Satellite network resource scheduling method, device and equipment

Abstract

This specification relates to the field of satellite network resource scheduling technology and provides a satellite network resource scheduling method, apparatus, and device. The method includes: identifying the user intent of a service request, wherein the user intent represents the service request's intended demand for satellite network resources; determining the target service type corresponding to the user intent based on a pre-trained neural network model, wherein the neural network model includes a nonlinear mapping relationship between user intent and service type; and allocating satellite network resources corresponding to the target service type to the service request. The embodiments of this specification facilitate adaptively meeting users' complex satellite network service needs.

Description

Technical Field

[0001] This specification relates to the technical field of satellite network resource scheduling, and in particular to a satellite network resource scheduling method, device, and equipment. Background Art

[0002] Traditional technologies use fixed-division satellite network slicing to schedule satellite network resources. However, this approach struggles to meet users' complex satellite network service needs. Summary of the Invention

[0003] The purpose of the embodiments of this specification is to provide a satellite network resource scheduling method, apparatus, and device to facilitate adaptively meeting users' complex satellite network service needs.

[0004] To achieve the above objectives, on the one hand, an embodiment of this specification provides a satellite network resource scheduling method, including:

[0005] Identifying a user intention of a service request, where the user intention represents the demand intention of the service request for satellite network resources;

[0006] Determining a target service type corresponding to the user intent based on a pre-trained neural network model, wherein the neural network model includes a nonlinear mapping relationship between user intent and service type;

[0007] Allocate satellite network resources corresponding to the target service type to the service request.

[0008] In the satellite network resource scheduling method of the embodiment of this specification, the satellite network resources include satellite network resources of multiple satellite networks; the multiple satellite networks include low-orbit satellite networks, medium-orbit satellite networks and / or high-orbit satellite networks.

[0009] In the satellite network resource scheduling method of the embodiment of this specification, the neural network model includes a graph neural network model, and the resource topology structure of the multiple satellite networks is represented based on an undirected graph. In the undirected graph, different nodes represent different satellite network resources, and the edges connecting the nodes represent the association relationship between different satellite network resources.

[0010] In the satellite network resource scheduling method of the embodiment of this specification, the association relationship includes the integration cost of different satellite network resources.

[0011] In the satellite network resource scheduling method of the embodiment of this specification, the association relationship also includes fusion restriction conditions of different satellite network resources.

[0012] In the satellite network resource scheduling method of the embodiment of this specification, the fusion restriction condition includes: among the multiple satellite networks, an upper limit on the resource ratio of satellite network resources provided by the first satellite network to the second satellite network resources.

[0013] In the satellite network resource scheduling method of the embodiment of this specification, the graph neural network model includes a graph convolutional network model, and determining the target service type corresponding to the user intention based on the pre-trained neural network model includes:

[0014] Based on the input layer of the graph convolutional network model, inputting the user intention and the undirected graph into the graph convolutional layer of the graph convolutional network model;

[0015] Extracting mapping features between the user intention and different business types in the undirected graph based on the graph convolution layer of the graph convolution network model;

[0016] Based on the probability layer of the graph convolutional network model, converting the mapping features into mapping probabilities between the user intent and each business type;

[0017] Based on the filtering layer of the graph convolutional network model, filtering out mapping probabilities that do not conform to the association relationship from the mapping probabilities, and obtaining mapping probabilities that conform to the association relationship;

[0018] Based on the output layer of the graph convolutional network model, the business type corresponding to the maximum value of the mapping probability that meets the association relationship is determined as the target business type corresponding to the user intention.

[0019] In the satellite network resource scheduling method of the embodiment of this specification, the graph convolutional network model is obtained by pre-training an initial graph convolutional network model based on a deep reinforcement learning method. The deep reinforcement learning method uses the weighted sum of the total fusion delay and the fusion benefit-cost ratio as the reward function, and the fusion benefit-cost ratio is the ratio of the fusion benefit to the fusion cost.

[0020] In the satellite network resource scheduling method of the embodiment of this specification, the total fusion delay includes: the sum of the delays additionally added by allocating the fused satellite network resources.

[0021] In the satellite network resource scheduling method of the embodiment of this specification, the fusion benefit includes the total bandwidth resources of the fused satellite network; the fusion cost includes: the sum of the computing resources additionally consumed by allocating the fused satellite network resources, the sum of the bandwidth resources additionally consumed by allocating the fused satellite network resources, and the sum of the delay additionally increased by allocating the fused satellite network resources.

[0022] In the satellite network resource scheduling method of the embodiment of this specification, after allocating the satellite network resources corresponding to the target service type to the service request, the method further includes:

[0023] Based on the user intention data of service requests in a specified historical period, satellite network resource allocation data, and user evaluation information, a linear mapping relationship between user intention and service type is fitted.

[0024] In the satellite network resource scheduling method of the embodiment of this specification, the linear mapping relationship includes a linear mapping relationship between resource subgraphs and user intentions, and each resource subgraph represents a set of satellite network resources required for a corresponding service type.

[0025] In the satellite network resource scheduling method of the embodiment of this specification, after fitting the linear mapping relationship between user intention and service type, the method further includes:

[0026] Identifying a user intention for a new service request; the user intention represents the demand intention of the new service request for satellite network resources;

[0027] Determining whether a service type corresponding to the user intention exists in the linear mapping relationship;

[0028] In response to the presence of a service type corresponding to the user intention in the linear mapping relationship, determining the service type corresponding to the user intention as a target service type;

[0029] In response to the fact that the business type corresponding to the user intention does not exist in the linear mapping relationship, a target business type corresponding to the user intention is determined according to the neural network model.

[0030] In the satellite network resource scheduling method of the embodiment of this specification, the user intention is represented based on a service quality level tuple, and the service quality level tuple includes one or more different service quality indicators.

[0031] In the satellite network resource scheduling method of the embodiment of this specification, identifying the user intention of the service request includes:

[0032] Determine the task time range, task space range, and task keywords corresponding to the business request;

[0033] Determining the service quality requirements corresponding to the business request based on the task keywords;

[0034] Dividing the quality of service requirement into a sub-quality of service requirement sequence according to the task time range and the task space range;

[0035] Each sub-quality of service requirement in the quality of service requirement sequence is mapped to a corresponding quality of service level tuple.

[0036] In the satellite network resource scheduling method of the embodiment of this specification, the service quality level tuple includes part or all of the following service quality indicators:

[0037] Access guarantee indication, access point number indication, bandwidth indication, latency indication, packet error rate indication, energy consumption guarantee indication; wherein, the access guarantee indication includes access priority, access preemption indication and access preemption indication, and the energy consumption guarantee indication includes single-satellite energy consumption upper limit, terminal energy consumption upper limit and upper limit breakthrough indication.

[0038] In the satellite network resource scheduling method of the embodiment of this specification, the access priority is determined according to the user type, service type and / or service concurrency corresponding to the service request.

[0039] On the other hand, an embodiment of this specification further provides a satellite network resource scheduling device, including:

[0040] An identification module, configured to identify a user intention of a service request, wherein the user intention represents the demand intention of the service request for satellite network resources;

[0041] a determination module, configured to determine a target service type corresponding to the user intent based on a pre-trained neural network model, wherein the neural network model includes a nonlinear mapping relationship between user intent and service type;

[0042] The allocation module is used to allocate satellite network resources corresponding to the target service type to the service request.

[0043] On the other hand, an embodiment of this specification also provides a computer device, comprising at least one processor; and at least one memory storing instructions, which, when executed individually or collectively by the at least one processor, enable the computer device to perform the above method.

[0044] On the other hand, an embodiment of this specification further provides a computer storage medium storing instructions, which, when executed individually or collectively by at least one processor of a computer device, enable the computer device to perform the above method.

[0045] On the other hand, an embodiment of this specification further provides a computer program product, wherein when the instructions are executed individually or collectively by at least one processor of a computer device, the computer device executes the above method.

[0046] On the other hand, an embodiment of this specification further provides a chip, wherein the chip includes a circuit system, and the circuit system is configured to execute the above method.

[0047] It can be seen from the technical solutions provided in the above embodiments of this specification that, after identifying the demand intention of the service request for satellite network resources, the embodiments of this specification can input the demand intention into the pre-trained neural network model, thereby predicting the service type corresponding to the demand intention; because the neural network model has pre-learned the nonlinear mapping relationship between user intention and service type, it has a strong generalization ability, and can more effectively and adaptively allocate satellite network resources according to complex and changeable user intentions, which is conducive to adaptively meeting the user's complex satellite network service needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are only some of the embodiments described in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without inventive work. In the drawings:

[0049] Figure 1 A schematic diagram of an application environment for satellite network resource scheduling in some embodiments of this specification is shown;

[0050] Figure 2 A schematic diagram showing a plurality of satellite networks in an exemplary embodiment of the present specification is shown;

[0051] Figure 3 A flowchart of a satellite network resource scheduling method according to some embodiments of this specification is shown;

[0052] Figure 4 Shown Figure 3 Flowchart of identifying user intent for a business request in the method shown;

[0053] Figure 5 Shown Figure 3 Flowchart of the method for determining the target business type corresponding to user intent based on a pre-trained neural network model;

[0054] Figure 6 Shown Figure 3 A flowchart of the method for allocating satellite network resources corresponding to a target service type to a service request;

[0055] Figure 7 A schematic diagram showing a satellite network providing continuous services to user equipment for a long period of time in an exemplary embodiment of this specification;

[0056] Figure 8 It shows a structural block diagram of a satellite network resource scheduling device in some embodiments of this specification;

[0057] Figure 9 It shows a structural block diagram of a computer device in some embodiments of this specification.

[0058] [Description of Reference Numerals]

[0059] 10. User equipment;

[0060] 20. Resource scheduling server;

[0061] 30. Satellite network;

[0062] 81. Identification module;

[0063] 82. Determine the module;

[0064] 83. Allocation module;

[0065] 902. Computer equipment;

[0066] 904, processor;

[0067] 906. Memory;

[0068] 908, driving mechanism;

[0069] 910, input / output interface;

[0070] 912. Input devices;

[0071] 914. Output device;

[0072] 916. Presentation equipment;

[0073] 918. Graphical User Interface;

[0074] 920, network interface;

[0075] 922, communication link;

[0076] 924. Communication bus. DETAILED DESCRIPTION

[0077] To help those skilled in the art better understand the technical solutions in this specification, the following will provide a clear and complete description of the technical solutions in the embodiments of this specification, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. All other embodiments derived by those skilled in the art based on the embodiments in this specification without creative effort shall fall within the scope of protection of this specification.

[0078] It should be noted that in the embodiments of this specification, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved are all information and data authorized and agreed by the user and fully authorized by all parties, that is, the acquisition, transmission, storage, use, and processing of data in the technical solution of this application comply with the relevant provisions of national laws and regulations.

[0079] Currently, with the rapid development of satellite communication technology, satellite networks are demonstrating unique advantages in achieving global coverage and high-availability communications. In satellite networks, satellites have extensive coverage areas, filling the gaps left by ground-based base stations. Furthermore, some large-scale communication satellite networks can even provide global communication services. On this basis, satellite networks can support complex, long-duration, long-distance services worldwide, such as those for transcontinental aircraft and ocean shipping. Network assurance for large-scale satellite network services requires the scheduling of multiple satellites, multiple spectrums, and / or various types of satellite network resources. During long-duration satellite network services, user intent may change depending on user needs, current network conditions, and the environment. Mapping user intent into service requirements in long-duration, large-scale satellite network services and scheduling corresponding satellite network resources accordingly to ensure smooth service delivery has become a critical requirement in satellite networks.

[0080] Traditional technologies schedule satellite network resources based on fixed satellite network slices. These slices are pre-defined based on applications, industries, or users, with each slice allocated a fixed range of satellite network resources. However, in complex, large-scale, and long-term satellite network services, service types typically do not follow a fixed pattern, and resources often depend on each other. This makes it difficult to schedule satellite network resources based on fixed satellite network slices, making it difficult to meet users' complex satellite network service needs.

[0081] In view of this, the embodiments of this specification provide an improved satellite network resource scheduling solution to facilitate adaptively meeting users' complex satellite network service needs. Figure 1Schematic diagram of an application environment for satellite network resource scheduling in some embodiments of this specification is shown; the application environment includes user equipment (UE) 10, a resource scheduling server 20, and a satellite network 30. UE 10 can send a service request to resource scheduling server 20; resource scheduling server 20 can identify the user intent of the service request, which represents the service request's demand for satellite network resources; determine the target service type corresponding to the user intent based on a pre-trained neural network model, which includes a nonlinear mapping relationship between user intent and service type; and allocate satellite network resources corresponding to the target service type to the service request; wherein the satellite network resources are a portion of the satellite network 30's satellite network resources.

[0082] In some embodiments of this specification, the user device 10 may be a fixed terminal device or a mobile terminal device; wherein the mobile terminal device may include, but is not limited to, a smartphone, a tablet computer, a laptop computer, a smart wearable device, a vehicle (such as a vehicle, a ship, an airplane, etc.), etc. wherein the smart wearable device may include a smart bracelet, a smart watch, smart glasses, or a smart helmet, etc.

[0083] In some embodiments of the present specification, the resource scheduling server 20 can be a ground control center or a ground station of the satellite network 30, or a satellite with a resource scheduling function in the satellite network 30, or it can be software running in a ground control center or a satellite with a resource scheduling function to provide business logic for resource scheduling.

[0084] In some embodiments of the present specification, the satellite network 30 may be one or more satellite networks; wherein, multiple satellite networks (e.g. Figure 2 As shown in the figure, the satellite network can be a homogeneous or heterogeneous satellite network, thereby forming a multi-dimensional satellite network.

[0085] In some embodiments of this specification, satellite network resources may include satellite network resources of different levels and dimensions. For example, satellite network resources may include high earth orbit satellites, medium earth orbit satellites, low earth orbit satellites, cells, beams, frequency bands, time slots, physical resource blocks (PRBs), etc.

[0086] The embodiment of this specification provides a satellite network resource scheduling method, which can be applied to the resource scheduling service side mentioned above, with reference to Figure 3 As shown, in some embodiments of this specification, the satellite network resource scheduling method may include the following steps:

[0087] Step 301: Identify the user intention of a service request, where the user intention represents the demand intention of the service request for satellite network resources.

[0088] Step 302: Determine the target business type corresponding to the user intention based on a pre-trained neural network model, where the neural network model includes a nonlinear mapping relationship between user intention and business type.

[0089] Step 303: Allocate satellite network resources corresponding to the target service type to the service request.

[0090] In an embodiment of the present specification, after identifying the demand intention of a service request for satellite network resources, the demand intention can be input into a pre-trained neural network model to predict the service type corresponding to the demand intention; since the neural network model has pre-learned the nonlinear mapping relationship between user intention and service type, it has a strong generalization ability, and can thus more effectively and adaptively allocate satellite network resources according to complex and changeable user intentions, thereby facilitating adaptively meeting the user's complex satellite network service needs.

[0091] In some embodiments of this specification, a service request may be a service request sent by a user device via any suitable means, such as a short message. In some exemplary embodiments, the service request may include, for example, "requesting a voice call with ship A," "querying tomorrow's weather forecast for sea area B," "viewing real-time satellite images of the disaster situation at location C," "requesting access to a video conference with conference number xxxxx," and so on.

[0092] In some embodiments of this specification, user intent refers to the demand for satellite network resources for a service request, rather than the user's intended purpose for initiating the service request. For example, if the service request is "requesting a voice call with ship A," the corresponding user intent is the set of satellite network resources required to allocate or utilize in order to enable a normal voice call between the user and ship A.

[0093] In some embodiments of this specification, user intent can be represented using a Quality of Service (QoS) tuple, which can include one or more different Quality of Service (QoS) metrics. Representing user intent using QoS tuples not only facilitates vectorized representation of users' comprehensive demands for network service quality, but also serves as input for pre-trained neural network models.

[0094] In some embodiments of the present specification, the service quality level tuple may include, for example, part or all of an access guarantee indication, an access point quantity indication, a bandwidth indication, a delay indication, a packet error rate indication, an energy consumption guarantee indication, and the like.

[0095] The access guarantee indicator is used to indicate the requirement for guaranteed network access. In some embodiments of this specification, the access guarantee indicator may include an access priority, an access preemption indicator, and an access preemption indication. The access preemption indicator indicates whether preemption of other users' access resources is permitted when satellite network resources are insufficient. The access preemption indicator can take a value of 0 or 1. When the access preemption indicator takes a value of 1, preemption of other users' access resources is permitted; when the access preemption indicator takes a value of 0, preemption of other users' access resources is not permitted (or prohibited). The access preemption indication indicates whether preemption of one user's access resources is permitted. The access preemption indication can take a value of 0 or 1. When the access preemption indicator takes a value of 1, preemption of other users' access resources is permitted; when the access preemption indicator takes a value of 0, preemption of other users' access resources is not permitted (or prohibited).

[0096] Access priority is used to indicate the priority for different users and / or services accessing the satellite network. Priority settings can be customized by the user or the system. When network resources are insufficient, high-priority users and / or services can be prioritized based on access assurance indicators. This means that access priority can be determined based on the user type, service type, and / or concurrent service volume corresponding to the service request.

[0097] The access point quantity indicator is used to represent the minimum number requirement for access points; the bandwidth indicator is used to represent the minimum bandwidth requirement (Bps); the delay indicator is used to represent the maximum allowed delay, that is, the maximum tolerable delay (ms); and the packet error rate indicator is used to represent the maximum allowed packet error rate, that is, the maximum tolerable packet error rate.

[0098] Energy consumption guarantee indicators represent the requirements for service energy consumption. Energy consumption guarantee indicators can include single-star energy consumption caps, terminal energy consumption caps, and limit-breaking indicators. The limit-breaking indicator indicates whether service energy consumption exceeding the energy consumption cap is acceptable. The limit-breaking indicator can be represented by 0 or 1.

[0099] It should be understood that the service quality level tuple adopts the access guarantee indication, access point number indication, bandwidth indication, delay indication, packet error rate indication and energy consumption guarantee indication, which are only exemplary examples of some embodiments of this specification; in other embodiments of this specification, the service quality level tuple can also be represented by more or fewer QoS indicators. Therefore, this specification does not limit which QoS indicators are used in the service quality level tuple, and specific selection can be made as needed.

[0100] refer to Figure 4 As shown, in some embodiments of this specification, taking the user intent represented by the service quality level tuple as an example, identifying the user intent of a service request may include the following steps:

[0101] Step 401: Determine the task time range, task space range, and task keywords corresponding to the business request.

[0102] In some embodiments of the present specification, tasks may be mined from time and space dimensions to obtain the evolution characteristics of the task in the time dimension and the movement characteristics of the task in the space dimension.

[0103] In some embodiments of the present specification, the evolution characteristic of the task time dimension refers to the task time range from the start time to the end time of the task. The movement characteristic of the task space dimension refers to the task space range from the start position to the end position of the task. For example, in an exemplary embodiment, the business request is "request to provide navigation for a cargo ship departing from Port X along route M to Port Y, with a departure time of November 1, 2020 and an estimated arrival time of November 15, 2020". It can be extracted that the task time range corresponding to the business request is November 1, 2020 to November 15, 2020, and the task space range corresponding to the business request can be extracted as the position of Port X to the position of Port Y in route M.

[0104] In some embodiments of this specification, task keywords refer to keywords that can characterize QoS, or keywords related to characterizing QoS. For example, in one exemplary embodiment, the service request is "Requesting access to the video conference with conference number xxxxx, requiring a priority of 1, an access preemption indicator of 1, an access preemption indicator of 0, and a maximum latency of 200ms." Then, "priority of 1," "access preemption indicator of 1," "access preemption indicator of 0," and "latency of 200ms" are all task keywords.

[0105] In other embodiments of this specification, a service request may not indicate or not indicate all QoS indicators. In such cases, the subject intent (i.e., the primary purpose) of the service request can be extracted to subsequently map and generate a quality of service level tuple based on the typical QoS indicators corresponding to the subject intent (typical QoS indicators in satellite networks). For example, in an exemplary embodiment, if the service request is "Request to access video conference with conference number xxxxx," "video conference," "video conference" can be extracted as the subject intent, so that the quality of service level tuple can be subsequently mapped and generated based on the typical QoS indicators corresponding to "video conference."

[0106] In some embodiments of this specification, the task time range, task space range, and task keywords corresponding to the business request may be determined based on technical means such as keyword matching, regular expressions, or natural language processing.

[0107] Step 402: Determine the service quality requirement corresponding to the business request according to the task keyword.

[0108] In some embodiments of the present application, in scenarios where a task keyword can quantitatively represent QoS, the task keyword is the service quality requirement corresponding to the service request. For example, "200ms delay" in the above embodiment is the service quality requirement corresponding to the service request.

[0109] In some embodiments of the present application, in scenarios where the task keywords do not quantitatively characterize QoS, service quality requirement indicators corresponding to typical services can be pre-constructed. For example, service quality indicators such as bandwidth corresponding to positioning and navigation services, service quality indicators such as bandwidth corresponding to voice call services, service quality indicators such as bandwidth corresponding to video call services, and service quality indicators such as bandwidth corresponding to on-site monitoring services. On this basis, the corresponding service quality requirements can be matched according to the task keywords extracted from the service request. For example, if the task keyword of a service request is voice call, and the bandwidth requirement corresponding to the voice call service is 10MB, it can be determined that the service quality requirement corresponding to the service request is 10MB bandwidth.

[0110] Step 403: Divide the QoS requirement into sub-QoS requirement sequences according to the task time range and the task space range.

[0111] In some embodiments of the present application, to improve service quality, the entire service quality requirement can be divided into multiple sub-service quality requirements according to the task time range and task space range, forming a sub-service quality requirement sequence. Each sub-service quality requirement in the sub-service quality requirement sequence includes: a service period, a tracking area code (TAC), and a service quality indicator. The service period is the time segment divided by the task time range, and the tracking area code is the spatial area divided by the task space range.

[0112] Step 404: Map each sub-quality of service requirement in the quality of service requirement sequence to a corresponding quality of service level tuple.

[0113] In some embodiments of this specification, taking the example of a QoS level tuple employing an access guarantee indicator, an access point quantity indicator, a bandwidth indicator, a latency indicator, a packet error rate indicator, and an energy consumption guarantee indicator, a mapping relationship between each QoS indicator in the QoS level tuple can be pre-established to map a corresponding QoS indicator to each sub-QoS requirement, and then combine them into a QoS level tuple. This facilitates flexible adjustment of QoS indicators.

[0114] In some embodiments of this specification, the pre-trained neural network model is a pre-trained artificial neural network (ANN) model. Because neural network models have nonlinear adaptive information processing capabilities, they can be trained to learn the nonlinear mapping relationship between user intent and service types. In some embodiments of this specification, because satellite networks can be represented by graph structures, using a graph neural network model (e.g., a graph convolutional network) as the pre-trained neural network model not only enables the model to learn the nonlinear mapping relationship between user intent and service types, but also makes it more intuitive and easier to understand.

[0115] In some embodiments of this specification, the resource topology of multiple satellite networks can be represented using an undirected graph. In this undirected graph, different nodes represent different satellite network resources, and the edges connecting the nodes represent the relationships between different satellite network resources. In some embodiments of this specification, the relationships can include the integration costs of different satellite network resources (e.g., the additional latency, bandwidth, and computational costs incurred due to network integration). In other embodiments of this specification, the relationships can include integration costs and integration constraints. Integration constraints can include, for example, resource sharing ratios. Specifically, in a scenario where multiple satellite networks are integrated, a first satellite network can provide an upper limit on the proportion of satellite network resources that a second satellite network can provide.

[0116] For example, in some embodiments of this specification, satellite network resources can be constructed as an undirected weighted graph. ,in is a node in the graph, representing a collection of various network resources. The edges in the graph represent the fusion cost set between various satellite network resources. Represents connected nodes and The time cost can be specifically expressed as: when the i-th type of satellite network resources Insufficient to support the business, additional j-type satellite network resources were allocated When the two types of network resources aggregate delay, Indicates the additional computing resources consumed when allocating network resources. In other embodiments of this specification, when considering the fusion constraint, the undirected weighted graph can also add the i-th type of satellite network resources. When the jth type of satellite network resources are insufficient to support the business Can be the i-th type of satellite network resources The upper limit of the resource ratio provided, that is, What percentage of your own resources can be temporarily provided to use.

[0117] In some embodiments of this specification, taking the graph convolutional network model in the graph neural network model as an example, the model training process includes the following steps:

[0118] (1) Obtain a dataset and an undirected graph for representing satellite network resources.

[0119] The data set refers to a training set obtained by summarizing and preprocessing the historical service requests of the satellite network, the network resources allocated for the historical service requests, and the historical service completion status (service quality, customer satisfaction, etc.).

[0120] (2) Taking the dataset and the undirected graph as input, an initial graph convolutional network model is trained based on a deep reinforcement learning method until a graph convolutional network model that meets expectations (e.g., preset model evaluation indicators) is obtained.

[0121] In some embodiments of the present specification, during the model training process, training can be performed in the order of different user priorities or service priorities, so that users with different priorities or users with different priority services can obtain the required network resources in an orderly manner.

[0122] In some embodiments of this specification, a graph convolutional network model may include an input layer, a graph convolutional layer, a probability layer, a filtering layer, and an output layer. During model training, user intent, the state and features of an undirected graph (e.g., network resource allocation) are input. The input layer transmits this state information to the graph convolutional layer. The graph convolutional layer extracts and learns useful features from the input, which contribute to solving specific tasks. For satellite network topology graph data, the graph convolutional layer can gradually extract information from the graph by considering the relationships between nodes and their neighbors at each layer to obtain network resource vectors, thereby enabling the probability layer to generate user intent mapping probabilities that better align with network resource relationships. The probability layer is primarily responsible for calculating the probability of mapping user intent to service types based on service completion status. The filtering layer removes nodes that do not meet the constraints, and the output layer outputs the service type with the highest probability of being mapped to the current user intent.

[0123] In some embodiments of the present specification, during model training, a deep reinforcement learning method uses the weighted sum of the total fusion latency and the fusion benefit-cost ratio as a reward function. The total fusion latency refers to the sum of the additional latency incurred by allocating the fused satellite network resources; the fusion benefit-cost ratio is the ratio of the fusion benefit to the fusion cost; the fusion benefit includes the total bandwidth resources of the fused satellite network; and the fusion cost includes the sum of the additional computing resources consumed by allocating the fused satellite network resources, the sum of the additional bandwidth resources consumed by allocating the fused satellite network resources, and the sum of the additional latency incurred by allocating the fused satellite network resources. This allows the model to comprehensively consider the additional costs of network resource fusion, such as latency, node computing resource consumption, and link bandwidth resource consumption, and balance the fusion benefit (meeting user service needs). This ensures that subsequent model application predictions are based on a balance between fusion benefit and fusion cost. Consequently, resource allocation can achieve low latency and high resource utilization.

[0124] In some embodiments of this specification, the total fusion delay is calculated according to the formula Calculated.

[0125] in, For the integrated satellite network The total delay (i.e., the integrated total delay), The additional delay cost for allocating the i-th type of satellite network resources, is the integration cost of the satellite network.

[0126] In some embodiments of this specification, the fusion benefit cost ratio can be calculated according to the formula Calculated;

[0127] in, is the fusion benefit-cost ratio, For satellite network revenue, is the satellite network cost, is the revenue conversion rate of satellite network bandwidth, is the i-th type of satellite network resources, A collection of satellite network resources. is the bandwidth function of the i-th type of satellite network resources, is the cost coefficient of computing resources, The additional computing resources consumed to allocate the i-th type of satellite network resources, is the cost coefficient of satellite network resources, The additional bandwidth resources consumed to allocate the i-th category satellite network resources, is the cost coefficient of network delay, The additional delay cost for allocating the i-th type of satellite network resources, is the integration cost of the satellite network.

[0128] refer to Figure 5 As shown, in some embodiments of this specification, determining the target service type corresponding to the user intention based on the pre-trained neural network model may include the following steps:

[0129] Step 501: Based on the input layer of the graph convolutional network model, the user intention and the undirected graph are input into the graph convolutional layer of the graph convolutional network model.

[0130] Among them, user intention is a user intention vector represented by a service quality level tuple; the user intention vector and the node state features in the undirected graph together form the initial feature matrix of the node; the adjacency matrix of the undirected graph forms another input of the graph convolutional network model; among them, the adjacency matrix is used to describe the connection relationship between each node in the undirected graph.

[0131] Step 502: Based on the graph convolution layer of the graph convolutional network model, extract mapping features between the user intention and different business types in the undirected graph.

[0132] In some embodiments of the present specification, the graph convolution layer aggregates the feature vector of each node with the feature vectors of adjacent nodes in a nonlinear manner (convolution operation) to extract mapping features between user intent and different business types in an undirected graph.

[0133] Step 503: Based on the probability layer of the graph convolutional network model, the mapping features are converted into mapping probabilities between the user intention and each business type.

[0134] Step 504: Based on the filtering layer of the graph convolutional network model, the mapping probabilities that do not conform to the association relationship are filtered out to obtain the mapping probabilities that conform to the association relationship.

[0135] Step 505: Based on the output layer of the graph convolutional network model, the business type corresponding to the maximum value of the mapping probability that meets the association relationship is determined as the target business type corresponding to the user intention.

[0136] In an embodiment of the present specification, in a scenario where the target business type corresponding to the user intention is determined based on a pre-trained neural network model, the business type is a personalized, non-standardized business type customized by the neural network model based on the user intention, rather than selecting one from multiple known, determined business types.

[0137] In some embodiments of this specification, allocating satellite network resources corresponding to the target service type to the service request means allocating satellite network resources corresponding to the target service type to process the service request to meet the user's service needs. Of course, when actually allocating satellite network resources, the location information of the UE when initiating the service request may also be taken into consideration.

[0138] In addition, in satellite network services, the user intention of the same user may change at any time according to user needs, current network conditions, environment (such as UE location), etc. Therefore, when determining the target service type corresponding to the user intention based on the pre-trained neural network model, a time series of target service types {S1, S2, S3, S4, ...} for the same user can be formed according to the time axis.

[0139] For example, in Figure 7 In the exemplary embodiment shown, the user equipment is a ship (with satellite communication equipment installed on board). When the ship arrives at position 1 and issues a service request for "requesting a voice call with ship A", the resource scheduling server can Figure 3 The satellite network resource scheduling method shown in FIG. 1 allocates satellite network resources S11 for the service request. When the ship sends a service request of "querying the weather forecast of sea area B tomorrow" at position 2, the resource scheduling server can allocate satellite network resources S11 according to the satellite network resource scheduling method shown in FIG. Figure 3 The satellite network resource scheduling method shown in FIG. 1 allocates satellite network resources for the service request S12. When the ship arrives at position 3 and issues a service request for "requesting to access the video conference with conference number xxxxx", the resource scheduling server can allocate satellite network resources for the service request S12. Figure 3 The satellite network resource scheduling method shown allocates satellite network resources to the service request S13. Therefore, for the user equipment, a target service type time sequence {S11, S12, S13} can be formed.

[0140] refer to Figure 6 As shown, in some other embodiments of this specification, the satellite network resource scheduling method may include the following steps:

[0141] Step 601: Identify the user intention of the service request.

[0142] Step 602: Determine the target business type corresponding to the user intention based on a pre-trained neural network model, where the neural network model includes a nonlinear mapping relationship between user intention and business type.

[0143] That is, the user intention and the state and characteristics of the undirected graph of satellite network resources are input into a pre-trained neural network model, thereby obtaining the target service type corresponding to the user intention of the service request.

[0144] Step 603: Allocate satellite network resources corresponding to the target service type to the service request.

[0145] Step 604: Fit a linear mapping relationship between user intent and service type based on user intent data of service requests in a specified historical period, satellite network resource allocation data, and user evaluation information.

[0146] In some embodiments of this specification, a linear mapping relationship includes a linear mapping relationship between resource subgraphs and user intent, where each resource subgraph represents the set of satellite network resources required for a corresponding service type. This makes the representation of the set of satellite network resources required for a service type more flexible, adaptable, and easy to understand. Unlike the nonlinear mapping relationship described above, in service type matching based on a linear mapping relationship, a service type is selected from multiple known, standard service types.

[0147] Step 605: Identify the user intention of the new service request.

[0148] Step 606: Determine whether the service type corresponding to the user's intention exists in the linear mapping relationship. If the service type corresponding to the user's intention exists in the linear mapping relationship, execute step 607a; otherwise, execute step 607b.

[0149] Step 607a: Determine the service type corresponding to the user intention as the target service type.

[0150] Step 607b: Determine the target business type corresponding to the user intention based on the neural network model.

[0151] Step 608: Allocate satellite network resources corresponding to the target service type to the new service request.

[0152] Since the processing efficiency of business type matching based on linear mapping relationship is higher than that of determining the target business type corresponding to the user intention based on the neural network model; Figure 6 The illustrated embodiment can improve the efficiency of satellite network resource scheduling, thereby increasing the response speed to user service requests and improving user experience.

[0153] Although the process flows described above include multiple operations occurring in a particular order, it should be understood that these processes may include more or fewer operations, and that the operations may be performed sequentially or in parallel (eg, using parallel processors or a multi-threaded environment).

[0154] Corresponding to the above-mentioned satellite network resource scheduling method, the embodiment of this specification also provides a satellite network resource scheduling device, which can be configured on the above-mentioned resource scheduling server, referring to Figure 8 As shown, in some embodiments of this specification, satellite network resource scheduling may include:

[0155] An identification module 81 is configured to identify a user intention of a service request, wherein the user intention represents the demand intention of the service request for satellite network resources;

[0156] a determination module 82 for determining a target service type corresponding to the user intention based on a pre-trained neural network model, wherein the neural network model includes a nonlinear mapping relationship between user intention and service type;

[0157] The allocation module 83 is configured to allocate satellite network resources corresponding to the target service type to the service request.

[0158] For the convenience of description, the above devices are described as being divided into various units according to their functions. Of course, when implementing this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0159] An embodiment of this specification further provides a chip, which includes a circuit system configured to perform the above-mentioned satellite network resource scheduling.

[0160] The embodiment of this specification also provides a computer device. Figure 9 As shown, in some embodiments of this specification, the computer device 902 may include one or more processors 904, such as one or more central processing units (CPUs) or graphics processing units (GPUs). Each processing unit may implement one or more hardware threads. The computer device 902 may also include any memory 906 for storing any type of information, such as code, settings, data, etc. In one specific embodiment, the memory 906 may contain a computer program executable by the processor 904. When executed by the processor 904, the computer program may execute instructions of the satellite network resource scheduling method described in any of the above embodiments. For example, and without limitation, the memory 906 may include any one or more combinations of the following: any type of RAM, any type of ROM, a flash memory device, a hard disk, an optical disk, etc. More generally, any memory may use any technology to store information. Furthermore, any memory may provide volatile or non-volatile retention of information. Furthermore, any memory may represent a fixed or removable component of the computer device 902. In one embodiment, when the processor 904 executes the associated instructions stored in any memory or combination of memories, the computer device 902 may perform any operation of the associated instructions. The computer device 902 also includes one or more drive mechanisms 908 for interacting with any storage, such as a hard disk drive mechanism, an optical disk drive mechanism, and the like.

[0161] Computer device 902 may also include an input / output interface 910 (I / O) for receiving various inputs (via input devices 912) and providing various outputs (via output devices 914). A specific output mechanism may include a presentation device 916 and an associated graphical user interface 918 (GUI). In other embodiments, input / output interface 910 (I / O), input devices 912, and output devices 914 may not be included, and the computer device 902 may simply function as a computer device on a network. Computer device 902 may also include one or more network interfaces 920 for exchanging data with other devices via one or more communication links 922. One or more communication buses 924 couple the components described above.

[0162] The communication link 922 can be implemented in any manner, for example, via a local area network, a wide area network (e.g., the Internet), a point-to-point connection, etc., or any combination thereof. The communication link 922 can include any combination of hardwired links, wireless links, routers, gateway functions, name servers, etc., governed by any protocol or combination of protocols.

[0163] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), computer-readable storage media, and computer program products of some embodiments of this specification. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processor to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processor generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0164] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processor to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, the instruction device being implemented in the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0165] These computer program instructions can also be loaded onto a computer or other programmable data processor so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0166] In a typical configuration, a computer device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0167] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0168] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can be implemented using any method or technology to store information. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computer device. As defined in this specification, computer-readable media does not include transitory computer-readable media such as modulated data signals and carrier waves.

[0169] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Thus, the embodiments of this specification may take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware. Furthermore, the embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROMs, optical storage, etc.) containing computer-usable program code.

[0170] Embodiments of this specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. Embodiments of this specification may also be practiced in distributed computing environments where tasks are performed by remote processors connected via a communications network. In distributed computing environments, program modules may be located in local and remote computer storage media, including storage devices.

[0171] It should also be understood that in the embodiments of this specification, the term "and / or" is merely a description of the association relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone. Furthermore, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0172] The various embodiments in this specification are described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the system embodiments are generally similar to the method embodiments, so the description is relatively simple. For relevant parts, refer to the description of the method embodiments.

[0173] In the description of this specification, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the embodiments of this specification. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0174] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A satellite network resource scheduling method, characterized in that: include: Identifying a user intention of a service request, where the user intention represents the demand intention of the service request for satellite network resources; The satellite network resources include satellite network resources of multiple satellite networks; Determining a target service type corresponding to the user intent based on a pre-trained neural network model, wherein the neural network model includes a nonlinear mapping relationship between the user intent and the service type; the neural network model includes a graph neural network model, and the resource topology structure of the multiple satellite networks is represented based on an undirected graph. In the undirected graph, different nodes represent different satellite network resources, and edges connecting the nodes represent associations between the different satellite network resources; the associations include fusion costs and fusion constraints of the different satellite network resources; Allocate satellite network resources corresponding to the target service type to the service request.

2. The satellite network resource scheduling method according to claim 1, wherein: The plurality of satellite networks include a low-orbit satellite network, a medium-orbit satellite network and / or a high-orbit satellite network.

3. The satellite network resource scheduling method according to claim 1, wherein: The fusion restriction condition includes: among the multiple satellite networks, an upper limit on the resource ratio of satellite network resources provided by the first satellite network to the second satellite network resources.

4. The satellite network resource scheduling method according to claim 1, wherein: The graph neural network model includes a graph convolutional network model; and determining the target service type corresponding to the user intention based on the pre-trained neural network model includes: Based on the input layer of the graph convolutional network model, inputting the user intention and the undirected graph into the graph convolutional layer of the graph convolutional network model; Extracting mapping features between the user intention and different business types in the undirected graph based on the graph convolution layer of the graph convolution network model; Based on the probability layer of the graph convolutional network model, converting the mapping features into mapping probabilities between the user intent and each business type; Based on the filtering layer of the graph convolutional network model, filtering out mapping probabilities that do not conform to the association relationship from the mapping probabilities, and obtaining mapping probabilities that conform to the association relationship; Based on the output layer of the graph convolutional network model, the business type corresponding to the maximum value of the mapping probability that meets the association relationship is determined as the target business type corresponding to the user intention.

5. The satellite network resource scheduling method according to claim 4, wherein: The graph convolutional network model is obtained by pre-training an initial graph convolutional network model based on a deep reinforcement learning method. The deep reinforcement learning method uses the weighted sum of the total fusion delay and the fusion benefit-cost ratio as a reward function, and the fusion benefit-cost ratio is the ratio of the fusion benefit to the fusion cost.

6. The satellite network resource scheduling method according to claim 5, wherein: The total fusion delay includes: the sum of the delays additionally added by allocating the fused satellite network resources.

7. The satellite network resource scheduling method according to claim 5, wherein: The fusion benefit includes the total bandwidth resources of the integrated satellite network; the fusion cost includes: the sum of the computing resources additionally consumed by allocating the integrated satellite network resources, the sum of the bandwidth resources additionally consumed by allocating the integrated satellite network resources, and the sum of the delay additionally increased by allocating the integrated satellite network resources.

8. The satellite network resource scheduling method according to claim 1, wherein: After allocating the satellite network resources corresponding to the target service type to the service request, the method further includes: Based on the user intention data of service requests in a specified historical period, satellite network resource allocation data, and user evaluation information, a linear mapping relationship between user intention and service type is fitted.

9. The satellite network resource scheduling method according to claim 8, wherein: The linear mapping relationship includes a linear mapping relationship between resource subgraphs and user intentions, and each resource subgraph represents a set of satellite network resources required for a corresponding service type.

10. The satellite network resource scheduling method according to claim 8, wherein: After fitting the linear mapping relationship between user intent and business type, it also includes: Identifying a user intention for a new service request; the user intention represents the demand intention of the new service request for satellite network resources; Determining whether a service type corresponding to the user intention exists in the linear mapping relationship; In response to the presence of a service type corresponding to the user intention in the linear mapping relationship, determining the service type corresponding to the user intention as a target service type; In response to the fact that the business type corresponding to the user intention does not exist in the linear mapping relationship, a target business type corresponding to the user intention is determined according to the neural network model.

11. The satellite network resource scheduling method according to claim 1, wherein: The user intention is represented based on a quality of service level tuple, where the quality of service level tuple includes one or more different quality of service indicators.

12. The satellite network resource scheduling method according to claim 11, wherein: The identifying of the user intent of the service request includes: Determine the task time range, task space range, and task keywords corresponding to the business request; Determining the service quality requirements corresponding to the business request based on the task keywords; Dividing the quality of service requirement into a sub-quality of service requirement sequence according to the task time range and the task space range; Each sub-quality of service requirement in the quality of service requirement sequence is mapped to a corresponding quality of service level tuple.

13. The satellite network resource scheduling method according to claim 11, wherein: The service quality level tuple includes some or all of the following service quality indicators: Access guarantee indication, access point number indication, bandwidth indication, latency indication, packet error rate indication, energy consumption guarantee indication; wherein, the access guarantee indication includes access priority, access preemption indication and access preemption indication, and the energy consumption guarantee indication includes single-satellite energy consumption upper limit, terminal energy consumption upper limit and upper limit breakthrough indication.

14. The satellite network resource scheduling method according to claim 13, wherein: The access priority is determined according to the user type, service type and / or service concurrency corresponding to the service request.

15. A satellite network resource scheduling device, characterized in that: include: An identification module, configured to identify a user intention of a service request, wherein the user intention represents the demand intention of the service request for satellite network resources; The satellite network resources include satellite network resources of multiple satellite networks; A determination module is configured to determine a target service type corresponding to the user intent based on a pre-trained neural network model, wherein the neural network model includes a nonlinear mapping relationship between user intent and service type; the neural network model includes a graph neural network model, and the resource topology structure of the multiple satellite networks is represented based on an undirected graph. In the undirected graph, different nodes represent different satellite network resources, and edges connecting the nodes represent associations between the different satellite network resources; the associations include fusion costs and fusion constraints of the different satellite network resources; The allocation module is used to allocate satellite network resources corresponding to the target service type to the service request.

16. A computer device, characterized in that: include: at least one processor; as well as At least one memory having instructions stored thereon, which, when executed individually or collectively by the at least one processor, cause the computer device to perform the method according to any one of claims 1 to 14.

17. A computer storage medium storing instructions, characterized in that: When the instructions are executed individually or collectively by at least one processor of a computer device, the instructions cause the computer device to perform the method according to any one of claims 1 to 14.

18. A computer program product comprising instructions, characterized in that When the instructions are executed individually or collectively by at least one processor of a computer device, the instructions cause the computer device to perform the method according to any one of claims 1 to 14.

19. A chip, characterized in that: The chip comprises circuitry configured to perform the method according to any one of claims 1 to 14.

Images