Low-altitude equipment collaborative awareness method, system and equipment

Abstract

The invention discloses a low-altitude equipment cooperative sensing method, system and equipment, and belongs to the technical field of low-altitude intelligent management and control. The method comprises the following steps: firstly, receiving a sensing task and analyzing to generate a structured task demand, triggering dynamic access management according to the structured task demand, and solving an optimal equipment combination from a candidate equipment resource pool through a multi-target optimization model to construct a virtual sensing cluster; the digital twinborn deduction module performs multi-strategy prospective deduction through a game model based on the mapped twinborn state and business intention to generate a global collaborative awareness strategy; during strategy execution, the airspace cognitive map module is utilized to monitor abnormal evidence and perform causal reasoning, and a reasoning result is fed back to the digital twinborn deduction module to form closed-loop control. According to the method, the problems of resource scheduling stiffness and decision passive lag of a traditional low-altitude sensing system are solved, elastic on-demand scheduling and active closed-loop management and control of the low-altitude sensing resources are realized, and the intelligence, the self-adaptability and the robustness of the system are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of low-altitude intelligent control technology, and more specifically, to a method, system, and device for collaborative sensing of low-altitude equipment. Background Technology

[0002] With the booming development of the low-altitude economy, application scenarios such as drone logistics, urban air traffic, and low-altitude inspection are becoming increasingly diverse, and low-altitude airspace is gradually transforming into a high-density, highly dynamic, and complex operating environment. To ensure the safety and orderliness of low-altitude flight activities, building a comprehensive, all-weather low-altitude sensing network has become a key aspect of infrastructure construction. Currently, a massive number of heterogeneous sensing devices are deployed in low-altitude airspace, including various radars, optoelectronic pods, radio monitoring stations, and ground sensors. These devices continuously generate multi-dimensional spatiotemporal data, forming the underlying foundation for physical sensing. Simultaneously, the rapid advancements in digital twin technology and artificial intelligence algorithms have provided a technological foundation for the digital mapping and intelligent management of the physical world. This makes the unified management of physical devices and the simulation and extrapolation of airspace situations through digital means an inevitable trend in industry development. How to efficiently utilize these technologies to achieve accurate understanding of the low-altitude environment has become a current research hotspot. However, existing low-altitude sensing systems still have significant limitations when facing complex and ever-changing on-demand tasks. On the one hand, traditional methods of configuring sensing resources are mostly static and pre-defined, with fixed network topologies between devices. This makes it difficult to scale up and down flexibly and schedule resources on demand based on sudden monitoring tasks or dynamically changing airspace environments, resulting in low resource utilization and delayed response. On the other hand, when faced with multi-source heterogeneous devices, existing technologies often focus on the simple aggregation of underlying data, lacking a deep understanding of business intent and forward-looking strategy deduction. This causes the system to only passively respond to the current state and cannot predict future trends. Summary of the Invention

[0003] The purpose of this invention is to provide a method, system, and device for collaborative sensing of low-altitude equipment, which realizes the transformation of low-altitude sensing from static configuration to task-driven. By using digital twin game theory and cognitive graph causal reasoning, it solves the problems of rigid resource scheduling and delayed decision-making, and improves the system's adaptive collaborative sensing and closed-loop control capabilities in complex environments.

[0004] This invention is achieved through the following technical solution:

[0005] A collaborative sensing method for low-altitude equipment, the method comprising the following steps: Receive perception task instructions, parse and generate structured task requirement vectors, and trigger dynamic access management accordingly to update the candidate device resource pool; Based on the task requirement vector, a multi-objective optimization model is constructed to solve the candidate device resource pool and generate the Pareto optimal device combination to construct a virtual sensing cluster. Multi-source data acquisition is initiated based on the configuration parameters of the virtual sensing cluster. The acquired data is synchronously transmitted to the spatial cognitive map module for semantic association updates and to the digital twin inference module for state mapping. The digital twin simulation module, based on the mapped twin's state and business intent, performs multi-strategy forward simulation through a game theory model to generate a global collaborative perception strategy. The global collaborative perception strategy is mapped to device control commands and sent to the virtual perception cluster for execution. At the same time, the spatial cognitive map module is used to monitor abnormal evidence in the execution results, perform causal reasoning, and feed it back to the digital twin inference module to form a closed-loop control.

[0006] Optionally, the parsing and generation of the structured task requirement vector specifically includes: Create task instructions To Intent Model mapping relationship ; The intent model Construct into triples ,in, For the set of objective functions, For the set of constraints, Set a set for preferences. This is a parsing mapping function.

[0007] Optionally, updating the candidate device resource pool specifically includes constructing a dynamic capability profile for each device, the dynamic capability profile containing schedulable status. ; The schedulable state Based on device load rate With remaining power Solve .

[0008] Optionally, the dynamic capability profile may also include a spatiotemporal coverage model. It is defined as the effective sensing range of a device in three-dimensional space:

[0009] in, For the geographical location of the equipment, For the detection radius, For the range of elevation angles, This is the lower limit of the effective elevation angle for the equipment in the vertical direction. This represents the upper limit of the effective elevation angle for the equipment in the vertical direction. This is a function for calculating the elevation angle. For three-dimensional real space, Let be the coordinate vector of any spatial point within the task area.

[0010] Optionally, the multi-objective optimization model aims to maximize the coverage of the task region. Minimize average response time Maximize system lifetime For the objective function, construct the following optimization problem:

[0011] And satisfy constraints ; in, For the equipment combination to be selected, This is a pool of candidate devices.

[0012] Optionally, the acquisition frequency of the multi-source data acquisition Based on basic data collection capabilities Real-time equipment status and task requirements priority They jointly determine and satisfy the functional relationship: .

[0013] Optionally, the multi-strategy forward deduction specifically adopts the Stackelberg game model, wherein the system strategy is... The environmental response is ,Strategy In the scene Expected utility The specific calculation formula is as follows:

[0014] in, To find the expectation value operator, As a follower in a game, in the face of a strategy, the airspace environment or target is considered. The optimal response strategy at that time. This is a quantitative return function.

[0015] Optionally, the generation of the global collaborative awareness strategy specifically includes: Calculation strategy based on multi-attribute utility theory In all scenarios The overall effect :

[0016] in, For the scene The weight, For set The i-th specific scene in This is a set of key inference scenarios generated for prediction. For strategy In the i-th specific scenario Expected utility under the following conditions; Select to make The strategy for maximizing this is the global collaborative perception strategy.

[0017] A low-altitude equipment collaborative sensing system, comprising: The dynamic access management module is used to parse task instructions, build dynamic capability profiles based on device load rate and remaining power, and construct virtual sensing clusters using multi-objective optimization algorithms; The multi-source data acquisition module is used to perform data acquisition based on the virtual sensing cluster configuration and task priority function; The digital twin inference module is used to parse the business intent model, calculate the expected utility of strategies in different scenarios based on the game model, and generate a global collaborative perception strategy. The spatial cognitive graph module is used to construct an entity semantic relationship graph and perform root cause reasoning on anomalous evidence based on a random causal graph model. The strategy generation and scheduling module receives the strategies generated by the digital twin inference module, maps them into device control commands, and performs closed-loop adjustments based on the inference results of the spatial cognitive map module.

[0018] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of a low-altitude device collaborative sensing method.

[0019] The technical solution of the present invention has at least the following advantages and beneficial effects: This invention introduces a task-driven dynamic cognitive access mechanism, which can dynamically assemble virtual perception clusters in real time based on high-level business needs, and achieves elastic scheduling and precise matching of heterogeneous resources through multi-objective optimization algorithms. On the other hand, this invention utilizes a digital twin inference module, combined with a game theory model, to perform forward-looking simulation and evaluation of perception strategies. This enables the generation of globally robust strategies in complex and uncertain environments, achieving a leap from passive monitoring to proactive prediction. Furthermore, this invention integrates spatial domain cognitive mapping technology, endowing the system with the ability to deeply understand the root causes of anomalies through modeling semantic relationships between entities and random causal reasoning. This constructs a complete closed loop of perception-inference-decision-execution-feedback, greatly enhancing the system's adaptability and anti-interference capabilities. Attached Figure Description

[0020] Figure 1 A flowchart illustrating the low-altitude equipment collaborative sensing method provided by the present invention; Figure 2 This is a schematic diagram illustrating the principle of the low-altitude equipment collaborative sensing system provided by the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0022] The core inventive concept of this invention lies in the following: First, a dynamic cognitive access framework is constructed. This framework can perform real-time capability assessment and dynamic discovery of massive heterogeneous low-altitude sensing devices based on upper-layer task intent. It then automatically selects devices through a multi-objective optimization algorithm to form virtual sensing clusters that meet specific task requirements, thereby achieving elastic scaling and on-demand scheduling of sensing resources. Next, digital twin technology is used to digitally map and deeply analyze the sensing state of the physical world with business intent. By introducing a Stackelberg game model for multi-strategy forward-looking deduction, a globally robust collaborative sensing strategy is generated in complex and uncertain environments, achieving a decision-making upgrade from passive response to proactive prediction. Finally, an innovative airspace cognitive graph is introduced to model the deep semantic relationships between physical entities, spatial entities, and event entities in the low-altitude environment. When an anomaly is detected, a stochastic causal graph model is used for automated root cause tracing, and the reasoning results are fed back to the digital twin inference module, forming an intelligent closed-loop control of "perception-inference-decision-execution-feedback". This fundamentally solves the technical problems of rigid resource scheduling, delayed decision-making, and lack of adaptive optimization capabilities in traditional low-altitude perception systems.

[0023] Example 1 like Figure 1 As shown, this embodiment provides a low-altitude equipment collaborative sensing method, which mainly includes the following steps: Step S1: Receive the perception task instruction, parse and generate a structured task requirement vector, and trigger dynamic access management accordingly to update the candidate device resource pool.

[0024] It's important to explain that this step is the starting point of the entire collaborative perception task. The task instructions received by the system can be in natural language form, such as "conduct high-precision drone intrusion detection in area A," or they can be structured API calls. The intent parser within the digital twin inference module is responsible for transforming these instructions into a machine-understandable structured task requirement vector. Specifically, the intent parser establishes a structured task requirement vector from the instructions... To Intent Model Mapping relationship: As an example, the intent model It can be constructed into a triple: .in, This is a set of objective functions, such as maximizing coverage or minimizing detection latency. A set of constraints, such as total power consumption not exceeding a threshold or the requirement to use a specific type of device; Set up a set of preferences, such as prioritizing the use of high-precision radar.

[0025] After the task requirement vector is generated, the dynamic access management module is triggered. This module first performs device discovery and registration. Specifically, it adopts a hybrid discovery strategy to adapt to heterogeneous network environments: for devices following public protocols (such as ONVIF), discovery is achieved through network listening and proactive polling; for devices using proprietary protocols, discovery beacons are broadcast by probes deployed at the network edge; for devices in sleep or low-power states, wake-up commands are sent through dedicated low-power channels (such as LoRa). After all discovered devices pass security authentication, the system assigns them a globally unique logical identifier (UID) and includes them in the candidate device resource pool.

[0026] Subsequently, the system builds and updates a dynamic capability profile for each device in the resource pool in real time. This profile is the core basis for subsequent device selection. As an optional example, this dynamic capability profile at least includes schedulable status. This state is not static, but rather depends on the device's current real-time load rate. With remaining power Through function The result is obtained through dynamic calculation. It is an element in a discrete state set, such as {ready, light load, heavy load, low battery, fault}, which allows the system to intuitively determine whether the equipment is suitable for undertaking a new task.

[0027] Step S2: Based on the task requirement vector, construct a multi-objective optimization model to solve the candidate device resource pool and generate the Pareto optimal device combination to construct a virtual sensing cluster.

[0028] The core of this step lies in how to intelligently select an optimal combination of devices from numerous candidate devices to perform the task. Specifically, the system models this problem as a multi-objective optimization problem. This is based on the objective function defined in the task requirement vector, such as maximizing the task area coverage. Minimize average response time Maximize system lifetime The following optimization model is constructed:

[0029] Its constraints are the combination of candidate equipment. It must be a candidate device resource pool A subset of, i.e. .

[0030] Among them, coverage The calculations rely on the spatiotemporal coverage model in the device dynamic capability profile. This model precisely defines the effective sensing range of each device in three-dimensional space, and its mathematical expression is as follows:

[0031] here, These are the geographical coordinates of the device. That is its maximum detection radius. It is its effective vertical elevation angle range, and It is a calculation of spatial points The elevation angle function relative to the equipment location. This is calculated jointly by the selected equipment combination. All devices The coverage can be obtained by finding the intersection of the union and the task region. .

[0032] To solve this complex multi-objective optimization problem, this embodiment employs the Fast Non-Dominated Sorting Genetic Algorithm (NSGA-II) with an elitist strategy. This algorithm can effectively explore the vast solution space and find a set of Pareto optimal solutions without dominance relationships. Each solution represents a combination of devices that makes different trade-offs between different objectives (such as coverage, latency, and energy consumption). Finally, the system can set preferences based on the task requirement vector. Automatically select the most suitable combination of devices from the Pareto optimal solution set. Alternatively, the solution set can be provided to the operator for manual decision-making. The selected equipment combination This constitutes the virtual perception cluster for this mission.

[0033] Step S3: Start multi-source data acquisition according to the configuration parameters of the virtual sensing cluster, and synchronously send the acquired data to the spatial cognitive map module for semantic association update, and send it to the digital twin inference module for state mapping.

[0034] Once the virtual sensing cluster is constructed, the system sends acquisition commands to all devices within the cluster. These commands are highly customized. For example, the acquisition frequency of each device. It is no longer fixed, but rather based on its own basic data collection capabilities. Current real-time status (From the ability profile) and the priority of task requirements The dynamic adjustment can be represented by a function: This approach ensures that critical equipment operates at high frequencies, while non-critical equipment can operate at lower frequencies to save energy.

[0035] The collected multi-source heterogeneous data undergoes preliminary processing at edge nodes (such as format standardization and timestamp alignment) before being sent in parallel to two core modules. One data stream flows to the spatial domain cognitive graph module for real-time updates of entity states and relationships within the graph. The other data stream flows to the digital twin inference module for synchronous updates of the digital twin, ensuring consistency between the virtual and physical worlds.

[0036] Step S4: Based on the mapped twin state and business intent, the digital twin simulation module performs multi-strategy forward simulation through a game theory model to generate a global collaborative perception strategy.

[0037] This step is the core of the system's intelligent decision-making. After receiving real-time data and updating the twin's state, the digital twin simulation module combines it with the previously analyzed business intent model. A multi-strategy forward-looking simulation engine is launched. The goal of this engine is no longer to simply respond to the current state, but to predict multiple possible future scenarios and generate the optimal response strategy for each scenario.

[0038] Specifically, this embodiment uses the Stackelberg game model to evaluate the merits of different strategies. In this model, the collaborative perception system is considered the "leader," and its decision-making strategy is... Other intelligent agents in the airspace (such as non-cooperative drones) or environmental changes are considered "followers," and their reactions are... The system needs to predict the optimal response of its followers and, based on this, choose the strategy that is most advantageous to itself. In a specific simulation scenario Expected utility It is obtained by solving for the equilibrium of this game, and its calculation formula is as follows: Here It is a utility function that quantifies the extent to which a system achieves its task objective given a policy and environmental response.

[0039] To generate a global policy that performs well under various uncertain scenarios, the system aggregates policies. In all the key scenarios considered The lower utility forms a comprehensive utility. One feasible aggregation method is to use multi-attribute utility theory for weighted summation:

[0040] in, For the scene The probability of occurrence or its importance weight. Ultimately, the system chooses to maximize overall utility. Maximization strategy The global collaborative perception strategy is the output.

[0041] Step S5: The global collaborative perception strategy is mapped into device control commands and sent to the virtual perception cluster for execution. At the same time, the spatial domain cognitive map module is used to monitor abnormal evidence in the execution results, perform causal reasoning, and feed it back to the digital twin inference module to form a closed-loop control.

[0042] Generated global policy This is a high-level description, such as "Command Radar 1 to switch to sector scan mode, focusing on area B, and simultaneously command Electro-optical pod 2 to track targets detected by the radar." The strategy generation and scheduling module is responsible for "translating" this high-level strategy into a specific sequence of control commands that can be executed by the underlying devices.

[0043] During command execution, the spatial cognitive map module acts as an "inspector." It continuously receives and analyzes data transmitted back from the virtual sensing cluster, comparing it with existing knowledge in the map to monitor execution effectiveness. Once abnormal evidence is detected... For example, if a target's trajectory is suddenly interrupted or the data reported by the device deviates significantly from the expected model, the graph module will immediately activate the causal inference engine.

[0044] In this embodiment, causal reasoning employs a stochastic causal graph model. When abnormal evidence is detected... At that time, the model will calculate each potential root cause. Posterior probabilities (such as equipment failure, signal interference, target avoidance behavior, etc.):

[0045] in, This is abnormal evidence. For the j-th potential root cause, as a potential root cause The posterior probability of occurrence, for The prior probability, In order to occur The abnormal evidence appeared under the following conditions. The conditional probability; k is the summation index of the potential root causes. For the k-th potential root cause, as a potential root cause The prior probability, In order to cause Abnormal evidence appears under certain conditions The conditional probability is calculated. The result is a root cause ranking list with confidence levels.

[0046] This formula is based on Bayes' theorem and incorporates the prior probability of the root cause. and the conditional probability of this anomalous evidence occurring given the root cause. The calculation result is a root cause ranking list with confidence levels. This inference result is immediately fed back to the digital twin inference module, acting as a new disturbance event in the environment, triggering a local, rapid re-inference to generate adjustment strategies, which are then distributed to the devices to cope with the anomaly. This process forms a complete, fast-responding closed-loop control, greatly improving the system's robustness and adaptability.

[0047] Example 2 like Figure 2 As shown, this embodiment provides a low-altitude equipment collaborative sensing system, which serves as the physical carrier for implementing the above-described method. The system includes the following functional modules: The dynamic access management module is configured to perform the discovery, authentication, and registration of heterogeneous devices, and to build a dynamic capability profile for each device. Specifically, this module integrates functions for calculating the schedulable state of devices and algorithms for constructing a three-dimensional spatiotemporal coverage model. Furthermore, this module is also configured to construct and utilize multi-objective optimization algorithms, such as a fast non-dominated sorting genetic algorithm with an elitist strategy, to solve for candidate devices based on the requirements of upper-layer tasks, in order to automatically construct the optimal virtual sensing cluster.

[0048] The multi-source data acquisition module is configured to receive the virtual sensing cluster configuration generated by the dynamic access management module and issue acquisition commands to devices within the cluster accordingly. Internally, this module includes a function for dynamically calculating the device acquisition frequency, taking into account device capabilities, real-time status, and task priority. Simultaneously, this module is also responsible for allocating high-priority channels for data transmission, ensuring the real-time and reliable flow of data.

[0049] The digital twin simulation module is configured as the system's "decision-making brain." This module first parses the high-level business intent model, then updates the twin's state based on real-time data synchronized from the data acquisition module. Its core function is to launch a multi-strategy forward-looking simulation engine. This engine incorporates a Stackelberg game model to calculate and evaluate the expected utility of various alternative strategies under different simulation scenarios, and ultimately generates the globally optimal collaborative perception strategy through utility aggregation.

[0050] The airspace cognitive graph module is configured as the system's "knowledge center." This module is responsible for building and maintaining a graph containing various entities in the low-altitude environment and their complex semantic relationships. It keeps synchronized with real-time data through a streaming update algorithm. Its key function is that when an anomaly is detected in the system, it can activate an inference engine based on a stochastic causal graph model to analyze the abnormal evidence, calculate and output the most likely root cause and its confidence level.

[0051] The strategy generation and scheduling module is configured as a bridge connecting decision-making and execution. It receives the global collaborative perception strategy generated by the digital twin inference module and translates it into a sequence of specific control commands that can be recognized and executed by the underlying devices. At the same time, it also receives causal inference results from the spatial domain cognitive graph module and triggers a closed-loop adjustment process accordingly, issuing adjustment commands to relevant devices.

[0052] The service and visualization module is configured as the interface for system-user interaction. It provides standardized data query APIs and generates decision recommendations based on inference results. Furthermore, it can render the real-time status of the digital twin and the structure and content of the spatial cognitive map into intuitive two-dimensional or three-dimensional visualizations, facilitating user monitoring and understanding of the overall system's operational status.

[0053] Example 3 In addition, embodiments of the present invention also provide an electronic device, including: a memory for storing a computer program; and a processor for executing the computer program stored in the memory, wherein when the computer program is executed, it implements the low-altitude equipment collaborative sensing method described in any of the above embodiments of the present invention.

[0054] In one example, the electronic device could be a server deployed in the cloud or at an edge computing node. Its processor is responsible for running the core computational tasks described above, such as multi-objective optimization, digital twin simulation, and causal reasoning. Its memory stores a dynamic capability profile model, multi-objective optimization algorithms, a game theory engine, a spatial cognitive map structure, and a causal reasoning model. The device communicates with various sensing devices and user terminals deployed in low-altitude environments via a network.

[0055] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For system embodiments, since they largely correspond to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0056] The basic principles of the present invention have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, effects, etc., mentioned in the present invention are merely examples and not limitations. The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of the invention. Therefore, the present invention is not intended to be limited to the aspects shown herein, but is to be carried out within the widest scope consistent with the principles and novel features disclosed herein.

[0057] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for low-altitude device cooperative perception, characterized in that, The method comprises the following steps: Receiving a perception task instruction, parsing a structured task demand vector, and triggering dynamic access management according to the vector to update a candidate device resource pool; Based on the task demand vector, a multi-objective optimization model is constructed to solve the candidate device resource pool, and a Pareto optimal device combination is generated to construct a virtual perception cluster; According to the configuration parameters of the virtual perception cluster, multi-source data acquisition is started, and the collected data is synchronously transmitted to the airspace cognitive graph module for semantic correlation update and to the digital twin deduction module for state mapping; The digital twin deduction module performs multi-strategy forward deduction based on the mapped twin state and business intent through a game model to generate a global collaborative perception strategy; The global collaborative perception strategy is mapped to device control instructions and executed by the virtual perception cluster, while the airspace cognitive graph module is used to monitor abnormal evidence in the execution results, perform causal reasoning, and feedback to the digital twin deduction module to form a closed-loop control.

2. The low-altitude device collaborative perception method of claim 1, wherein, The structured task demand vector is parsed, which specifically includes: Establishment task instruction To the mapping relationship of the intention model ;​ The intent model constructed as triples wherein, is a set of objective functions, is a set of constraints, is a set of preference settings, is a set of parsing mapping functions.

3. The low-altitude device collaborative perception method of claim 1, wherein, updating the pool of candidate device resources, in particular comprising building a dynamic capability profile for each device, the dynamic capability profile containing schedulable states ; The schedulable state Based on device load rate With the remaining power Solving .

4. The low-altitude device cooperative perception method of claim 3, wherein, The dynamic capability profile further comprises a spatio-temporal coverage model defined as the effective sensing range of the device in three-dimensional space: wherein, is a device geographic position, is a detection radius, is an elevation range, is a lower limit of the effective detection elevation of the device in the vertical direction, is an upper limit of the effective detection elevation of the device in the vertical direction, is an elevation calculation function, is a three-dimensional real space, is an arbitrary spatial point coordinate vector within the task area.

5. The low-altitude device collaborative perception method of claim 1, wherein, The multi-objective optimization model aims to maximize task area coverage , minimize average response latency , maximize system lifetime as an objective function, the optimization problem is constructed as follows: and satisfy constraints ; wherein, is a set of candidate devices, is a pool of candidate device resources.

6. The low-altitude device collaborative perception method of claim 1, wherein, The acquisition frequency of the multi-source data acquisition is determined by the basic acquisition capability , the real-time state of the equipment and the priority of the task requirement together, satisfying the functional relationship: 。 7. The low-altitude device coordinated perception method of claim 1, wherein, The multi-strategy forward deduction specifically adopts a Stackelberg game model, wherein a system strategy is , an environmental reaction is , a strategy is expected utility under a scene , and a specific calculation formula is as follows: ​ wherein, is the expected value operator, is the optimal reaction strategy of the follower in the game against the strategy of the leader, is the quantization function.

8. The low-altitude device collaborative perception method of claim 7, wherein, The global collaborative perception strategy is generated, which specifically includes: By multi-attribute utility theory In all scenarios Under the integrated utility : wherein, is the weight of the scenario, is the weight of the scenario, is the ith particular scenario in the set is the set of key inferred scenarios generated from the prediction, is the set of key inferred scenarios generated from the prediction, is the expected utility under the ith particular scenario is the expected utility under the ith particular scenario is the expected utility under the ith particular scenario selecting a strategy that maximizes the global co-awareness strategy.

9. A low-altitude device coordination and perception system, comprising: It includes: A dynamic access management module is used to parse the task instruction, construct a dynamic capability profile based on the device load rate and remaining power, and construct a virtual perception cluster using a multi-objective optimization algorithm; A multi-source data acquisition module is used to perform data acquisition according to the virtual perception cluster configuration and task priority function; A digital twin deduction module is used to parse the business intent model, calculate the expected utility of the strategy in different scenarios based on a game model, and generate a global collaborative perception strategy; An airspace cognitive graph module is used to construct an entity semantic relationship graph, and perform root cause reasoning on abnormal evidence based on a stochastic causal graph model; A strategy generation and scheduling module is used to receive the strategy generated by the digital twin deduction module, map it to device control instructions, and perform closed-loop adjustment according to the reasoning results of the airspace cognitive graph module.

10. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to realize the steps of the low-altitude device collaborative perception method according to any one of claims 1 to 8.

Images