, and the second edge attributes corresponding to each first node, a first graph structure can be constructed to represent the task offloading relationship of multiple satellite constellations.

[0076] Optionally, the graph neural network model includes multiple first MLP models, and the operation of using the graph neural network model and iteratively updating the first graph structure according to a pre-defined calculation rule to generate the second graph structure includes: inputting the first graph structure into the graph neural network model, and constructing an edge attribute vector using a first calculation rule corresponding to the first edge, wherein the edge attribute vector includes the edge attribute of the first edge and the node attributes of the two first nodes adjacent to the first edge; and iteratively updating the edge attribute vector of the first edge using multiple first MLP models to generate edge features corresponding to the second graph structure.

[0077] Specifically, Figure 7 This is a model architecture diagram of the graph neural network model according to Embodiment 1 of this application. (Reference) Figure 7 As shown, the graph neural network model includes multiple first MLP models. These multiple first MLP models are used to iteratively update the edge attribute vector of the first edge in the first graph structure and generate edge features corresponding to the second graph structure.

[0078] Therefore, the iterative update module first constructs the edge attribute vector using the first calculation rule corresponding to the first edge. For example, the iterative update module constructs the edge attribute vector in the first graph structure according to the first calculation rule, based on the satellite constellation. First side Corresponding edge attribute vector .in, With satellite constellations First side Corresponding edge attribute vector .in, And so on. (Regarding satellite constellations) First side Corresponding edge attribute vector .in, .

[0079] For example, in the iterative update module constructing the first graph structure according to the first calculation rule, the satellite constellation... First side Corresponding edge attribute vector .in, With satellite constellations First side Corresponding edge attribute vector .in, And so on. (Regarding satellite constellations) First side Corresponding edge attribute vector .in, .

[0080] For example, in the satellite constellation in the first graph structure constructed by the iterative update module according to the first calculation rule... With satellite constellations The first side between two adjacent satellites Corresponding edge attribute vector .in, Satellite constellation With satellite constellations The first side between two adjacent satellites Corresponding edge attribute vector .in, Satellite constellation With satellite constellations The first side between two adjacent satellites Corresponding edge attribute vector .in, .

[0081] Then, the iterative update module inputs the constructed edge attribute vectors corresponding to each first edge into multiple first MLP models, thereby outputting the iteratively updated edge features corresponding to the second graph structure. This process continues. The iterative update module can use multiple first MLP models to iteratively update the edge attribute vectors of each first edge in the first graph structure and generate edge features corresponding to the second graph structure. The iterative update method is the same as described above, so it will not be repeated here.

[0082] Optionally, the graph neural network model includes multiple second MLP models, and the operation of using the graph neural network model to iteratively update the first graph structure according to a pre-defined calculation rule to generate the second graph structure includes: inputting the first graph structure into the graph neural network model, and constructing a node attribute vector using the second calculation rule corresponding to the first node, wherein the node attribute vector includes the node attribute of the first node, the edge attributes of the two adjacent first edges of the first node, and the node attributes of the first nodes corresponding to the two adjacent first edges respectively; and iteratively updating the node attribute vector of the first node using multiple second MLP models to generate node features corresponding to the second graph structure.

[0083] Specifically, refer to Figure 7 As shown, the graph neural network model includes multiple second MLP models. These second MLP models are used to iteratively update the node attribute vector of the first node in the first graph structure and generate node features corresponding to the second graph structure.

[0084] Therefore, firstly, the iterative update module constructs a node attribute vector using the second calculation rule corresponding to the first node. For example, the iterative update module constructs the node attribute vector corresponding to the first node in the first graph structure according to the second calculation rule. Corresponding node attribute vector .in, , , , For example, the iterative update module constructs the first node in the first graph structure according to the second calculation rule. Corresponding node attribute vector .in, , , , .

[0085] And so on.

[0086] For example, the iterative update module constructs the first node in the first graph structure according to the second calculation rule. Corresponding node attribute vector .in, , , , .

[0087] For example, the iterative update module constructs the first node in the first graph structure according to the second calculation rule. Corresponding node attribute vector .in, , , , .

[0088] Then, the iterative update module inputs the constructed node attribute vectors corresponding to each first node into multiple second MLP models, thereby outputting the iteratively updated node features corresponding to the second graph structure. This process continues. The iterative update module can use multiple second MLP models to iteratively update the node attribute vectors of each first node in the first graph structure and generate node features corresponding to the second graph structure. The iterative update method is the same as described above, so it will not be repeated here.

[0089] It is worth noting that the graph neural network model includes multiple first MLP models and multiple second MLP models. The first first MLP model and the first second MLP model can iteratively update the first graph structure to generate intermediate graph structures. The last first MLP model and the last second MLP model can iteratively update the intermediate graph structure to generate the second graph structure. Apart from the first first MLP model and the last first MLP model, and the first second MLP model and the last second second MLP model, the intermediate first MLP model and the intermediate second MLP model are used to iteratively update the previous intermediate graph structure and generate the next intermediate graph structure. Further details will not be elaborated here.

[0090] Optionally, the operation of constructing a first target feature vector corresponding to the virtual task unloading relationship in the second graph structure, and determining the second task unloading relationship between each satellite constellation in the second graph structure based on the first target feature vector, includes: determining the edge features of the second edge corresponding to the unloading task relationship in the virtual task unloading relationship and the node features of the two second nodes corresponding to the second edge, and constructing a first target feature vector corresponding to the unloading task relationship based on the edge features and node features; determining the edge features of the second edge corresponding to the receiving unloading task relationship in the virtual task unloading relationship and the node features of the two second nodes corresponding to the second edge, and constructing a second target feature vector corresponding to the receiving unloading task relationship based on the edge features and node features; and calculating the difference vector between the first target feature vector and the second target feature vector, and determining the second task unloading relationship based on the difference vector.

[0091] Specifically, refer to Figure 6 As can be seen, since the first task offloading relationship between two adjacent satellites in multiple satellite constellations is a virtual task offloading relationship, it is necessary to predict the real task offloading relationship (i.e., the second task offloading relationship) between two adjacent satellites in multiple satellite constellations.

[0092] Therefore, firstly, the task unloading relationship determination module determines the edge characteristics of the second side corresponding to the unloading task relationship in the virtual task unloading relationship, as well as the node characteristics of the two second nodes corresponding to the second side. For example, satellite constellations. satellites in With satellite constellations satellites in Communication links can be established between them, and satellites With satellite The unloading task relationship between them is based on satellite As a precursor node, with satellite The edge feature of the second edge corresponding to the successor node .

[0093] Furthermore, the two second nodes corresponding to the second side are related to the satellite. The corresponding second node and satellite The corresponding second node Thus, with the second node The corresponding node features are , with the second node The corresponding node features are .

[0094] Therefore, the task unloading relationship determination module can be based on edge features. Node features and node features The first target feature vector corresponding to the relationship between construction and unloading tasks .

[0095] Similarly, the task unloading relationship determination module can determine the edge features of the second side corresponding to the receiving unloading task relationship in the virtual task unloading relationship, as well as the node features of the two second nodes corresponding to the second side. For example, a satellite constellation. satellites in With satellite constellations satellites in Communication links can be established between them, and satellites With satellite The relationship between receiving and unloading tasks is based on satellites. As a precursor node, with satellite The edge feature of the second edge corresponding to the successor node .

[0096] Furthermore, the two second nodes corresponding to the second side are related to the satellite. The corresponding second node and satellite The corresponding second node Thus, with the second node The corresponding node features are , with the second node The corresponding node features are .

[0097] Therefore, the task unloading relationship determination module can be based on edge features. Node features and node features Constructing and receiving the second target feature vector corresponding to the relationship between the unloading task and the task. .

[0098] The task unloading relationship determination module then calculates the first target feature vector. With the second target feature vector The difference vector between them. That is, .

[0099] Then, the task unloading relationship determination module determines... The relationship between positive and negative values. If, If it is positive, it means that the satellite As a precursor node, with satellite This is the successor node. Conversely, it indicates that the satellite is the primary node. As a precursor node, with satellite It is the successor node.

[0100] It is worth noting that the mission offloading relationship between two adjacent satellites in each satellite constellation can also be determined based on the above method, so it will not be elaborated here.

[0101] Therefore, even if two adjacent satellites in a satellite constellation have not actually unloaded their missions, the actual mission unloading relationship between these two satellites in the next time period can be determined using the above method. This facilitates the ground system in allocating mission unloading among satellite constellations.

[0102] Optionally, the operation of inputting the third target feature vector into a pre-trained first prediction model to predict the task offloading amount between each satellite constellation includes: determining the third target feature vector corresponding to the associated node in the second graph structure, wherein the third target feature vector includes the node feature corresponding to the associated node, the edge features of the two adjacent second edges of the associated node, and the node features of the second nodes corresponding to the two adjacent second edges respectively; and inputting the third target feature vector into the first prediction model to determine the task offloading amount between each satellite constellation, wherein the first prediction model includes an input layer, a fully connected layer, an activation function layer, and an output layer.

[0103] Specifically, Figure 8 This is a schematic diagram of the first prediction model according to Embodiment 1 of this application. (Reference) Figure 8 As shown, the first prediction model includes an input layer, a fully connected layer, an activation function layer, and an output layer.

[0104] Furthermore, the task unloading prediction module first determines the third target feature vector corresponding to the associated nodes in the second graph structure. This third target feature vector includes the node features corresponding to the associated node, the edge features of the two adjacent second edges of the associated node, and the node features of the second nodes corresponding to the two adjacent second edges. For example, a satellite constellation... satellites in With satellite constellations satellites in If two satellites are adjacent, then the associated node is: (i.e., with associated nodes) Corresponding satellite To satellite (Unload the task). Therefore, the third target feature vector is... Among them, the aforementioned third target feature vector represents the associated node. Corresponding satellite To the satellite constellation respectively Satellites within and satellite constellations Satellites within Uninstall the task.

[0105] The mission unloading prediction module then inputs the third target feature vector into the first prediction model to determine the mission unloading amount between each satellite constellation.

[0106] Thus, according to the first aspect of this embodiment, the technical effect of ensuring the smooth completion of computing tasks and improving the utilization rate of computing resources of each satellite constellation is achieved.

[0107] In addition, refer to Figure 1 As shown, according to a second aspect of this embodiment, a storage medium is provided. The storage medium includes a stored program, wherein, when the program is executed, a processor performs any of the methods described above.

[0108] Thus, according to this embodiment, the technical effect of ensuring the smooth completion of computing tasks and improving the utilization rate of computing resources of each satellite constellation is achieved.

[0109] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0110] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0111] Example 2 Figure 9 An edge computing offloading prediction apparatus 900 for multi-constellation collaboration according to this embodiment is shown, which corresponds to the method described according to Embodiment 1. Reference Figure 9 As shown, the device 900 includes: a task computation volume determination module 910, used to determine the task computation volume of multiple satellite constellations and each satellite in the constellation for the same region; a first graph structure construction module 920, used to construct a first graph structure with each satellite in the satellite constellation as the first node and the first task unloading relationship of each satellite as the first edge, wherein the first task unloading relationship includes the real task unloading relationship between two adjacent satellites in the satellite constellation and the virtual task unloading relationship between two adjacent satellites corresponding to different satellite constellations, and the virtual task unloading relationship includes unloading task relationship and receiving unloading task relationship; and an iterative update module 930, used to utilize a pre-trained graph neural network model and according to a pre-set calculation... The algorithm iteratively updates the first graph structure to generate the second graph structure; the task unloading relationship determination module 940 is used to construct the first target feature vector and the second target feature vector corresponding to the virtual task unloading relationship in the second graph structure, and determine the second task unloading relationship between each satellite constellation in the second graph structure based on the first target feature vector and the second target feature vector, wherein the second task unloading relationship represents the unloading task relationship or the receiving unloading task relationship; and the task unloading amount prediction module 950 is used to determine the second target feature vector corresponding to the associated node of the third task unloading relationship in the second graph structure, and input the third target feature vector into the pre-trained first prediction model to predict the task unloading amount between each satellite constellation.

[0112] Optionally, the task computation volume determination module 910 includes: a task computation volume calculation module, used to determine the task computation volume of each satellite in multiple satellite constellations for a region under historical time periods; a task computation vector construction module, used to construct a task computation vector corresponding to each satellite based on the task computation volume of each satellite under each historical time period; and a task computation volume output module, used to input each task computation vector into a pre-trained second prediction model and output the task computation volume of each satellite for a region under the current time period.

[0113] Optionally, the first graph structure construction module 920 includes: a node attribute determination module, used to determine the node attributes corresponding to the first node based on the task computation amount of each satellite; a first edge attribute determination module, used to determine the first edge attributes corresponding to the first edge based on the task unloading weight, distance weight, and attenuation coefficient weight corresponding to the real task unloading relationship; a second edge attribute determination module, used to determine the second edge attributes corresponding to the first edge based on the distance weight and attenuation coefficient weight corresponding to the virtual task unloading relationship; and a first graph structure construction submodule, used to construct the first graph structure based on the first node, the first edge attributes, and the second edge attributes.

[0114] Optionally, the graph neural network model includes multiple first MLP models, and the iterative update module 930 includes: an edge attribute vector construction module, used to input the first graph structure into the graph neural network model and construct an edge attribute vector using a first calculation rule corresponding to the first edge, wherein the edge attribute vector includes the edge attribute of the first edge and the node attributes of the two first nodes adjacent to the first edge; and a first iterative update submodule, used to iteratively update the edge attribute vector of the first edge using multiple first MLP models to generate edge features corresponding to the second graph structure.

[0115] Optionally, the graph neural network model includes multiple second MLP models, and the iterative update module 930 includes: a node attribute vector construction module, used to input the first graph structure into the graph neural network model and construct a node attribute vector using the second calculation rule corresponding to the first node, wherein the node attribute vector includes the node attribute of the first node, the edge attribute of the two adjacent first edges of the first node, and the node attribute of the first node corresponding to the two adjacent first edges respectively; and a second iterative update submodule, used to iteratively update the node attribute vector of the first node using multiple second MLP models to generate node features corresponding to the second graph structure.

[0116] Optionally, the task unloading relationship determination module 940 includes: a first target feature vector construction module, used to determine the edge features of the second side corresponding to the unloading task relationship in the virtual task unloading relationship and the node features of the two second nodes corresponding to the second side, and construct a first target feature vector corresponding to the unloading task relationship based on the edge features and node features; a second target feature vector construction module, used to determine the edge features of the second side corresponding to the receiving unloading task relationship in the virtual task unloading relationship and the node features of the two second nodes corresponding to the second side, and construct a second target feature vector corresponding to the receiving unloading task relationship based on the edge features and node features; and a difference vector calculation module, used to calculate the difference vector between the first target feature vector and the second target feature vector, and determine the second task unloading relationship based on the difference vector.

[0117] Optionally, the task offloading prediction module 950 includes: a third target feature vector determination module, used to determine the third target feature vector corresponding to the associated node in the second graph structure, wherein the third target feature vector includes the node feature corresponding to the associated node, the edge feature of the two adjacent second edges of the associated node, and the node feature of the second node corresponding to the two adjacent second edges respectively; and a task offloading determination module, used to input the third target feature vector into the first prediction model and determine the task offloading amount between each satellite constellation, wherein the first prediction model includes an input layer, a fully connected layer, an activation function layer, and an output layer.

[0118] Thus, according to this embodiment, the technical effect of ensuring the smooth completion of computing tasks and improving the utilization rate of computing resources of each satellite constellation is achieved.

[0119] Example 3 Figure 10 An edge computing offloading prediction apparatus 1000 for multi-constellation collaboration according to this embodiment is shown, which corresponds to the method described according to Embodiment 1. Reference Figure 10 As shown, the device 1000 includes: a processor 1010; and a memory 1020 connected to the processor 1010, used to provide the processor 1010 with instructions to process the following steps: determining multiple satellite constellations and the computational workload of each satellite in the constellation for the same region; constructing a first graph structure with each satellite in the constellation as a first node and the first task offloading relationship of each satellite as a first edge, wherein the first task offloading relationship includes the actual task offloading relationship between two adjacent satellites within the constellation and the virtual task offloading relationship between two adjacent satellites corresponding to different satellite constellations, and the virtual task offloading relationship includes offloading task relationship and receiving offloading task relationship; utilizing A pre-trained graph neural network model iteratively updates the first graph structure according to pre-defined calculation rules to generate a second graph structure. A first target feature vector and a second target feature vector corresponding to the virtual task unloading relationship in the second graph structure are constructed. Based on the first and second target feature vectors, a second task unloading relationship between each satellite constellation in the second graph structure is determined, where the second task unloading relationship represents an unloading task relationship or a receiving unloading task relationship. A third target feature vector corresponding to the associated node in the second graph structure with the second task unloading relationship is determined, and the third target feature vector is input into a pre-trained first prediction model to predict the task unloading amount between each satellite constellation.

[0120] Optionally, the operation of determining the task computation volume of multiple satellite constellations and each satellite in the constellation for the same region includes: determining the task computation volume of each satellite in the multiple satellite constellations for the region under historical time periods; constructing a task computation vector corresponding to each satellite based on the task computation volume of each satellite under each historical time period; and inputting each task computation vector into a pre-trained second prediction model and outputting the task computation volume of each satellite for the region under the current time period.

[0121] Optionally, the operation of constructing a first graph structure with each satellite of the satellite constellation as the first node and the first task offloading relationship of each satellite as the first edge includes: determining the node attributes corresponding to the first node based on the task computation amount of each satellite; determining the first edge attributes corresponding to the first edge based on the task offloading weight, distance weight, and attenuation coefficient weight corresponding to the real task offloading relationship; determining the second edge attributes corresponding to the first edge based on the distance weight and attenuation coefficient weight corresponding to the virtual task offloading relationship; and constructing the first graph structure based on the first node, the first edge attributes, and the second edge attributes.

[0122] Optionally, the graph neural network model includes multiple first MLP models, and the operation of using the graph neural network model and iteratively updating the first graph structure according to a pre-defined calculation rule to generate the second graph structure includes: inputting the first graph structure into the graph neural network model, and constructing an edge attribute vector using a first calculation rule corresponding to the first edge, wherein the edge attribute vector includes the edge attribute of the first edge and the node attributes of the two first nodes adjacent to the first edge; and iteratively updating the edge attribute vector of the first edge using multiple first MLP models to generate edge features corresponding to the second graph structure.

[0123] Optionally, the graph neural network model includes multiple second MLP models, and the operation of using the graph neural network model to iteratively update the first graph structure according to a pre-defined calculation rule to generate the second graph structure includes: inputting the first graph structure into the graph neural network model, and constructing a node attribute vector using the second calculation rule corresponding to the first node, wherein the node attribute vector includes the node attribute of the first node, the edge attributes of the two adjacent first edges of the first node, and the node attributes of the first nodes corresponding to the two adjacent first edges respectively; and iteratively updating the node attribute vector of the first node using multiple second MLP models to generate node features corresponding to the second graph structure.

[0124] Optionally, the operation of constructing a first target feature vector corresponding to the virtual task unloading relationship in the second graph structure, and determining the second task unloading relationship between each satellite constellation in the second graph structure based on the first target feature vector, includes: determining the edge features of the second edge corresponding to the unloading task relationship in the virtual task unloading relationship and the node features of the two second nodes corresponding to the second edge, and constructing a first target feature vector corresponding to the unloading task relationship based on the edge features and node features; determining the edge features of the second edge corresponding to the receiving unloading task relationship in the virtual task unloading relationship and the node features of the two second nodes corresponding to the second edge, and constructing a second target feature vector corresponding to the receiving unloading task relationship based on the edge features and node features; and calculating the difference vector between the first target feature vector and the second target feature vector, and determining the second task unloading relationship based on the difference vector.

[0125] Optionally, the operation of inputting the third target feature vector into a pre-trained first prediction model to predict the task offloading amount between each satellite constellation includes: determining the third target feature vector corresponding to the associated node in the second graph structure, wherein the third target feature vector includes the node feature corresponding to the associated node, the edge features of the two adjacent second edges of the associated node, and the node features of the second nodes corresponding to the two adjacent second edges respectively; and inputting the third target feature vector into the first prediction model to determine the task offloading amount between each satellite constellation, wherein the first prediction model includes an input layer, a fully connected layer, an activation function layer, and an output layer.

[0126] Thus, according to this embodiment, the technical effect of ensuring the smooth completion of computing tasks and improving the utilization rate of computing resources of each satellite constellation is achieved.

[0127] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0128] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0129] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0130] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0131] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0132] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0133] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A multi-constellation collaborative edge computing offloading prediction method, characterized in that, include: Determine the computational workload of multiple satellite constellations and the task workload of each satellite in the constellation for the same region. Using each satellite of the satellite constellation as the first node and the first task offloading relationship of each satellite as the first edge, a first graph structure is constructed. The first task offloading relationship includes the real task offloading relationship between two adjacent satellites in the satellite constellation and the virtual task offloading relationship between two adjacent satellites corresponding to different satellite constellations. The virtual task offloading relationship includes offloading task relationship and receiving offloading task relationship. The process of generating a second graph structure by iteratively updating the first graph structure using a pre-trained graph neural network model according to pre-defined computation rules includes: The first graph structure is input into the graph neural network model, and an edge attribute vector is constructed using the first calculation rule corresponding to the first edge, wherein the edge attribute vector includes the edge attribute of the first edge and the node attribute of the two first nodes adjacent to the first edge. The edge attribute vector of the first edge is iteratively updated using the multiple first MLP models to generate edge features corresponding to the second graph structure; The first graph structure is input into the graph neural network model, and a node attribute vector is constructed using the second calculation rule corresponding to the first node. The node attribute vector includes the node attribute of the first node, the edge attribute of the two adjacent first edges of the first node, and the node attribute of the first node corresponding to the two adjacent first edges respectively. The node attribute vector of the first node is iteratively updated using the multiple second MLP models to generate node features corresponding to the second graph structure; Construct a first target feature vector and a second target feature vector in the second graph structure that correspond to the virtual task unloading relationship, and determine the second task unloading relationship between each satellite constellation in the second graph structure based on the first target feature vector and the second target feature vector, wherein the second task unloading relationship represents an unloading task relationship or a receiving unloading task relationship; A third target feature vector is determined corresponding to the associated node in the second graph structure that is related to the second task unloading relationship, and the third target feature vector is input into a pre-trained first prediction model to predict the task unloading amount between the satellite constellations. The third target feature vector includes the node feature corresponding to the associated node, the edge feature of the two adjacent second edges of the associated node, and the node feature of the second node corresponding to the two adjacent second edges respectively.

2. The method according to claim 1, characterized in that, The operations for determining multiple satellite constellations and the computational workload of each satellite in the constellation for the same region include: The computational workload of each satellite in the multiple satellite constellations for the region under a defined historical period; Based on the computational workload of each satellite in various historical time periods, a corresponding computational vector is constructed for each satellite; and Each task calculation vector is input into a pre-trained second prediction model, and the task calculation amount of each satellite for the region in the current time period is output.

3. The method according to claim 1, characterized in that, The operation of constructing a first graph structure, using each satellite of the satellite constellation as the first node and the first task offloading relationship of each satellite as the first edge, includes: Based on the task computation volume of each satellite, determine the node attributes corresponding to the first node; Based on the task unloading weight, distance weight, and attenuation coefficient weight corresponding to the actual task unloading relationship, determine the attributes of the first side corresponding to the first side. Based on the distance weight and attenuation coefficient weight corresponding to the virtual task unloading relationship, determine the attributes of the second side corresponding to the first side; and The first graph structure is constructed based on the first node, the first edge attribute, and the second edge attribute.

4. The method according to claim 1, characterized in that, The operation of constructing a first target feature vector corresponding to the virtual task offloading relationship in the second graph structure, and determining the second task offloading relationship between each satellite constellation in the second graph structure based on the first target feature vector, includes: Determine the edge features of the second edge corresponding to the unloading task relationship in the virtual task unloading relationship and the node features of the two second nodes corresponding to the second edge, and construct a first target feature vector corresponding to the unloading task relationship based on the edge features and the node features; Determine the edge features of the second edge corresponding to the receiving and unloading task relationship in the virtual task unloading relationship, and the node features of the two second nodes corresponding to the second edge; and construct a second target feature vector corresponding to the receiving and unloading task relationship based on the edge features and the node features; and Calculate the difference vector between the first target feature vector and the second target feature vector, and determine the second task unloading relationship based on the difference vector.

5. The method according to claim 1, characterized in that, The operation of inputting the third target feature vector into a pre-trained first prediction model to predict the mission offloading amount between the various satellite constellations includes: Determine the third target feature vector corresponding to the associated node in the second graph structure; and The third target feature vector is input into the first prediction model, and the task offloading amount between each satellite constellation is determined, wherein the first prediction model includes an input layer, a fully connected layer, an activation function layer, and an output layer.

6. A storage medium, characterized in that, The storage medium includes a stored program, wherein, when the program is executed, the method described in any one of claims 1 to 5 is performed by a processor.

7. A multi-constellation collaborative edge computing offloading prediction device, characterized in that, include: The task computation volume determination module is used to determine the task computation volume of multiple satellite constellations and each satellite in the satellite constellation for the same area. The first graph structure construction module is used to construct a first graph structure with each satellite of the satellite constellation as the first node and the first task unloading relationship of each satellite as the first edge. The first task unloading relationship includes the real task unloading relationship between two adjacent satellites in the satellite constellation and the virtual task unloading relationship between two adjacent satellites corresponding to different satellite constellations. The virtual task unloading relationship includes unloading task relationship and receiving unloading task relationship. An iterative update module is used to iteratively update the first graph structure using a pre-trained graph neural network model according to pre-defined calculation rules to generate a second graph structure. The graph neural network model includes multiple first MLP models and multiple second MLP models. The operation of iteratively updating the first graph structure using the graph neural network model according to pre-defined calculation rules to generate the second graph structure includes: The first graph structure is input into the graph neural network model, and an edge attribute vector is constructed using the first calculation rule corresponding to the first edge, wherein the edge attribute vector includes the edge attribute of the first edge and the node attribute of the two first nodes adjacent to the first edge. The edge attribute vector of the first edge is iteratively updated using the multiple first MLP models to generate edge features corresponding to the second graph structure; The first graph structure is input into the graph neural network model, and a node attribute vector is constructed using the second calculation rule corresponding to the first node. The node attribute vector includes the node attribute of the first node, the edge attribute of the two adjacent first edges of the first node, and the node attribute of the first node corresponding to the two adjacent first edges respectively. The node attribute vector of the first node is iteratively updated using the multiple second MLP models to generate node features corresponding to the second graph structure; The task unloading relationship determination module is used to construct a first target feature vector and a second target feature vector in the second graph structure corresponding to the virtual task unloading relationship, and to determine a second task unloading relationship between the satellite constellations in the second graph structure based on the first target feature vector and the second target feature vector, wherein the second task unloading relationship represents an unloading task relationship or a receiving unloading task relationship; and The task unloading volume prediction module is used to determine the third target feature vector corresponding to the associated node in the second graph structure that is related to the second task unloading relationship, and input the third target feature vector into the pre-trained first prediction model to predict the task unloading volume between the satellite constellations. The third target feature vector includes the node feature corresponding to the associated node, the edge feature of the two adjacent second edges of the associated node, and the node feature of the second node corresponding to the two adjacent second edges respectively.

8. A multi-constellation collaborative edge computing offloading prediction device, characterized in that, include: processor; as well as A memory, connected to the processor, for providing the processor with instructions to perform the following processing steps: Determine the computational workload of multiple satellite constellations and the task workload of each satellite in the constellation for the same region. Using each satellite of the satellite constellation as the first node and the first task offloading relationship of each satellite as the first edge, a first graph structure is constructed. The first task offloading relationship includes the real task offloading relationship between two adjacent satellites in the satellite constellation and the virtual task offloading relationship between two adjacent satellites corresponding to different satellite constellations. The virtual task offloading relationship includes offloading task relationship and receiving offloading task relationship. The process of generating a second graph structure by iteratively updating the first graph structure using a pre-trained graph neural network model according to pre-defined computation rules includes: The first graph structure is input into the graph neural network model, and an edge attribute vector is constructed using the first calculation rule corresponding to the first edge, wherein the edge attribute vector includes the edge attribute of the first edge and the node attribute of the two first nodes adjacent to the first edge. The edge attribute vector of the first edge is iteratively updated using the multiple first MLP models to generate edge features corresponding to the second graph structure; The first graph structure is input into the graph neural network model, and a node attribute vector is constructed using the second calculation rule corresponding to the first node. The node attribute vector includes the node attribute of the first node, the edge attribute of the two adjacent first edges of the first node, and the node attribute of the first node corresponding to the two adjacent first edges respectively. The node attribute vector of the first node is iteratively updated using the multiple second MLP models to generate node features corresponding to the second graph structure; Construct a first target feature vector and a second target feature vector in the second graph structure corresponding to the virtual task unloading relationship, and determine a second task unloading relationship between the satellite constellations in the second graph structure based on the first target feature vector and the second target feature vector, wherein the second task unloading relationship represents an unloading task relationship or a receiving unloading task relationship; determine a third target feature vector corresponding to the associated node in the second graph structure that is associated with the second task unloading relationship, and input the third target feature vector into a pre-trained first prediction model to predict the task unloading amount between the satellite constellations, wherein the third target feature vector includes the node feature corresponding to the associated node, the edge feature of the two adjacent second edges of the associated node, and the node feature of the second node corresponding to the two adjacent second edges respectively.

Images