A method and system for dynamic allocation of resources for space-air-ground integrated network

Abstract

The application discloses a kind of resource dynamic allocation methods and systems for space-air-ground integrated network, the method includes the following steps: deploying multiple task processing models with different processing accuracies on all nodes of space-air-ground integrated network;Collect the task generated by user equipment;The node of space-air-ground integrated network adaptively selects task processing model according to the characteristics and data type of the task;According to the accuracy of task processing model, the total cost is the objective function, and task allocation is carried out;When task allocation is carried out, it is first allocated to ground network and space-based network, and when the number of the task exceeds the task quantity threshold of ground network and space-based network, the task exceeding the task quantity threshold is allocated to air-based network for processing.The application improves the task processing capability of space-air-ground integrated network and the service quality of network by innovative computing offloading method and adaptive distributed computing model.

Description

Technical Field

[0001] The present invention relates to the field of wireless communications, and in particular to a method and system for dynamic resource allocation in an air-ground-space integrated network. Background Art

[0002] With the development of wireless communication and telecommunications technologies, 5G networks have made significant progress in bandwidth, capacity, and latency. The heterogeneous data generated by large-scale intelligent devices has driven the rapid development of the telecommunications industry, but the growing demand for data computing and communication between devices has placed higher demands on terrestrial communication and computing networks.

[0003] A single terrestrial communication and computing network is far from meeting current communication needs, let alone supporting the strategic imperative of a digital society seamlessly transitioning from 5G to 6G. As one of the core technologies of the 6G era, the integrated space-air-ground network, based on ground-based networks and supplemented and extended by space-based and air-based networks, provides intelligent and efficient information services for a wide range of network applications across a wide spatial area, playing a key role in the development of the 6G network architecture. Unlike a single terrestrial communication network, the integrated space-air-ground-ground network transcends geographical constraints and is applicable not only to scenarios like smart cities but also to smart agriculture, natural environment monitoring, disaster management and prevention, and marine exploration.

[0004] Satellites, as a crucial component of integrated space-ground networks, are crucial to the network's overall performance. Currently, most satellites are limited to data transmission and communication. A small number of computing satellites have relatively limited computing capabilities due to limitations in power consumption, heat dissipation, computing hardware, and physical size. Therefore, for integrated space-ground networks, dynamically allocating heterogeneous resources based on the mission requirements of the satellite's corresponding region and the satellite's own availability is crucial. However, most resource allocation methods for integrated space-ground networks only consider the satellite's orbit, ignoring the satellite's inherent computing power and the quality of service requirements of the region in which it is located. Summary of the Invention

[0005] The present invention proposes a dynamic resource allocation method and system for an air-space-ground integrated network, which solves the problem that the existing resource allocation method for the air-space-ground integrated network ignores the service quality requirements and the computing capacity of the satellite itself.

[0006] To solve the above technical problems, the present invention provides a method for dynamic resource allocation for an air-space-ground integrated network, comprising the following steps:

[0007] Deploy multiple task processing models with different processing accuracy on all nodes of the integrated air-space-ground-integrated network; collect tasks generated by user devices; the nodes of the integrated air-space-ground-integrated network adaptively select task processing models according to the characteristics and data types of the tasks; allocate tasks based on the accuracy of the task processing models and with total cost as the objective function; when allocating tasks, first allocate them to the ground-based network and the space-based network; when the number of tasks exceeds the task number threshold of the ground-based network and the space-based network, allocate the tasks that exceed the task number threshold to the air-based network for processing.

[0008] Preferably, the characteristics of the task include accuracy priority, efficiency priority, minimum energy consumption, accuracy and efficiency balance, accuracy and energy consumption balance, efficiency and energy consumption balance, accuracy, efficiency and energy consumption balance, and the data types of the task include text tasks, image tasks, voice tasks, text and image tasks, text, image and voice tasks.

[0009] Preferably, the total cost objective function includes communication delay, computing time cost and global task processing accuracy during node aggregation. When the task is assigned to airborne network processing, the total cost objective function also includes the time cost of node synchronization between satellites.

[0010] Preferably, the expression of the global task processing accuracy is:

[0011]

[0012]

[0013] Where A max is the global task processing accuracy; T i P is the task quantity threshold of node i; T i G is the number of tasks assigned to node i; M is the number of task processing models; Indicates whether model m is selected by node i. If selected, otherwise A m is the accuracy of the task processing model m; α is the weight coefficient; N is the number of nodes; E i is the energy consumption of node i; is the maximum energy that node i can use for task processing; represents the total energy of node i at time T; represents the total energy of node i at time Tt; It represents the minimum energy consumption required by node i to maintain its normal operation.

[0014] Preferably, the expression of the communication delay during node aggregation is:

[0015] T comm =δ1*(T SG +T GA +T AS +N SS *T SS );

[0016]

[0017] In the above formula, T comm is the communication delay when nodes are aggregated; δ1 and δ2 are weight coefficients; T SG represents the transmission delay between the satellite and the ground equipment; T GA represents the transmission delay between the ground equipment and the UAV in the air; T AS represents the transmission delay between the drone and the satellite; T SS represents the transmission delay between satellites; N SS Indicates the number of retransmissions between satellites; represents the transmission delay between nodes; M model Represents the decay ratio of the task processing model; BW AB represents the bandwidth between node A and node B; p AB represents the transmission power between node A and node B; h AB Indicates that the channel fading between node A and node B is sparse; σ is the noise function; is the propagation delay between node A and node B; L AB is the length of the physical link between node A and node B; V AB is the propagation speed in the wireless medium between node A and node B.

[0018] Preferably, the expression for the computation time cost is:

[0019]

[0020] Where, T comp is the calculation time cost; δ1 and δ2 are weight coefficients; T train represents the computation time cost of nodes in the ground-based network; and represents the aggregate time cost of aerial drone nodes and satellite nodes respectively; FLOPS model Indicates the floating-point operations of the task processing model; Indicates the amount of data for the task; N ep Indicates the number of calculations; FLOPS indicates the floating-point operations per second of the node; Indicates the number of task processing models selected by the node.

[0021] Preferably, the time cost of inter-satellite node synchronization is expressed as:

[0022]

[0023] Where N S is the number of satellites; and Represents the transmission time cost and propagation delay between satellites; FLOPS S Indicates the floating point operations per second of the satellite node; M model Represents the decay ratio of the task processing model.

[0024] Preferably, the expression of the total cost objective function is:

[0025]

[0026] T SAG =T comm +T comp +T sync ;

[0027] Where A represents the number of node coordination strategies; λ1 and λ2 are adjustable weight coefficients; T comm T is the communication delay when nodes aggregate; comp is the calculation time cost; T sync A is the synchronization time cost of the satellite node; max Processing accuracy for global tasks.

[0028] A dynamic resource allocation system for an air-space-ground integrated network, applicable to the above-mentioned dynamic resource allocation method for an air-space-ground integrated network, comprising: a data acquisition module, a judgment module, an adaptive collaboration module, and a result aggregation module;

[0029] The data acquisition module is used to collect and update parameter information of all nodes in the air-ground-space integrated network;

[0030] The judgment module is used to poll the ground-based network in the integrated space-ground-air network and the space-based network covering the ground-based network area to see whether there is a need to switch the task processing model. When there is a switching demand, it determines the relationship between the number of tasks generated by the user equipment and the task thresholds of the ground-based network and the space-based network, and sends the judgment result to the adaptive collaboration module;

[0031] The adaptive collaboration module is used to adaptively select a collaborative network and a task processing model for each node to participate in task processing;

[0032] The result aggregation module is used to receive the task processing results of the adaptive collaboration module, notify the judgment module and feed back the corresponding task processing results to the terminal device.

[0033] Preferably, the parameter information includes wireless communication bandwidth, data size of the task, accuracy of the task processing model, energy consumption of the task processing model, energy consumption of the node, and maximum energy that the node can use for task processing.

[0034] The benefits of the present invention include at least:

[0035] 1. The computation offloading method adopted by the present invention has less computational effort during execution and reduces computational complexity, which not only reduces the burden on individual computing nodes but also improves the operating efficiency of the entire network.

[0036] 2. Taking full account of user task requirements, we have designed an adaptive distributed computing model that can dynamically adjust the allocation and scheduling of computing resources based on task requirements, ensuring that each task can be processed on the most appropriate computing node;

[0037] In general, the present invention not only improves the task processing capability of the integrated air-space-ground network but also significantly improves the service quality of the network through an innovative computing offloading method and an adaptive distributed computing model. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is a flow chart of a method according to an embodiment of the present invention;

[0039] Figure 2 A structural diagram of an adaptive distributed computing system according to an embodiment of the present invention;

[0040] Figure 3 This is a comparison chart of the average accuracy of the method according to the embodiment of the present invention and the existing method;

[0041] Figure 4 A comparison chart of average energy consumption between the method according to an embodiment of the present invention and the existing method;

[0042] Figure 5 2 is a comparison chart of the average processing time of the method according to the embodiment of the present invention and the existing method. DETAILED DESCRIPTION

[0043] The following is a clear and complete description of the technical solutions in the embodiments of the present invention, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts are within the scope of protection of the present invention.

[0044] When providing various IoT information services, integrated air-space-ground networks must meet varying quality of service requirements across metrics such as data transmission latency and reliability, and computational accuracy and efficiency. However, simultaneously optimizing all these metrics from the perspectives of network design, data communication, and task processing incurs significant energy, storage, communication, and computing resource overhead, making them unsuitable for resource-constrained nodes in integrated air-space-ground networks. Therefore, designing model-adaptive distributed computing methods that combine information service quality with the goal of meeting information service requirements while simultaneously reducing integrated air-space-ground computational latency is crucial for optimizing inter-satellite resource allocation and improving the task processing capabilities and service quality of integrated air-space-ground networks.

[0045] Example 1

[0046] like Figure 1 As shown, an embodiment of the present invention provides a method for dynamic resource allocation for an air-ground integrated network, comprising the following steps:

[0047] Step S1: Set up multiple task processing models based on the task type and service characteristics required by the user, and deploy the perception model on each node of the air-ground integrated network. The perception model includes various task processing models. The perception model collects and updates the parameter information of all nodes, including: wireless communication bandwidth BW, task data size D, task processing model accuracy A m , by model m i Energy consumption E for processing information m , node energy consumption E i , the maximum energy that a node can use for information processing

[0048] Specifically, the integrated air-space-ground network established by the present invention comprises three parts: a ground-based network, a space-based network, and an air-based network. The ground-based and space-based networks pre-deploy complete task processing models. Due to its own resource limitations, the air-based network only deploys the segmented task processing models. Multiple task processing models with different parameters are deployed simultaneously in each node. The nodes involved in this invention are no longer limited to the limitations of traditional computer architectures. Any device with wireless transmission and computing capabilities can be considered a node, such as mobile phones, smart cars, drones flying in the sky, and satellites in various orbits in space.

[0049] The selection of the task processing model in the present invention is performed with reference to the settings in Table 1 below. In Table 1, the first row indicates the set data type, the first column indicates the set task characteristics, and the rest is the set task processing model.

[0050] Table 1 Task processing model selection comparison table

[0051]

[0052] Step S2: Tasks generated by user devices are collected and assigned to the nodes closest to the user devices in the integrated air-ground-space network, randomly selecting a task processing model. When the network environment changes, the task processing model selection of the ground edge nodes that make up the ground-based network and the UAV nodes covering the ground-based network area is polled to see if it has changed. Tasks are reallocated if the task type or data characteristics processed by the nodes change.

[0053] Specifically, network environment changes include factors that may cause network environment fluctuations, such as task type conversion, node sleep and wake-up, and changes in user needs. The criteria for judging these network environment changes are as follows:

[0054] Task type conversion: The judgment criterion is whether the data type processed by the task processing model changes. If it changes, the task type changes; otherwise, it does not change.

[0055] Node sleep and wake-up: The judgment standard is whether the working state of the node has changed. If the node changes from a non-working state to a working state or is always in a working state within a specified continuous time, the node is in the awake state; if the node changes from a working state to a non-working state or is always in a non-working state within a specified continuous time, the node is in the sleep state.

[0056] Changes in user needs: The judgment standard is whether the task characteristics have changed, such as whether accuracy priority, efficiency priority, energy saving priority, etc. have changed. If so, the user needs have changed, otherwise, they will not change.

[0057] Step S3: Determine the number of tasks T assigned to each node in the ground-based network and the space-based network i G and the number of tasks T that are processed on node i within the specified time i P The relationship between the local network and the space-based network is adaptively selected to participate in the collaborative network task processing. When the number of tasks processed by the nodes in the local network and the space-based network exceeds the set task processing threshold, step S4 is executed; otherwise, step S5 is executed. The collaborative network adaptively selected to participate in information processing should meet the following conditions:

[0058]

[0059] Step S4: Collect the data type and characteristics of the new task, and adaptively extract the data from the perception model according to the data type and characteristics of the new task. Select the corresponding task processing model in the ground-based network and the space-based network, according to the number of tasks T assigned to each node. iG and the number of tasks T that are processed on node i within the specified time i P Calculate the overall task processing accuracy A of the ground-based network and the space-based network sum , time transmission cost T ga , we can get the total system mission cost C of the ground-based network and the space-based network AG .

[0060] Node i can adaptively select the optimal task processing model, which not only meets the processing accuracy A required by the task but also reduces the system energy consumption. The overall task processing accuracy A of the ground-based network and the space-based network sum In addition to being related to the task processing accuracy of each node i, it is also related to the number of tasks T assigned to node i. i G And the number of tasks T that can be completed on node i within the specified control time i P Related, that is, satisfying the following relationship:

[0061]

[0062] Where M is the number of task processing models; Indicates whether model m is selected by node i. If selected, otherwise A m is the accuracy of the task processing model m; α is the weight coefficient; N is the number of nodes; E i is the energy consumption of node i; is the maximum energy that node i can use for task processing; represents the total energy of node i at time T; represents the total energy of node i at time Tt; It represents the minimum energy consumption required by node i to maintain its normal operation.

[0063] The time transmission cost T of all nodes in the ground-based network and the space-based network ga The calculation method is:

[0064]

[0065] Where, Indicates whether the communication path established by device q between node i and node j is used. If so, otherwise represents the task that device q forwards from node i to node j for processing; D represents the data size of the task; h ij represents the number of hops between node i and node j; L represents the time cost of a single hop.

[0066] Generated in node i and in the node set Cost of tasks completed The calculation method is:

[0067]

[0068] Among them, v ij represents the cost of the task forwarded from node i to node j, which is calculated as:

[0069]

[0070] The total system mission cost C of the ground-based network and the space-based network AG The calculation method is:

[0071]

[0072] Where A represents the number of node coordination strategies, which are automatically generated by the system; β, ε, and η are adjustable weight coefficients.

[0073] After the system automatically generates multiple node system strategies, it will calculate the total cost of the system tasks C AG The node coordination strategy with the smallest total task cost is selected as the node coordination strategy within the specified time of this task.

[0074] Step S5: Assign the tasks that exceed the task threshold to the air-based network for processing, and let the ground-based network and space-based network process the remaining tasks, collect the data type and characteristics of the new tasks, and adaptively extract the data from the perception model according to the data type and characteristics of the new tasks. Select the corresponding task processing model and calculate the communication delay T when the task processing model of each node in the airborne network is aggregated. comm , calculation time cost T comp , the time cost T of synchronizing the task processing model between satellites sync , the global task processing accuracy of the air-space-ground integrated network A max , we get the total system mission cost C of the air-ground integrated network SAG .

[0075] Specifically, the global task processing accuracy A of the air-ground integrated network max The calculation method is the same as the overall task processing accuracy A of the ground-based network and the space-based network in step S4 sum The calculation method is the same as that of , and the node range involved is all nodes in the integrated air-space-ground network, while step S4 only involves nodes in the ground-based network and the space-based network.

[0076] Communication delay T when task processing model of each node in the air-space-ground integrated network is aggregated comm It is composed of the communication delay between devices at different levels in the network and is calculated as:

[0077] T comm =δ1*(T SG +T GA +T AS +N SS *T SS );

[0078] Among them, δ1 is the weight coefficient; T SG represents the transmission delay between the satellite and the ground equipment; T GA represents the transmission delay between the ground equipment and the UAV in the air; T AS represents the transmission delay between the drone and the satellite; T SS represents the transmission delay between satellites; N SS Indicates the number of retransmissions between satellites.

[0079] T SG 、T GA 、T AS and T SS The calculation method is:

[0080]

[0081] Among them, AB∈{SG,GA,AS,SS}, G represents ground nodes, A represents aerial drone nodes, and S represents satellite nodes; represents the transmission delay between nodes; M model Represents the decay ratio of the task processing model; BW AB Indicates the bandwidth between device A and device B; p AB Indicates the transmission power between device A and device B; h AB Indicates that the channel fading between device A and device B is sparse; σ is the noise function; The propagation delay can be expressed as the length L of the physical link between device A and device B. AB and the propagation speed V in the wireless medium AB Calculated; T AB Represents the end-to-end delay of transmission between nodes.

[0082] The computational time cost of the integrated air-space-ground network T comp It consists of the computational time cost and aggregation time cost of the task processing model, and is calculated as follows:

[0083]

[0084]

[0085] Among them, δ2 is the weight coefficient; T train represents the computation time cost of nodes in the ground-based network; and represents the aggregate time cost of aerial drone nodes and satellite nodes respectively; FLOPS model Indicates the floating-point operations of the task processing model; Indicates the amount of data for the task;

[0086] N ep Indicates the number of calculations; FLOPS indicates the floating-point operations per second of ground equipment; Indicates the number of models selected by the node.

[0087] The synchronization time cost T of the task processing model between satellites in the space-based network sync The calculation method is:

[0088]

[0089] Among them, N S is the number of satellites; and Represents the transmission time cost and propagation delay between satellites; FLOPS S Indicates the number of floating-point operations per second used for satellite calculations.

[0090] The total time cost T of task processing in the air-ground integrated network SAG The calculation method is:

[0091] T SAG =T comm +T comp +T sync .

[0092] The total system mission cost C of the integrated air-space-ground network SAG The calculation method is:

[0093]

[0094] Among them, λ1 and λ2 are adjustable weight coefficients.

[0095] According to the total cost of system tasks C SAG The node coordination strategy with the minimum cost is selected as the node coordination strategy within the specified time of this task.

[0096] Step S6: According to the strategy selected in step S4 or step S5, the task is completed within the specified time and the task processing result is fed back to the user terminal. When the specified time is over, return to step S1.

[0097] Example 2

[0098] This embodiment provides a resource dynamic allocation system for an air-ground integrated network. Figure 2 As shown, the system includes:

[0099] The data acquisition module 201 is used to collect and update the parameter information of all computing nodes before the specified time of task execution, including: wireless communication bandwidth, data size of the task, accuracy of the task model, energy consumption of processing the task according to the model, node energy consumption, and the maximum energy that the node can use for task processing.

[0100] The judgment module 202 is used to poll the ground edge nodes that constitute the ground-based network in the integrated air-ground-space network and the UAV nodes covering the ground-based network area to see whether there is a need to switch the task model when the network environment changes. When there is a switching demand, it further judges the relationship between the number of information generated by the node and the number of tasks completed within the specified time, and sends the judgment result to the adaptive collaboration module.

[0101] Adaptive collaboration module 203 is used to adaptively select collaborative networks for task processing. Based on the results transmitted by the modules, the adaptive collaboration module selects a collaborative network consisting of ground-based and space-based nodes, or a collaborative network consisting of ground-based, space-based, and airborne nodes. It also calculates the total cost of the system task under various scenarios, selects the scenario with the lowest cost as the node collaboration strategy, executes the task, and sends the results to the result aggregation module.

[0102] The result aggregation module 204 is configured to receive the task processing results of the adaptive collaboration module, notify the judgment module 202 and feed back the corresponding task processing results to the terminal device.

[0103] The present invention provides a method and system for dynamic resource allocation for an integrated air-space-ground-integrated network. The computational offloading method adopted has a small computational amount, low computational complexity, easy device implementation, and strong practicality. Different from other methods of processing data using specific types of artificial intelligence models, the present invention fully considers the characteristics of the node such as energy consumption and computing resource consumption, and service requirements such as data processing accuracy, and proposes a matching task processing model adaptive decision-making method based on the node's own characteristics, thereby improving the task processing capability of the integrated air-space-ground-integrated network.

[0104] Overall, this invention effectively reduces energy consumption in integrated air-ground networks and reduces processing latency for user devices, while still meeting service requirements. This has significant practical significance and application value for information service processing in underdeveloped and underdeveloped areas where terrestrial communication and computing networks are underdeveloped.

[0105] Figure 3 、 Figure 4 、 Figure 5 The Enhanced method of the present invention was compared with the existing Self-Processing algorithm, Greedy algorithm, Shortest Path Algorithm (SP), and Proportional Distribution Algorithm (PD) in terms of average task accuracy, average task energy consumption, and average algorithm processing time. It can be seen that the Enhanced method of the present invention improves task accuracy while reducing average task energy consumption and average processing time.

[0106] The technical features of the above embodiments may be combined in any manner. To simplify the description, not all possible combinations of the technical features in the above embodiments are described. Only preferred embodiments of the present invention are presented. While the description is relatively specific and detailed, it should not be construed as limiting the scope of the present invention. As long as there are no contradictions in the combination of these technical features, they should be considered to be within the scope of this specification.

[0107] It should be noted that those skilled in the art may make various modifications and improvements without departing from the scope of the present invention, and these modifications and improvements fall within the scope of protection of the present invention. Therefore, the scope of protection of the patent for this invention shall be based on the appended claims.

Claims

1. A method for dynamic resource allocation in an air-ground-space integrated network, characterized by: The following steps are involved: Deploy multiple task processing models with different processing accuracies on all nodes of the integrated air-ground-space network; collect tasks generated by user devices; The nodes of the integrated air-space-ground network adaptively select a task processing model based on the characteristics and data types of the tasks; perform task allocation based on the accuracy of the task processing model and with total cost as the objective function; when allocating tasks, first allocate tasks to the ground-based network and the space-based network; when the number of tasks exceeds a threshold number of tasks for the ground-based network and the space-based network, the tasks exceeding the threshold number of tasks are allocated to the air-based network for processing; the total cost objective function includes global task processing accuracy, and the expression for the global task processing accuracy is: Where A max Processing accuracy for global tasks; T i P is the task quantity threshold of node i; T i G is the number of tasks assigned to node i; M is the number of task processing models; Indicates whether model m is selected by node i. If selected, otherwise A m is the accuracy of the task processing model m; α is the weight coefficient; N is the number of nodes; E i is the energy consumption of node i; is the maximum energy that node i can use for task processing; represents the total energy of node i at time T; represents the total energy of node i at time Tt; It represents the minimum energy consumption required by node i to maintain its normal operation.

2. The method for dynamic resource allocation for an air-ground-integrated network according to claim 1, characterized in that: The characteristics of the tasks include accuracy priority, efficiency priority, minimum energy consumption, accuracy and efficiency balance, accuracy and energy consumption balance, efficiency and energy consumption balance, accuracy, efficiency and energy consumption balance; the data types of the tasks include text tasks, image tasks, speech tasks, text and image tasks, text, image and speech tasks.

3. The method for dynamic resource allocation for an air-ground-integrated network according to claim 1, characterized in that: The total cost objective function includes communication delay, computing time cost and global task processing accuracy when nodes are aggregated. When tasks are assigned to air-based network processing, the total cost objective function also includes the time cost of node synchronization between satellites.

4. The method for dynamic resource allocation for an air-ground-integrated network according to claim 3, characterized in that: The expression of the communication delay when the nodes are aggregated is: T comm =δ1*(T SG +T GA +T AS +N SS *T SS ); In the above formula, T comm is the communication delay when nodes are aggregated; δ1 and δ2 are weight coefficients; T SG represents the transmission delay between the satellite and the ground equipment; T GA represents the transmission delay between the ground equipment and the aerial drone; T AS represents the transmission delay between the drone and the satellite; T SS represents the transmission delay between satellites; N SS Indicates the number of retransmissions between satellites; represents the transmission delay between nodes; M model Represents the decay ratio of the task processing model; BW AB represents the bandwidth between node A and node B; p AB represents the transmission power between node A and node B; h AB Indicates that the channel fading between node A and node B is sparse; σ is the noise function; is the propagation delay between node A and node B; L AB is the length of the physical link between node A and node B; V AB is the propagation speed in the wireless medium between node A and node B.

5. The method for dynamic resource allocation for an air-ground-integrated network according to claim 3, characterized in that: The expression of the computation time cost is: Where, T comp is the calculation time cost; δ1 and δ2 are weight coefficients; T train represents the computation time cost of nodes in the ground-based network; and represents the aggregate time cost of aerial drone nodes and satellite nodes respectively; FLOPS model Indicates the floating-point operations of the task processing model; Indicates the amount of data for the task; N ep Indicates the number of calculations; FLOPS indicates the floating-point operations per second of the node; Indicates the number of task processing models selected by the node; M model Represents the decay ratio of the task processing model.

6. The method for dynamic resource allocation for an air-ground-integrated network according to claim 3, characterized in that: The expression of the time cost of inter-satellite node synchronization is: Where N S is the number of satellites; and Represents the transmission time cost and propagation delay between satellites; FLOPS S Indicates the floating point operations per second of the satellite node; M model Represents the decay ratio of the task processing model.

7. The method for dynamic resource allocation for an air-ground-integrated network according to claim 3, characterized in that: The expression of the total cost objective function is: T SAG =T comm +T comp +T sync ; Where A represents the number of node coordination strategies; λ1 and λ2 are adjustable weight coefficients; T comm T is the communication delay when nodes aggregate; comp is the calculation time cost; T sync A is the synchronization time cost of the satellite node; max Processing accuracy for global tasks.

8. A dynamic resource allocation system for an integrated air-space-ground-integrated network, applicable to a dynamic resource allocation method for an integrated air-space-ground-integrated network as claimed in any one of claims 1 to 7, characterized in that: It includes data collection module, judgment module, adaptive collaboration module and result aggregation module; The data acquisition module is used to collect and update parameter information of all nodes in the air-ground-space integrated network; The judgment module is used to poll the ground-based network in the integrated space-ground-air network and the space-based network covering the ground-based network area to see whether there is a need to switch the task processing model. When there is a switching demand, it determines the relationship between the number of tasks generated by the user equipment and the task thresholds of the ground-based network and the space-based network, and sends the judgment result to the adaptive collaboration module; The adaptive collaboration module is used to adaptively select a collaborative network and a task processing model for each node to participate in task processing; The result aggregation module is used to receive the task processing results of the adaptive collaboration module, notify the judgment module and feed back the corresponding task processing results to the terminal device.

9. The resource dynamic allocation system for an integrated air-ground-space network according to claim 8, characterized in that: The parameter information includes wireless communication bandwidth, data size of the task, accuracy of the task processing model, energy consumption of the task processing model, energy consumption of the node, and maximum energy that the node can use for task processing.

Images