A ground sensor station network distribution method and related device

Abstract

This application discloses a method and related apparatus for allocating ground sensor station networks, relating to the field of satellite operation control technology. The method includes: acquiring target operational condition data corresponding to a satellite communication mission; generating target prompt words based on a preset prompt word template and the target operational condition data; processing the target prompt words using a preset large language model to generate behavior tree description text; parsing the behavior tree description text using a preset text model mapping rule set to generate a tree behavior model; and allocating ground sensor station networks based on the tree behavior model. This application's solution, through the synergy of prompt word templates, a large language model, and a mapping rule set, can intelligently infer behavior tree description text from the target operational condition data corresponding to a satellite communication mission, and instantly convert the behavior tree description text into an executable tree behavior model, achieving high-efficiency, high-accuracy, and unmanned intelligent allocation of ground sensor station networks.

Description

Technical Field

[0001] This application relates to the field of satellite operation and control technology, and in particular to a ground sensor station network allocation method and related apparatus. Background Technology

[0002] Ground-based sensor network is the core ground infrastructure supporting satellite orbit tracking and control, satellite-to-ground data backhaul, and two-way satellite communication. It consists of multiple distributed ground-based sensor base stations, communication antennas, signal processing units, and timing control units. Its core function is to perform one-to-one matching, time slot allocation, frequency band reuse, and resource avoidance between single or multiple satellites and ground base stations based on the real-time position, communication frequency band, communication duration, and mission priority requirements of the on-orbit satellites.

[0003] Currently, the allocation of ground sensor station networks mainly relies on two traditional schemes: One method is manual allocation, where maintenance personnel manually match and set communication frequency bands, start and end times, and antenna pointing parameters for each satellite based on satellite orbit forecasts and ground station availability. This method is inefficient and unsuitable for high-concurrency, multi-satellite communication scenarios. A single round of scheduling can take 10 to 30 minutes, failing to meet the short-window communication requirements for rapid satellite transit.

[0004] Another approach is the fixed-rule, pre-set allocation method, which involves pre-entering static scheduling rules and periodically executing assigned tasks according to a pre-set script. This is currently a common solution in commercial spaceflight. However, this method lacks flexibility and struggles to cope with dynamic changes such as high-speed orbital motion, sudden maneuvers, and frequency band switching of low-Earth orbit satellites. It also cannot be compatible with heterogeneous communication systems of different satellite models, resulting in high idle rates of ground station networks and a serious waste of spectrum and hardware resources.

[0005] Existing ground sensor network allocation schemes are difficult to simultaneously meet core requirements such as dynamic adaptability, scheduling efficiency, and unmanned operation. How to provide a ground sensor network allocation method with strong dynamic adaptability and efficient resource utilization has become one of the technical problems that urgently need to be solved in the field of satellite operation and control technology. Summary of the Invention

[0006] Based on the above problems, this application provides a ground sensor station network allocation method, which can solve the problems of low efficiency and poor flexibility of fixed rules in the prior art, and realize efficient, dynamic and unmanned intelligent allocation of ground sensor station networks.

[0007] The embodiments of this application disclose the following technical solutions: The first aspect of this application provides a method for allocating a ground sensor station network, including: Acquire target operating condition data corresponding to the satellite communication mission; the target operating condition data includes communication data from the satellite and operational data from the ground sensor network; Based on the preset prompt word template and the target operating condition data, a target prompt word is generated; the prompt word template is used to structure and arrange the target operating condition data, set the priority of the constraints in the satellite communication mission, and perform fault tolerance processing for missing or abnormal data in the target operating condition data. The target prompt words are processed by a pre-defined large language model to generate a behavior tree description text. The behavior tree description text includes node identifiers, node types, node determination conditions, execution parameters corresponding to the nodes, and a set of child node identifiers corresponding to each parent node identifier. The behavior tree description text is parsed using a preset text model mapping rule set to generate a tree behavior model. The text model mapping rule set is used to map semantic keywords in the behavior tree description text to nodes, construct hierarchical relationships between nodes, and inject the execution parameters and node determination conditions of the nodes in the behavior tree description text into the attribute fields of the nodes. Based on the tree behavior model, the ground sensor network is allocated.

[0008] In one optional approach, the preset prompt word template includes a four-layer hierarchical structure, a constraint priority encoding mechanism, and an abnormal operating condition fault tolerance mechanism. The four-layer hierarchical structure includes a task layer, a condition layer, a constraint layer, and an output format layer. The task layer defines the objectives of the satellite communication mission. The condition layer organizes the condition data corresponding to the satellite communication mission. The constraint layer defines the priority of the constraints of the satellite communication mission. The output format layer defines the text format output by the large language model. The constraints include physical constraints and objective constraints. The objective constraints include maximizing the link signal-to-noise ratio and / or minimizing the communication downtime. The constraint priority coding mechanism includes a first priority, a second priority, and a third priority; the first priority corresponds to the physical constraint; the second priority corresponds to the target constraint; and the third priority corresponds to the constraint related to operation and maintenance costs in the satellite communication mission; the priorities are ranked from high to low as the first priority, the second priority, and the third priority. The abnormal operating condition fault tolerance mechanism includes a data completion mechanism, a parameter degradation mechanism, and a task degradation mechanism. The data completion mechanism is used to complete the data based on the historical average value of the parameter when the first type of parameter is missing. The parameter degradation mechanism is used to remove the missing parameter from the operating condition layer when the second type of parameter is missing. The task degradation mechanism is used to reduce the priority of the task corresponding to the third type of parameter when the third type of parameter is missing. The importance of the parameters, from high to low, is the third type of parameter, the second type of parameter, and the first type of parameter.

[0009] In one optional approach, generating the target prompt word based on a preset prompt word template and the target operating condition data includes: The target operating condition data is preprocessed to generate a standardized operating condition dataset; The metadata in the task layer is used as the content of the task layer; the metadata includes task ID, task type, priority, and time window; Based on the abnormal operating condition fault tolerance processing mechanism, the parameters in the standardized operating condition dataset are processed to generate updated operating condition data, and the updated operating condition data is used as the content of the operating condition layer. Based on the constraint priority encoding mechanism, the physical constraints, the target constraints, and the constraints related to operation and maintenance costs in the satellite communication mission, the content of the constraint layer is generated. According to the text format determined by the output format layer, the content of the task layer, the content of the working condition layer, and the content of the constraint layer are assembled to generate the initial prompt words; The initial prompt word is validated to obtain the target prompt word.

[0010] In one optional approach, the step of validating the initial prompt word to obtain the target prompt word includes: Verify whether the initial prompt word meets the text format specified by the output format layer; If the initial prompt word meets the text format, then a constraint compliance check is performed on the initial prompt word; the constraint compliance check is used to verify whether the decision logic in the initial prompt word meets the physical constraints; If the initial prompt word meets the text format and passes the constraint compliance check, then the initial prompt word is used as the target prompt word; If the initial prompt word does not meet the text format, or the initial prompt word fails the constraint compliance check, the initial prompt word is modified until the modified initial prompt word meets the text format and passes the constraint compliance check, thus obtaining the target prompt word.

[0011] In one alternative approach, the text model mapping rule set includes a behavior tree node library, a mapping rule library, topology generation rules, and node parameter injection rules; The behavior tree node library contains multiple types of nodes; these types of nodes include root nodes, condition nodes, control nodes, execution nodes, and decorator nodes. The mapping rule base includes a preset semantic keyword and the correspondence between each node in the behavior tree node library; The topology generation rules include that the root node is unique, the control node includes any number of child nodes, and the condition node and the execution node do not include child nodes. The node parameter injection rules include writing the execution parameters and decision parameters into the attribute fields of the corresponding node based on the parameter name matching principle.

[0012] In one optional approach, the step of parsing the behavior tree description text using a preset text model mapping rule set to generate a tree behavior model includes: Semantic parsing is performed on the node type, node determination conditions, and execution parameters of each node in the behavior tree description text to generate semantic keywords and parameter values ​​corresponding to each node; Based on the mapping rule library, the semantic keywords corresponding to the nodes are mapped to target nodes in the behavior tree node library, and according to the node parameter injection rules, the parameter values ​​and the judgment conditions are written into the attribute fields of the target nodes to obtain the behavior tree nodes to be assembled. Based on the node identifiers in the behavior tree description text and the set of child node identifiers corresponding to the parent node identifiers, the behavior tree nodes to be assembled are hierarchically connected according to the topology generation rules to obtain the initial tree behavior model. Once the initial tree behavior model passes the logical consistency check, the initial tree behavior model is used as the tree behavior model; the logical consistency check is used to verify the topology legality, branch integrity, node reachability, parameter legality, and compliance of physical constraints.

[0013] In one alternative approach, after allocating the ground sensor network based on the tree behavior model, the method further includes: The ground sensor network collects feedback data during the execution of the tree behavior model; the feedback data includes link communication quality, equipment operating status, and task completion rate. The feedback data is input into the large language model so that the large language model can correct the tree behavior model based on the feedback data.

[0014] A second aspect of this application provides a ground sensor station network distribution device, comprising: The data acquisition module is used to acquire target operating condition data corresponding to the satellite communication mission; the target operating condition data includes communication data from the satellite and operational data from the ground sensor network. The prompt word generation module is used to generate target prompt words based on a preset prompt word template and the target operating condition data; the prompt word template is used to structure and arrange the target operating condition data, set the priority of the constraints in the satellite communication mission, and perform fault tolerance processing for missing or abnormal data in the target operating condition data; The description text generation module is used to process the target prompt words through a preset large language model to generate behavior tree description text; the behavior tree description text includes node identifier, node type, node judgment condition, execution parameters corresponding to the node, and a set of child node identifiers corresponding to each parent node identifier; The behavior model construction module is used to parse the behavior tree description text through a preset text model mapping rule set to generate a tree behavior model; the text model mapping rule set is used to map the semantic keywords in the behavior tree description text to nodes, construct the hierarchical relationship between nodes, and inject the execution parameters and node judgment conditions of the nodes in the behavior tree description text into the attribute fields of the nodes. The execution control module is used to allocate the ground sensor station network based on the tree behavior model.

[0015] A third aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any implementation of the first aspect.

[0016] A fourth aspect of this application provides an electronic device, comprising: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method described in any implementation of the first aspect.

[0017] Compared with the prior art, this application has the following beneficial effects: This application provides a method for allocating ground sensor network data, including: structured arrangement of acquired satellite communication data and ground sensor network operation data based on a preset prompt word template, setting constraint priorities and handling fault tolerance for abnormal operating conditions to generate target prompt words; processing the target prompt words using a large language model to output a behavior tree description text containing node identifiers, node types, node judgment conditions, execution parameters corresponding to the nodes, and a set of child node identifiers corresponding to each parent node identifier; then parsing the description text into an executable tree behavior model using a text model mapping rule set, and finally allocating the ground sensor network data according to this model. This application, through the synergy of prompt word templates, a large language model, and a mapping rule set, eliminates the reliance on manual experience intervention and fixed rule-based pre-set allocation methods in traditional methods. It can intelligently infer behavior tree description text from the target operating condition data corresponding to satellite communication tasks, instantly convert the behavior tree description text into an executable tree behavior model, and then allocate the ground sensor network data according to the tree behavior model, achieving high-efficiency, high-accuracy, and unmanned intelligent allocation of ground sensor network data. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating a ground sensor network allocation method provided in this application embodiment; Figure 2 A flowchart illustrating a method for generating prompt words for a large language model oriented towards satellite-to-ground scheduling, provided in an embodiment of this application; Figure 3 A flowchart illustrating a method for generating a behavior tree behavior model for satellite communication tasks, provided in this application embodiment; Figure 4 This is a schematic diagram of the structure of a ground sensor station network distribution device provided in an embodiment of this application. Detailed Implementation

[0020] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0021] Figure 1 This is a flowchart illustrating a ground sensor station network allocation method provided in an embodiment of this application. (In conjunction with...) Figure 1 As shown, the ground sensor station network allocation method provided in this application includes: S101, acquire target operating condition data corresponding to the satellite communication mission.

[0022] The target operating condition data in this application includes satellite communication data and ground sensor network operation data. Satellite communication data includes, but is not limited to, satellite number, real-time WGS84 latitude, longitude, and altitude, orbital angular velocity, uplink communication frequency band range, downlink communication frequency band range, minimum communication bandwidth requirement, effective transit duration, and orbital maneuver markers. Ground sensor network operation data includes, but is not limited to, the numbers, latitude, longitude, and altitude of all available ground stations, supported frequency band range, current idle time slot table, real-time power consumption of equipment, maximum allowable power consumption, terrain obstruction angle threshold, signal reception sensitivity, and interference intensity from adjacent base stations.

[0023] S102, Based on the preset prompt word template and the target operating condition data, generate the target prompt word.

[0024] The prompt word template in this application is a pre-designed and carefully arranged structured template. This prompt word template is used to structure the target operating condition data, prioritize the constraints in the satellite communication mission, and handle abnormal operating conditions with fault tolerance for missing or abnormal data in the target operating condition data.

[0025] In other words, the cue word template in this application is a dedicated hierarchical constraint cue word template applicable to satellite communication missions. By mathematically representing satellite communication domain knowledge, scheduling constraints, and optimization objectives, a parameterizable and scalable cue word template is constructed to achieve accurate understanding and efficient solution of complex satellite-to-ground scheduling problems by a large language model.

[0026] In one optional implementation, the preset prompt word template includes a four-layer hierarchical structure, a constraint priority encoding mechanism, and an abnormal operating condition fault tolerance mechanism.

[0027] The four-layer hierarchical structure includes a task layer, a condition layer, a constraint layer, and an output format layer. The task layer defines the objectives of the satellite communication mission. The condition layer organizes the condition data corresponding to the satellite communication mission. The constraint layer defines the priority of the constraints of the satellite communication mission. The output format layer defines the text format output by the large language model. The constraints include physical constraints and objective constraints. The objective constraints include maximizing the link signal-to-noise ratio and / or minimizing the communication downtime.

[0028] In other words, the four-layer hierarchical structure of the prompt word template specifically includes four major components: the task layer, the working condition layer, the constraint layer, and the output format layer. This four-layer hierarchical structure is used to drive the large language model to output decision content according to the standard logic of the behavior tree.

[0029] The task layer is used to clarify the objectives of the satellite communication mission. As the header fixed information of the prompt word template, the task layer includes a unique mission identifier, a mission type enumeration value (telemetry, tracking, and command communication, data transmission, emergency communication, orbital maneuvering), a global mission priority (levels 1-5, with level 1 being the highest and level 5 the lowest), and the start and end timestamps of the mission's effective time window. The content of the task layer provides the large language model with a global mission context, ensuring that the decision-making logic is consistent with the mission objectives.

[0030] The operational status layer is used to input real-time satellite position, communication frequency band, and overall status information of the ground station. The operational status layer is divided into two sub-parts: satellite-side operational status and ground station network operational status. All data is presented in key-value pair format: "parameter name=value".

[0031] The constraint layer is used to input physically insurmountable constraints, including frequency band compatibility constraints, line-of-sight communication constraints, and device power consumption constraints. The constraint layer adopts a separate design of "hard constraints + soft constraints." Hard constraints, also known as physical constraints, are constraints that cannot be broken at the physical level, including full communication frequency band compatibility constraints, line-of-sight communication constraints (satellite elevation angle greater than ground station terrain obstruction angle), device power consumption not exceeding the maximum allowable value, and no time slot overlap or conflict. Soft constraints are optimization objective constraints (referred to as objective constraints), which are constraints that the large language model tries to satisfy as much as possible while meeting the hard constraints. These include, but are not limited to, minimizing the number of ground station handovers, maximizing the communication link signal-to-noise ratio, minimizing total device power consumption, and prioritizing the use of high-priority ground stations.

[0032] The output format layer forces the large language model to output results according to a preset text format. As fixed information at the end of prompts, the output format layer uses an absolute imperative tone to explicitly specify the format and field requirements of the content the large language model must output, prohibiting the output of any explanatory text, redundant information, or content outside the specified format. For example, it mandates outputting a topology structure according to tree node specifications, requiring the output of a list of behavioral tree nodes in JSON format. Each node must contain six mandatory fields: node identifier, node type, parent node identifier, node decision condition, execution parameters, and a list of child node identifiers. This ensures that the content output by the large language model can be directly converted into an executable tree behavior model.

[0033] The constraint priority coding mechanism includes a first priority, a second priority, and a third priority; the first priority corresponds to the physical constraint; the second priority corresponds to the target constraint; and the third priority corresponds to the constraint related to operation and maintenance costs in the satellite communication mission; the priorities are from high to low as the first priority, the second priority, and the third priority.

[0034] Based on the above priority settings, this application designs a three-level constraint priority encoding system in the prompt word template to solve the problem of decision confusion in large language models when multiple constraints conflict. This encoding system transforms all constraints into numerical priorities that large language models can directly understand, as follows: Level P0 (first priority, i.e., absolutely forbidden to violate) contains all physical constraints, with a coded value of 0. When generating decision logic, the large language model must first traverse all level P0 constraints, and any decision branch that violates a level P0 constraint must be directly eliminated.

[0035] Level P1 (second priority, i.e., high priority optimization) contains soft constraints directly related to the core objectives of the satellite communication mission, namely the target constraints, such as maximizing the link signal-to-noise ratio and minimizing communication interruption time, with an encoded value of 1. Provided all P0 level constraints are satisfied, the large language model must prioritize satisfying the P1 level constraints.

[0036] Level P2 (third priority, i.e., low priority optimization) includes constraints related to operation and maintenance costs, such as minimizing total device power consumption and prioritizing the use of local ground stations, with a coded value of 2. While satisfying all Level P0 and Level P1 constraints, the large language model can weigh the satisfaction of Level P2 constraints according to the actual situation.

[0037] The abnormal operating condition fault tolerance handling mechanism includes a data completion mechanism, a parameter degradation mechanism, and a task degradation mechanism. The data completion mechanism is used to complete the data based on the historical average value of the parameter when the first type of parameter is missing. The parameter degradation mechanism is used to remove the missing parameter from the operating condition layer when the second type of parameter is missing. The task degradation mechanism is used to reduce the priority of the task corresponding to the third type of parameter when the third type of parameter is missing. The importance of the parameters, from high to low, is the third type of parameter, the second type of parameter, and the first type of parameter.

[0038] Based on the above three-level fault tolerance mechanism, this application designs a three-level fault tolerance processing strategy corresponding to the first type of parameter, the second type of parameter, and the third type of parameter for abnormal input data caused by satellite data transmission delay, ground station sensor failure, etc., as follows: Level 1 fault tolerance (corresponding to the data completion mechanism, applicable to the first type of parameters): For missing non-critical parameters (such as real-time power consumption of the device, interference intensity of adjacent base stations), the historical average value of the parameter is automatically used to complete the data, and the prompt text clearly indicates "This parameter is a historical completion value, and redundancy should be reserved when making decisions".

[0039] Level 2 fault tolerance (corresponding parameter degradation mechanism, applicable to the second type of parameters): For missing key parameters (such as satellite real-time position, ground station supported frequency band), the satellite or ground station is automatically marked as "unavailable", removed from the operating condition layer, and the prompt word is clearly marked "This object will not participate in this scheduling due to missing data".

[0040] Level 3 fault tolerance (corresponding to the task degradation mechanism, applicable to the third type of parameters): When ground station data missing or satellite core parameters are completely missing, exceeding the preset threshold, the task priority is automatically reduced by one level, and the prompt message is clearly marked "This scheduling is in degradation mode, prioritizing the connectivity of core communication links".

[0041] Figure 2 This is a flowchart illustrating a method for generating prompt words for a large language model oriented towards satellite-to-ground scheduling, provided in an embodiment of this application. (Combined with...) Figure 2 As shown, based on the preset prompt word template and target operating condition data, the target prompt word is generated, including the following steps: S201, preprocess the target working condition data to generate a standardized working condition dataset.

[0042] It is understandable that the target operating condition data collected in this application may have problems such as missing data, inconsistent format, or outliers. Therefore, it is necessary to preprocess the target operating condition data, such as data cleaning, outlier removal, and format standardization, so as to obtain a standardized operating condition dataset.

[0043] S202, the metadata in the task layer is used as the content of the task layer.

[0044] As can be seen from the foregoing embodiments, the metadata in the task layer includes a unique task identifier, a task type enumeration value (measurement and control communication, data feedback, emergency communication, track maneuvering tracking), a global task priority (levels 1-5, with level 1 being the highest priority and level 5 being the lowest priority), and the start and end timestamps of the task's effective time window. This metadata is directly filled into the task layer as fixed header information of the task layer, thereby obtaining the content of the task layer.

[0045] S203, Based on the abnormal operating condition fault tolerance processing mechanism, the parameters in the standardized operating condition dataset are processed to generate updated operating condition data, and the updated operating condition data is used as the content of the operating condition layer.

[0046] Specifically, based on the data completion mechanism, parameter degradation mechanism, and task degradation mechanism in the abnormal working condition fault handling mechanism, the missing or abnormal parameters in the standardized working condition dataset are processed accordingly to generate updated working condition data. The updated working condition data is then filled into the working condition layer in the key-value pair format of "parameter name=value" to obtain the content of the working condition layer.

[0047] S204. Based on the constraint priority encoding mechanism, the physical constraints, the target constraints, and the constraints related to operation and maintenance costs in the satellite communication mission, generate the content of the constraint layer.

[0048] Specifically, based on the constraint priority coding mechanism, physical constraints are classified as P0-level hard constraints, target constraints as P1-level soft constraints, and operation and maintenance cost-related constraints as P2-level soft constraints. These constraints are written into the constraint layer in order of priority to obtain the content of the constraint layer.

[0049] S205, according to the text format determined by the output format layer, assemble the content of the task layer, the content of the working condition layer, and the content of the constraint layer to generate the initial prompt word.

[0050] Specifically, the task layer content is used as the fixed information in the prompt word header, the working condition layer content is used as the intermediate dynamic data segment, the constraint layer content is used as the constraint rule segment, and finally, they are assembled according to the JSON format requirements specified by the output format layer to generate a complete initial prompt word.

[0051] It is understandable that the initial prompt words generated according to the steps in S201-S205 may have problems such as format not conforming to the output specifications or decision logic violating physical constraints. Therefore, it is necessary to perform a verification operation on the initial prompt words. Only the initial prompt words that pass the verification operation can be used as target prompt words.

[0052] S206, determine whether the initial prompt word meets the text format specified by the output format layer.

[0053] If the initial prompt word meets the text format determined by the output format layer, proceed to step S207; if the initial prompt word does not meet the text format determined by the output format layer, proceed to step S209.

[0054] S207, determine whether the initial prompt word passes the constraint compliance check.

[0055] The constraint compliance check verifies whether the decision logic in the initial prompt word satisfies all physical constraints. If the decision logic in the initial prompt word violates physical constraints, proceed to step S210; if the decision logic in the initial prompt word satisfies all physical constraints (hard constraints at the P0 level), proceed to step S208.

[0056] S208, Output target prompt words.

[0057] The initial prompt word that is the text format determined by the output format layer and passes the constraint compliance check is used as the target prompt word.

[0058] S209, call the large language model to generate new initial prompt words.

[0059] S210 marks the violated constraint clauses and calls the large language model to generate correction information.

[0060] For example, in this application, the large language model is invoked to generate new initial prompt words three times. However, those skilled in the art can set a reasonable upper limit for the number of retries according to actual needs.

[0061] Figure 2 The method shown can encapsulate real-time operating data into structured prompt words that can be directly recognized by a large language model. Under the premise of clearly defining the physical constraints of this scheduling (such as frequency band matching, communication duration, etc.) and soft optimization objectives (such as minimizing power consumption, maximizing link signal-to-noise ratio, minimizing the number of station-network switching, etc.), target prompt words that conform to the output specifications in terms of format and satisfy all physical constraints in terms of content are generated. This ensures that the decision logic output by the large language model based on the prompt words can be directly transformed into an executable tree behavior model.

[0062] S103, the target prompt words are processed by a preset large language model to generate behavior tree description text.

[0063] The large language model in this application is a pre-trained model that can understand multi-level constraints and output structured decision logic.

[0064] The behavior tree description text in this application is a structured text description used to define the attributes and hierarchical relationships of each node in the behavior tree. It includes node identifier, node type, node decision conditions, execution parameters corresponding to the node, and a set of child node identifiers corresponding to each parent node identifier.

[0065] In this step, the large language model, based on its natural language processing capabilities, performs deep semantic analysis on the target prompt words, autonomously deconstructs the multi-level scheduling decision logic in the target prompt words, and outputs a complete behavior tree description text containing node type, node topology relationship, node judgment threshold, and execution parameters.

[0066] S104, the behavior tree description text is parsed using a preset text model mapping rule set to generate a tree behavior model.

[0067] The text model mapping rule set in this application is used to map semantic keywords in behavior tree description text to nodes, construct hierarchical relationships between nodes, and inject the execution parameters and node determination conditions of the nodes in the behavior tree description text into the attribute fields of the nodes.

[0068] In one optional implementation, the text model mapping rule set includes a behavior tree node library, a mapping rule library, topology generation rules, node parameter injection rules, and logical consistency verification rules.

[0069] The behavior tree node library has multiple types of nodes pre-defined; these types of nodes include root nodes, condition nodes, control nodes, execution nodes, and decorator nodes.

[0070] For example, the behavior tree node library in the text model mapping rule set is customized for the satellite-to-ground station network allocation scenario, containing 18 types of nodes in five categories. All nodes have five basic attributes: unique identifier, node type, parent node, child node, and execution status.

[0071] The five main categories and 18 types of nodes are as follows: Root node class: contains only one type of node, "General Scheduling Task Root Node". The "General Scheduling Task Root Node" serves as the entry point for the entire behavior tree and is responsible for initiating the entire scheduling process. Its attributes include task ID, task priority, and task time window.

[0072] Condition Node Class: Contains seven types of nodes used to determine whether the satellite-to-ground operating conditions meet specific conditions. All condition nodes return either "success" or "failure".

[0073] The condition node class specifically includes: satellite transit condition nodes, frequency band matching condition nodes, line-of-sight communication condition nodes, ground station idle condition nodes, power consumption constraint condition nodes, interference avoidance condition nodes, and mission priority condition nodes. Its attributes include decision thresholds and decision parameters.

[0074] Control node class: It contains four types of nodes, which are used to control the execution flow of the behavior tree.

[0075] The control node classes specifically include: sequential execution nodes (child nodes are executed in sequence, and return success if all succeed), selection execution nodes (child nodes are executed in priority order, and return success if the first succeeds), parallel execution nodes (all child nodes are executed simultaneously, and return success if all succeed), and loop execution nodes (child nodes are executed in a loop until the termination condition is met).

[0076] Execution Node Class: This class contains five types of nodes used to perform specific station network control operations. All execution nodes return values ​​of "success," "failure," or "running." Specifically, this class includes: antenna pointing adjustment node, communication frequency band configuration node, communication time slot locking node, ground station switch control node, and link status monitoring node.

[0077] Decorator Node Class: Contains one type of node used to limit the execution time of other nodes. If the decorated node does not return a result within the specified time, the node is forcibly terminated and a failure is returned.

[0078] The mapping rule base includes a preset semantic keyword and a correspondence between the nodes in the behavior tree node library.

[0079] Based on the above correspondence, a mapping mechanism is established between the natural language decision semantics output by the large language model and the node types of the behavior tree. This application adopts a semantic keyword matching method, which identifies keywords in the output text of the large language model and maps them to the corresponding node types to ensure the accuracy and consistency of the mapping. The specific mapping rules are as follows: Condition node mapping rule: When the output of the large language model contains keywords such as "if", "when", and "satisfy", and the semantic content judges whether a certain condition is true, it is mapped to the corresponding condition node.

[0080] Control node mapping rules: When the output of the large language model contains keywords such as "in sequence", "in order", "first", "then", it is mapped to "sequential execution node"; when it contains keywords such as "or", "preferred choice", "if it fails", it is mapped to "selection execution node"; when it contains keywords such as "simultaneously", "in parallel", "together", it is mapped to "parallel execution node"; when it contains keywords such as "loop", "repeat", "until", it is mapped to "loop execution node".

[0081] Execution node mapping rules: When the output of the large language model contains verbs such as "adjust", "set", "configure", "lock", "open", "close", "monitor", etc., and the semantic content is to operate the ground station equipment, it is mapped to the corresponding execution node.

[0082] Decorator node mapping rule: When the output of the large language model contains keywords such as "timeout", "maximum wait", or "if it exceeds X seconds", it is mapped to a "timeout reset node", where the timeout threshold is set to X seconds.

[0083] The topology generation rules include that the root node is unique, the control node includes any number of child nodes, and the condition node and the execution node do not include child nodes.

[0084] Based on the above rules, this application automatically constructs the topology of the behavior tree by analyzing the parent-child relationships and execution order of the nodes in the behavior tree description text (hereinafter referred to as the behavior tree) output by the large language model. The construction process follows the following three core rules: The unique root node rule: The entire behavior tree has one and only one root node, and all other nodes are direct or indirect children of this root node.

[0085] Nested hierarchy rules: Control nodes can contain any number of child nodes; condition nodes and execution nodes can only be child nodes of leaf nodes or control nodes, and they cannot contain child nodes themselves.

[0086] Execution order rules: The child nodes of the selected execution node are arranged in descending order of priority; the child nodes of the sequential execution node are arranged in the order of execution; the child nodes of the parallel execution node have no order requirement.

[0087] The node parameter injection rules include writing execution parameters and decision parameters into the attribute fields of the corresponding nodes based on parameter name matching principles. Based on these rules, this application employs a "parameter name matching + type conversion" mechanism to automatically inject the specific execution parameters and decision parameters output by the large language model into the customized attributes of the corresponding nodes, ensuring the accuracy of parameter injection. The specific injection method is as follows: For conditional nodes: The decision threshold and decision parameters output by the large model are injected into the node's "decision threshold" and "decision parameter" attributes, respectively, and the corresponding data types are converted.

[0088] For execution nodes: inject the specific operation parameters output by the large model into the node's "execution parameters" attribute, and perform validity checks on the injected parameters.

[0089] For control nodes and decoration nodes: inject parameters such as priority and timeout from the large model output into the corresponding attribute fields.

[0090] The logical consistency check is used to verify the topological legality, branch integrity, node reachability, parameter legality, and compliance of physical constraints of the tree behavior model.

[0091] Based on the above verification objectives, and to avoid logical vulnerabilities in the generated behavior tree, this application designs a set of logical consistency verification rules. The specific rules are as follows: Topological validity: There must be no circular references between nodes in the behavior tree, that is, there cannot be a situation where node A is the parent node of node B, and at the same time node B is the parent node of node A.

[0092] Branch integrity: All condition nodes must have both a corresponding success branch and a failure branch, and there must be no unprocessed condition branches.

[0093] Node reachability: There must be a valid execution path from the root node to each execution node, and there must be no unreachable execution nodes.

[0094] Parameter validity: All node parameters must be within the valid value range and must not exceed physical limits.

[0095] Physical constraint compliance: The execution logic of all nodes must meet the P0 level hard constraints, and there must be no execution operations that violate the hard constraints.

[0096] Figure 3 A flowchart illustrating a method for generating a behavior tree model for satellite communication tasks, provided in an embodiment of this application. (Combined with...) Figure 3 As shown, this application parses the behavior tree description text using a preset text model mapping rule set to generate a tree behavior model, including the following steps: S301, Semantic parsing is performed on the node type, node determination conditions and execution parameters of each node in the behavior tree description text to generate semantic keywords and parameter values ​​corresponding to each node.

[0097] Specifically, it receives the JSON-formatted behavior tree description text output by the large language model, performs structured parsing, and extracts the node ID, node type, parent node ID, node decision conditions, execution parameters, and child node ID list for each node.

[0098] Based on this, semantic parsing is performed on the node type, decision conditions, and execution parameters contained in each node to extract the corresponding semantic keywords and parameter values.

[0099] S302, based on the mapping rule library, the semantic keywords corresponding to the node are mapped to the target node in the behavior tree node library, and according to the node parameter injection rule, the parameter value and the judgment condition are written into the attribute field of the target node to obtain the behavior tree node to be assembled.

[0100] First, based on the preset mapping rule library, the semantic keywords of each node extracted in S301 are matched one by one with the node types in the behavior tree node library corresponding to the satellite communication task to determine the target node type corresponding to each semantic keyword.

[0101] Then, based on the matched node type, the corresponding node object is instantiated. Following this, according to the node parameter injection rules, the decision parameters and execution parameters parsed in S301 are injected into the customized attribute fields of the node object, and the injected parameters are validated to ensure that the parameter values ​​are within the legal range.

[0102] Finally, we obtain behavior tree nodes that have completed parameter injection and validity verification, ready for subsequent assembly.

[0103] S303, based on the node identifiers in the behavior tree description text and the set of child node identifiers corresponding to the parent node identifiers, the behavior tree nodes to be assembled are hierarchically connected according to the topology generation rules to obtain the initial tree behavior model.

[0104] First, from the behavior tree description text parsed from S301, obtain the node ID (node ​​identifier) ​​of each node and its corresponding parent node ID and child node ID list (the set of child node identifiers corresponding to the parent node identifier).

[0105] Then, based on the preset topology generation rules, the hierarchical relationship between nodes is established one by one—each behavior tree node to be assembled is attached to the corresponding parent node according to its parent node ID, and each child node in its child node ID list is connected to that node in turn.

[0106] Through this recursive hierarchical connection operation, all nodes to be assembled that have completed parameter injection are combined according to the original topology of the behavior tree, and finally a complete behavior tree topology with correct parent-child relationships is constructed, resulting in the initial tree behavior model.

[0107] S304, after the initial tree behavior model passes the logical consistency check, the initial tree behavior model is used as the tree behavior model.

[0108] First, the initial tree behavior model is fully verified according to the preset logical consistency verification rules to ensure that it meets the verification requirements of topology legality, branch integrity, node reachability, parameter legality and physical constraint compliance. Then, the initial tree behavior model is further verified by various checks such as logical closed-loop verification, website resource conflict verification, and frequency band interference pre-detection.

[0109] If the initial tree behavior model successfully passes all the above checks, it is considered to have logical correctness and structural integrity; if there are logical defects, return to S103 and regenerate the initial tree behavior model until it passes the checks.

[0110] Subsequently, the validated initial tree behavior model is serialized into an executable behavior tree format, and a unique behavior tree identifier is generated for it to mark the model's identity and version information.

[0111] At this point, the initial tree behavior model, after verification and serialization processing, is officially output as the final tree behavior model for subsequent deployment and execution.

[0112] S105, Based on the tree behavior model, allocate the ground sensor network.

[0113] Specifically, after obtaining the tree behavior model, the system sequentially completes steps such as working condition determination, optimal ground station selection, communication parameter configuration, and hardware equipment control based on the tree behavior model, thereby realizing the autonomous allocation of satellite and ground sensor station network, that is, realizing intelligent scheduling and matching of ground station resources.

[0114] In one alternative implementation, after allocating the ground sensor network based on the tree behavior model, feedback data (link communication quality, equipment operating status, and task completion rate) can be collected from the ground sensor network during the execution of the tree behavior model. The feedback data is then input into the large language model, which updates the scheduling prior knowledge base to achieve long-term adaptive optimization so that the large language model can correct the tree behavior model based on the feedback data. The large language model iteratively optimizes the logic of the next round of behavior tree nodes based on the feedback data, thereby achieving continuous autonomous optimization of the scheduling strategy.

[0115] It can be understood that the process of autonomous communication scheduling between the ground sensor network and satellites using the ground sensor network allocation method provided in this application is divided into two scenarios: initial deployment and routine operation.

[0116] Initial deployment refers to the debugging of communication links between the device (hereinafter referred to as the target device) corresponding to the ground sensor station network allocation method and the ground sensor station network, satellite orbit prediction system and on-board communication payload, importing the behavior tree node library and preset prompt word templates corresponding to the satellite communication mission, and entering the static basic parameters of the ground station.

[0117] Routine operation refers to the continuous polling and collection of real-time orbital positions and communication frequency band requirements of passing satellites, as well as the idle status of ground stations, by the devices corresponding to the ground sensor station network allocation method. Based on preset prompt word templates and the collected target operating condition data, target prompt words are automatically generated. The target prompt words are processed by a large language model to generate behavior tree description text. Then, the behavior tree description text is parsed by a preset text model mapping rule set to generate a tree behavior model. After logical consistency verification, the ground sensor station network is allocated based on the tree behavior model. Control commands are issued to complete antenna pointing adjustment, frequency band matching, and communication link establishment. During the mission execution, feedback data is collected and input into the large language model to continuously optimize the behavior tree node logic. After the satellite passes over, the current scheduling branch is automatically settled and ground station resources are released.

[0118] In summary, the ground sensor station network allocation method provided in this application has at least the following advantages: First, it achieves fully automated scheduling, significantly reducing operation and maintenance costs. The large language model in this application can automatically generate and iterate behavior tree scheduling logic based on real-time operating conditions, thereby significantly reducing the manpower costs of site operation and maintenance, while also significantly reducing scheduling errors introduced by manual operation.

[0119] Second, it is highly adaptable to dynamic satellite operating conditions, significantly improving resource utilization. The proposed solution can respond in real time to dynamic changes such as satellite orbital maneuvers, frequency band switching, and sudden communication tasks. Compared with fixed-rule scheduling schemes, it can significantly improve the utilization rate of ground sensor network hardware resources and effectively reduce the idle rate of spectrum resources. It is particularly suitable for the high real-time requirements of high-speed dynamic communication scenarios in low-Earth orbit constellations.

[0120] Third, it exhibits exceptional generalization capabilities, compatible with all types of satellite-to-ground communication missions. This application's solution eliminates the need to redesign algorithm rules or behavior tree structures for different satellite models or communication missions. The large language model can autonomously adapt to various missions, including telemetry, tracking, and command (TT&C) communication, data transmission, two-way voice communication, and emergency communication. It is compatible with heterogeneous high and low Earth orbit satellites and different frequency band communication systems, significantly enhancing its versatility.

[0121] Fourth, closed-loop self-iterative capability ensures continuous performance optimization over long-term operation. This application's solution can autonomously accumulate scheduling experience during long-term operation, continuously optimize the behavior tree decision logic, and ensure that system performance evolves over time.

[0122] Based on the same inventive concept, this application also provides a ground sensor station network distribution device. Figure 4 This is a schematic diagram of a ground sensor station network distribution device provided in an embodiment of this application. (Combined with...) Figure 4As shown, the ground sensor station network distribution device 400 provided in this application includes: The data acquisition module 401 is used to acquire target operating condition data corresponding to the satellite communication mission; the target operating condition data includes communication data from the satellite and operational data from the ground sensor network. The prompt word generation module 402 is used to generate target prompt words based on a preset prompt word template and the target operating condition data; the prompt word template is used to structure and arrange the target operating condition data, set the priority of the constraints in the satellite communication mission, and perform fault tolerance processing for missing or abnormal data in the target operating condition data; The description text generation module 403 is used to process the target prompt words through a preset large language model to generate behavior tree description text; the behavior tree description text includes node identifier, node type, node judgment condition, execution parameters corresponding to the node, and a set of child node identifiers corresponding to each parent node identifier; The behavior model construction module 404 is used to parse the behavior tree description text through a preset text model mapping rule set to generate a tree behavior model; the text model mapping rule set is used to map the semantic keywords in the behavior tree description text to nodes, construct the hierarchical relationship between nodes, and inject the execution parameters and node judgment conditions of the nodes in the behavior tree description text into the attribute fields of the nodes. The execution control module 405 is used to allocate the ground sensor station network based on the tree behavior model.

[0123] In one alternative implementation, the prompt word generation module 402 includes: The data preprocessing submodule is used to preprocess the target working condition data to generate a standardized working condition dataset; The first content generation submodule is used to take the metadata in the task layer as the content of the task layer; the metadata includes task ID, task type, priority and time window; The second content generation submodule is used to process the parameters in the standardized working condition dataset based on the abnormal working condition fault tolerance processing mechanism, generate updated working condition data, and use the updated working condition data as the content of the working condition layer. The third content generation submodule is used to generate the content of the constraint layer based on the constraint priority encoding mechanism, the physical constraints, the target constraints, and the constraints related to operation and maintenance costs in the satellite communication mission. The initial prompt word generation submodule is used to assemble the content of the task layer, the content of the working condition layer, and the content of the constraint layer according to the text format determined by the output format layer to generate the initial prompt word; The target prompt word generation submodule is used to perform a verification operation on the initial prompt word to obtain the target prompt word.

[0124] In one alternative implementation, the target prompt word generation submodule includes: A text format verification unit is used to verify whether the initial prompt word meets the text format specified by the output format layer; A compliance check unit is used to perform a constraint compliance check on the initial prompt if the initial prompt meets the text format; the constraint compliance check is used to verify whether the decision logic in the initial prompt meets the physical constraints; The first target prompt word determination unit is configured to use the initial prompt word as the target prompt word if the initial prompt word meets the text format and the initial prompt word passes the constraint compliance check. The second target prompt word determination unit is used to modify the initial prompt word if the initial prompt word does not meet the text format or fails the constraint compliance check, until the modified initial prompt word meets the text format and passes the constraint compliance check, thereby obtaining the target prompt word.

[0125] In one alternative implementation, the behavior model building module 404 includes: The semantic parsing unit is used to perform semantic parsing on the node type, node determination conditions and execution parameters of each node in the behavior tree description text, and generate semantic keywords and parameter values ​​corresponding to each node. The node tree assembly unit is used to map the semantic keywords corresponding to the node to the target node in the behavior tree node library based on the mapping rule library, and write the parameter values ​​and the judgment conditions into the attribute fields of the target node according to the node parameter injection rules to obtain the behavior tree node to be assembled. The initial tree behavior construction unit is used to hierarchically connect the behavior tree nodes to be assembled according to the node identifiers in the behavior tree description text and the set of child node identifiers corresponding to the parent node identifiers, in accordance with the topology generation rules, to obtain the initial tree behavior model. The tree behavior model acquisition unit is used to use the initial tree behavior model as the tree behavior model after the initial tree behavior model passes the logical consistency check; the logical consistency check is used to verify the topology legality, branch integrity, node reachability, parameter legality and compliance of physical constraints.

[0126] In one alternative implementation, the ground sensor station network distribution device 400 further includes: The data acquisition module is used to collect feedback data from the ground sensor network during the execution of the tree behavior model; the feedback data includes link communication quality, equipment operating status, and task completion rate. The large model correction module is used to input the feedback data into the large language model so that the large language model can correct the tree behavior model based on the feedback data.

[0127] Based on the ground sensor network allocation method and apparatus provided in the foregoing embodiments, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements some or all of the steps in the ground sensor network allocation method mentioned above.

[0128] Based on the ground sensor station network allocation method and apparatus provided in the foregoing embodiments, this application also provides an electronic device, including: A memory on which computer programs are stored; A processor is configured to execute the computer program in the memory to implement some or all of the steps in the ground sensor network allocation method provided in the foregoing embodiments.

[0129] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for the device embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiments. The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separate. The components indicated as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment solution according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0130] The above description is merely one specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for allocating ground sensor station networks, characterized in that, The method includes: Acquire target operating condition data corresponding to the satellite communication mission; the target operating condition data includes communication data from the satellite and operational data from the ground sensor network; Based on the preset prompt word template and the target operating condition data, a target prompt word is generated; the prompt word template is used to structure and arrange the target operating condition data, set the priority of the constraints in the satellite communication mission, and perform fault tolerance processing for missing or abnormal data in the target operating condition data. The target prompt words are processed by a pre-defined large language model to generate a behavior tree description text. The behavior tree description text includes node identifiers, node types, node determination conditions, execution parameters corresponding to the nodes, and a set of child node identifiers corresponding to each parent node identifier. The behavior tree description text is parsed using a preset text model mapping rule set to generate a tree behavior model. The text model mapping rule set is used to map semantic keywords in the behavior tree description text to nodes, construct hierarchical relationships between nodes, and inject the execution parameters and node determination conditions of the nodes in the behavior tree description text into the attribute fields of the nodes. Based on the tree behavior model, the ground sensor network is allocated.

2. The method according to claim 1, characterized in that, The preset prompt word template includes a four-layer hierarchical structure, a constraint priority coding mechanism, and an abnormal operating condition fault tolerance mechanism. The four-layer hierarchical structure includes a task layer, a condition layer, a constraint layer, and an output format layer. The task layer defines the objectives of the satellite communication mission. The condition layer organizes the condition data corresponding to the satellite communication mission. The constraint layer defines the priority of the constraints of the satellite communication mission. The output format layer defines the text format output by the large language model. The constraints include physical constraints and objective constraints. The objective constraints include maximizing the link signal-to-noise ratio and / or minimizing the communication downtime. The constraint priority coding mechanism includes a first priority, a second priority, and a third priority; the first priority corresponds to the physical constraint; the second priority corresponds to the target constraint; and the third priority corresponds to the constraint related to operation and maintenance costs in the satellite communication mission; the priorities are ranked from high to low as the first priority, the second priority, and the third priority. The abnormal operating condition fault tolerance mechanism includes a data completion mechanism, a parameter degradation mechanism, and a task degradation mechanism; the data completion mechanism is used to complete the data based on the historical average value of the parameter when the first type of parameter is missing; the parameter degradation mechanism is used to remove the missing parameter from the operating condition layer when the second type of parameter is missing. The task degradation mechanism is used to reduce the priority of the task corresponding to the third type of parameter when the third type of parameter is missing; the importance of the parameters from high to low is the third type of parameter, the second type of parameter, and the first type of parameter.

3. The method according to claim 2, characterized in that, The generation of target prompt words based on the preset prompt word template and the target operating condition data includes: The target operating condition data is preprocessed to generate a standardized operating condition dataset; The metadata in the task layer is used as the content of the task layer; the metadata includes task ID, task type, priority, and time window; Based on the abnormal operating condition fault tolerance processing mechanism, the parameters in the standardized operating condition dataset are processed to generate updated operating condition data, and the updated operating condition data is used as the content of the operating condition layer. Based on the constraint priority encoding mechanism, the physical constraints, the target constraints, and the constraints related to operation and maintenance costs in the satellite communication mission, the content of the constraint layer is generated. According to the text format determined by the output format layer, the content of the task layer, the content of the working condition layer, and the content of the constraint layer are assembled to generate the initial prompt words; The initial prompt word is validated to obtain the target prompt word.

4. The method according to claim 3, characterized in that, The step of verifying the initial prompt word to obtain the target prompt word includes: Verify whether the initial prompt word meets the text format specified by the output format layer; If the initial prompt word meets the text format, then a constraint compliance check is performed on the initial prompt word; the constraint compliance check is used to verify whether the decision logic in the initial prompt word meets the physical constraints; If the initial prompt word meets the text format and passes the constraint compliance check, then the initial prompt word is used as the target prompt word; If the initial prompt word does not meet the text format, or the initial prompt word fails the constraint compliance check, the initial prompt word is modified until the modified initial prompt word meets the text format and passes the constraint compliance check, thus obtaining the target prompt word.

5. The method according to claim 1, characterized in that, The text model mapping rule set includes a behavior tree node library, a mapping rule library, topology generation rules, and node parameter injection rules; The behavior tree node library contains multiple types of nodes; these types of nodes include root nodes, condition nodes, control nodes, execution nodes, and decorator nodes. The mapping rule base includes a preset semantic keyword and the correspondence between each node in the behavior tree node library; The topology generation rules include that the root node is unique, the control node includes any number of child nodes, and the condition node and the execution node do not include child nodes. The node parameter injection rules include writing the execution parameters and decision parameters into the attribute fields of the corresponding node based on the parameter name matching principle.

6. The method according to claim 5, characterized in that, The step of parsing the behavior tree description text using a preset text model mapping rule set to generate a tree behavior model includes: Semantic parsing is performed on the node type, node determination conditions, and execution parameters of each node in the behavior tree description text to generate semantic keywords and parameter values ​​corresponding to each node; Based on the mapping rule library, the semantic keywords corresponding to the nodes are mapped to target nodes in the behavior tree node library, and according to the node parameter injection rules, the parameter values ​​and the judgment conditions are written into the attribute fields of the target nodes to obtain the behavior tree nodes to be assembled. Based on the node identifiers in the behavior tree description text and the set of child node identifiers corresponding to the parent node identifiers, the behavior tree nodes to be assembled are hierarchically connected according to the topology generation rules to obtain the initial tree behavior model. Once the initial tree behavior model passes the logical consistency check, the initial tree behavior model is used as the tree behavior model; the logical consistency check is used to verify the topology legality, branch integrity, node reachability, parameter legality, and compliance of physical constraints.

7. The method according to any one of claims 1-6, characterized in that, After allocating the ground sensor network based on the tree behavior model, the process further includes: The ground sensor network collects feedback data during the execution of the tree behavior model; the feedback data includes link communication quality, equipment operating status, and task completion rate. The feedback data is input into the large language model so that the large language model can correct the tree behavior model based on the feedback data.

8. A ground sensor station network distribution device, characterized in that, The device includes: The data acquisition module is used to acquire target operating condition data corresponding to the satellite communication mission; the target operating condition data includes communication data from the satellite and operational data from the ground sensor network. The prompt word generation module is used to generate target prompt words based on a preset prompt word template and the target operating condition data; the prompt word template is used to structure and arrange the target operating condition data, set the priority of the constraints in the satellite communication mission, and perform fault tolerance processing for missing or abnormal data in the target operating condition data; The description text generation module is used to process the target prompt words through a preset large language model to generate behavior tree description text; the behavior tree description text includes node identifier, node type, node judgment condition, execution parameters corresponding to the node, and a set of child node identifiers corresponding to each parent node identifier; The behavior model construction module is used to parse the behavior tree description text through a preset text model mapping rule set to generate a tree behavior model; the text model mapping rule set is used to map the semantic keywords in the behavior tree description text to nodes, construct the hierarchical relationship between nodes, and inject the execution parameters and node judgment conditions of the nodes in the behavior tree description text into the attribute fields of the nodes. The execution control module is used to allocate the ground sensor station network based on the tree behavior model.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method described in any one of claims 1-7.

10. An electronic device, characterized in that, include: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method according to any one of claims 1-7.

Images