Low-altitude industry chain information intelligent cooperation and airspace management platform based on federal learning

Abstract

The invention provides a low-altitude industry chain information intelligent cooperation and airspace management platform based on federal learning. Through cluster ad hoc network and decentralized federated learning, the method is divided into three stages of data acquisition, data preprocessing and analysis application, low-altitude scene data communication of dynamic topology and static topology is realized, and the efficient communication requirement in a low-altitude operation scene is met. According to the system, differential data transmission strategies are adopted, optimization design is carried out on a single machine, a co-task cluster, a cooperative cluster and a non-cooperative cluster, and the communication efficiency and the resource utilization rate are improved. In the data preprocessing stage, federated learning aggregation of multi-source data fusion, data anonymization and decentralization is combined, and data privacy protection and safe sharing of three parties of industry, university and research are ensured. In the analysis and application stage, low-altitude aircraft task scheduling, airspace flow control and environmental condition evaluation are realized through aircraft analysis, airspace analysis, environmental analysis and digital twinborn technologies.

Description

Technical Field

[0001] The present invention belongs to the field of low-altitude airspace management and intelligent information collaboration technology, and in particular, relates to a low-altitude industrial chain information intelligent collaboration and airspace management platform based on federated learning. Background Art

[0002] At present, the management of low-altitude airspace faces multiple challenges in the context of the rapid growth of demand for low-altitude aircraft applications, including limited airspace resources, complex scheduling of low-altitude aircraft clusters, and difficulties in protecting data privacy. However, the existing centralized and traditional airspace management methods have obvious shortcomings in dealing with these challenges:

[0003] Communication bottleneck and single point failure risk of centralized management: Traditional centralized airspace management methods rely on central servers or centralized control systems. All low-altitude aircraft data and mission instructions must be processed and issued through the central server. This model is prone to communication bottlenecks when the number of low-altitude aircraft increases, resulting in data transmission delays and increased communication burden. In addition, once the central server fails, the entire airspace management system may fail, affecting the safety of the low-altitude aircraft cluster and the continuity of mission execution.

[0004] The contradiction between multi-party data sharing and privacy protection: Low-altitude airspace management usually involves the collaboration of multiple entities (such as enterprises, scientific research institutions, government departments, etc.). The data of each party contains sensitive information, which requires data sharing to achieve joint airspace management, while also protecting the privacy of each party's data. The current centralized data sharing model has great risks in terms of data privacy protection. Traditional encryption technology is difficult to support multi-party collaborative modeling while ensuring data privacy, making it difficult to balance multi-party data sharing and privacy protection.

[0005] Heterogeneous and multi-source data integration is difficult: Low-altitude airspace management requires real-time collection and integration of dynamic and static data from different sources (such as aircraft status, meteorological information, environmental data, etc.). These data sources and formats are diverse, which increases the complexity of data fusion and unified processing. Existing data integration methods lack effective technical means when processing multi-source heterogeneous data, and it is difficult to provide a consistent and real-time global data view, resulting in limited real-time and accuracy of the system in task scheduling and airspace monitoring.

[0006] Poor adaptability to dynamic environments: Low-altitude aircraft need to cope with dynamic changes in meteorological conditions and airspace traffic when flying in low-altitude airspace. Traditional airspace management systems lack the perception of real-time environmental and airspace conditions, lack timely prediction and application of environmental data such as wind speed, temperature, and humidity, and are difficult to achieve dynamic optimization of low-altitude aircraft flight paths, resulting in reduced safety and mission completion rates of aircraft in complex environments.

[0007] Lack of intelligent collaboration mechanism applicable to the low-altitude economic industry chain: With the development of the low-altitude economy, low-altitude airspace management is no longer limited to a single task, but involves multiple application scenarios such as logistics distribution, inspection, and monitoring. However, the existing system does not provide sufficient support for intelligent collaboration among the four parties of industry, academia, research, and application, making it difficult to achieve data collaboration and task optimization among all parties in the low-altitude economy, and lacks an intelligent collaboration platform that adapts to the needs of the low-altitude economic industry chain. Summary of the invention

[0008] In view of this, the present invention aims to overcome the shortcomings of the above-mentioned problems in the prior art and proposes a low-altitude industrial chain information intelligent collaboration and airspace management platform based on federated learning.

[0009] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0010] A low-altitude industry chain information intelligent collaboration and airspace management platform based on federated learning, including a data acquisition module, a data preprocessing module, and an analysis application module connected in sequence;

[0011] The data acquisition module is used to collect static and dynamic data in the low-altitude airspace, and adopts differentiated transmission strategies according to the cluster type to optimize communication efficiency and resource utilization;

[0012] The data preprocessing module achieves data consistency integration, privacy protection and collaborative modeling through multi-source data fusion, data anonymization and sharding processing, and decentralized federated learning model aggregation;

[0013] The analysis application module is used to perform intelligent analysis of aircraft, airspace and environment, including aircraft safety assessment, path optimization, airspace traffic management, environmental condition prediction and digital twin simulation, to achieve intelligent support for low-altitude airspace management and low-altitude aircraft mission scheduling.

[0014] Furthermore, the data acquisition module obtains static topological data and dynamic topological data from the environment. The static topological data is basic information containing spatial topological distribution that does not change with time, and is used for long-term planning and airspace layout. The dynamic topological data is real-time data that changes with mission execution and airspace environment changes, and is used for real-time monitoring, path optimization and risk avoidance of low-altitude aircraft.

[0015] Furthermore, the data acquisition module adopts differentiated transmission strategies according to the cluster type to optimize communication efficiency and resource utilization, specifically including:

[0016] Stand-alone mode: The low-altitude aircraft transmits its mission data and status information directly to the nearest regional base station via unicast;

[0017] Same-task cluster mode: low-altitude aircraft use multicast and ad hoc networks for communication. Several reporting nodes are selected in the cluster to collect data within the cluster and perform data compression and aggregation, and finally generate a joint message and upload it to the regional base station;

[0018] Cooperative cluster mode: Low-altitude aircraft use multi-hop communication to transmit data, forward data through neighboring nodes, and achieve efficient task information sharing. At the same time, priority is set for uploading of critical task data to ensure that time-sensitive data can be transmitted quickly;

[0019] Non-cooperative cluster mode: Low-altitude aircraft nodes communicate independently with regional base stations and are allocated independent communication frequency bands or channels to avoid interference between clusters and ensure the independence of each task and data security.

[0020] Furthermore, in the data preprocessing module, multi-source data fusion is used to integrate the topological and spatiotemporal information of heterogeneous data; data anonymization and sharding are used to protect data privacy; decentralized federated learning model aggregation is used to achieve cross-party collaborative modeling without sharing original data, specifically including:

[0021] Multi-source data fusion: Multi-source data fusion integrates heterogeneous data from different sources and types, unifying static and dynamic data, topology and spatiotemporal information into a standardized data format;

[0022] The heterogeneous data is defined as follows: assuming that the data sources in the low-altitude airspace include a static data set and a dynamic data set;

[0023] Topological integration: For multi-source data distributed in space, a topological integration function is used to map data at different locations into a unified spatial coordinate system to form a spatial information matrix covering the entire area.

[0024] Spatiotemporal information integration integrates dynamic data that changes over time into a topological structure in time series to form a spatiotemporal data set. The spatiotemporal integrated data is used as the input of the global model to support the intelligent scheduling of low-altitude aircraft and optimization of airspace resources in a time-varying environment.

[0025] Data sharding divides the data of each participant into multiple shards according to content or spatial area. Data anonymization anonymizes the data of each shard and implements data privacy protection by adding differential privacy noise.

[0026] Decentralized federated learning model aggregation realizes collaborative model updates among different participants, ensuring that the joint training of the global model is completed without leaving the local data; specifically, it includes:

[0027] Local model training: Each participant trains its model parameters on the local data set to obtain updated local model parameters;

[0028] Decentralized model aggregation: Decentralized federated learning achieves parameter aggregation through model parameter exchange between adjacent participants;

[0029] Global model output: After completing multiple rounds of decentralized aggregation, a global model is generated. The global model aggregates the local models of each participant to form a joint model that contains the data features of multiple parties, thereby realizing cross-party collaborative modeling.

[0030] Furthermore, the analysis application module includes aircraft analysis, airspace analysis, environmental analysis and digital twin sub-modules;

[0031] Aircraft analysis is used for flight safety assessment and path optimization, airspace analysis is used for airspace traffic management and segment speed control strategy, environmental analysis is used for wind field and meteorological condition prediction, and the digital twin submodule is used to build a virtual airspace environment to support airspace simulation and visualization; specifically, it includes:

[0032] The aircraft analysis module evaluates the flight status and optimizes the flight path to ensure flight safety and minimize energy consumption during mission execution, including:

[0033] Flight safety assessment: Real-time monitoring of the flight status of low-altitude aircraft to ensure that they fly within a safe range;

[0034] Path optimization: Optimize the path according to mission requirements and environmental conditions, with the goal of minimizing energy consumption and flight risks;

[0035] The airspace analysis module analyzes the traffic conditions in the low-altitude airspace, and realizes the reasonable allocation of airspace resources through traffic management and segment speed control strategies to ensure the orderly operation of different aircraft in the same airspace, including:

[0036] Traffic management: optimize airspace traffic distribution, avoid route congestion, and ensure safe distance between aircraft;

[0037] Segment speed control strategy: Based on the real-time airspace environment and mission requirements, set the optimal speed of the aircraft in each segment to avoid congestion or collision risks

[0038] The environmental analysis module monitors and predicts the wind field and meteorological conditions in the airspace in real time, helping the aircraft to make necessary flight adjustments to ensure flight safety and mission stability, including:

[0039] Wind field prediction: Use historical wind field data and real-time meteorological data to predict the wind speed and direction in low-altitude airspace to ensure stable flight of the aircraft under different wind conditions;

[0040] Meteorological condition analysis: by analyzing environmental variables, predicting weather trends and providing early warning information for aircraft;

[0041] The digital twin submodule realizes simulation and visualization support for low-altitude airspace by building a virtual airspace environment, so as to help managers intuitively understand the airspace status and provide decision support, including:

[0042] Virtual airspace environment construction: Based on multi-source data fusion and federated learning model output, a virtual airspace environment model is constructed to simulate the airspace conditions of aircraft operations;

[0043] Simulation and visualization: Through simulation in a virtual environment, the real-time operating status of the aircraft and airspace traffic conditions can be displayed, facilitating real-time monitoring and management of low-altitude airspace.

[0044] Compared with the prior art, the low-altitude industry chain information intelligent collaboration and airspace management platform based on federated learning described in the present invention has the following advantages:

[0045] This invention adopts a federated learning framework to achieve multi-party collaborative modeling through decentralized model aggregation, enabling participants to jointly build a global model without sharing original data. Compared with the traditional centralized data sharing model, this platform supports cross-organizational data collaboration while ensuring data privacy, and adapts to the low-altitude airspace management needs of multiple parties.

[0046] Through multi-source data fusion technology, the platform realizes unified and standardized processing of static and dynamic data, integrates multi-dimensional data such as airspace environment, aircraft status and meteorological information, and provides a global view for airspace management. In addition, the platform designs differentiated communication strategies according to the type of low-altitude aircraft cluster, effectively reducing the communication burden, improving the efficiency of data transmission, and ensuring the adaptability of the system in various scenarios.

[0047] The platform of the present invention dynamically predicts the weather conditions in the airspace through real-time environmental analysis, combined with historical data and real-time meteorological information, and provides early warning support for low-altitude aircraft. This function enables low-altitude aircraft to adjust their flight strategies in complex and changeable meteorological environments to ensure the stable execution of missions, greatly improving the platform's environmental adaptability compared to existing technologies.

[0048] The present invention realizes the dynamic allocation and priority scheduling of low-altitude airspace resources through the flow management and flight segment speed control strategy of the airspace analysis module, avoids route congestion, and improves the resource utilization rate of the airspace. Compared with the traditional airspace management system, the platform can ensure the orderly operation of low-altitude aircraft in a high-density mission environment and reduce the waste of airspace resources.

[0049] Through digital twin technology, the present invention constructs a virtual airspace environment for simulation exercises of low-altitude aircraft missions and visualization of mission status. The digital twin module provides managers with an intuitive view of airspace status, supports mission prediction and real-time adjustment, and makes the platform's intelligent collaboration and airspace management more intuitive and effective.

[0050] The present invention is particularly suitable for the collaboration of the low-altitude economic industry chain among production, learning, research and application parties, and meets the data sharing, security collaboration and intelligent analysis needs of different participants in low-altitude airspace management. The decentralized design and intelligent analysis capabilities of the platform enable it to provide efficient and safe airspace management support in multiple application scenarios, laying a solid foundation for the intelligent development of the low-altitude economy. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0052] Figure 1 It is a schematic diagram of the structure of a low-altitude industrial chain information intelligent collaboration and airspace management platform based on federated learning of the present invention.

[0053] Figure 2 This is a schematic diagram of a differentiated communication strategy for a low-altitude industrial chain information intelligent collaboration and airspace management platform based on federated learning in the present invention.

[0054] Figure 3 This is a schematic diagram of low-altitude airspace zoning and functional distribution of a low-altitude industrial chain information intelligent collaboration and airspace management platform based on federated learning in the present invention. DETAILED DESCRIPTION

[0055] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0056] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and the like are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0057] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood by specific circumstances.

[0058] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0059] like Figure 1 The present invention provides a low-altitude industry chain information intelligent collaboration and airspace management platform based on federated learning, and its main modules are as follows:

[0060] Data acquisition module: used to obtain the data of spatial static topological distribution and spatial dynamic topological distribution in the low-altitude airspace environment, and design differentiated communication strategies based on cluster types to optimize communication efficiency and resource utilization. Specifically include:

[0061] Static and dynamic data acquisition: The data acquisition module obtains static topological data (such as airspace control information, base station information) and dynamic topological data (such as low-altitude aircraft status, real-time sensor information) from the environment. static , the dynamic data set is D dynamic , then the total data D is expressed as:

[0062] D=D static ∪D dynamic

[0063] The static topological data is basic information containing spatial topological distribution that does not change over time and is mainly used for long-term planning and airspace layout. This type of data includes but is not limited to the following:

[0064] Airspace control information: covers airspace division, flight restriction areas, flight altitude restrictions, etc. This information provides boundary constraints for low-altitude aircraft path planning and mission execution to ensure compliance with airspace management regulations and safety requirements.

[0065] Base station information: The geographical location, signal coverage and channel allocation of communication base stations in low-altitude airspace. Base station information determines the optimal selection of data transmission paths and provides stable communication support for low-altitude aircraft during flight.

[0066] Land-based sensor data: mainly includes basic environmental data collected by ground-based meteorological sensors (such as temperature and humidity sensors, air pressure sensors, etc.). These data provide basic parameters of the ground environment for low-altitude aircraft flight.

[0067] Meteorological remote sensing data: Wide-area meteorological information obtained through remote sensing equipment, including wind speed, wind direction, temperature, humidity, cloud cover and rainfall conditions. These meteorological information provide the basic meteorological background of low-altitude airspace in the form of static data, especially providing a wide range of meteorological conditions reference before the start of the mission. The spatial distribution information of meteorological remote sensing data is set as G weather , which can be used to predict the meteorological conditions of low-altitude aircraft routes, is expressed as:

[0068] G weather = {W wind ,W temperature ,W humidity ,W precipitation}

[0069] Among them, W wind , W temperature , W humidity and W precipitation Respectively represent the distribution of wind speed, temperature, humidity and rainfall.

[0070] The dynamic topology data is data acquired in real time that changes with mission execution and airspace environment, and is mainly used for real-time monitoring, path optimization and risk avoidance of low-altitude aircraft. This type of data includes but is not limited to the following:

[0071] Aircraft status: including the low-altitude aircraft's location information (GPS coordinates), speed, acceleration, attitude (yaw, pitch, roll angle), remaining battery power and other flight status information.

[0072] Airborne sensor data: Various sensors carried by low-altitude aircraft collect detailed information and environmental conditions of the mission execution area in real time. The various sensors include but are not limited to cameras, lidar, infrared sensors, meteorological sensors, etc. Airborne sensor data D onboard It can be expressed as:

[0073] D onboard ={image, lidar, infrared sensor, wind speed}

[0074] Among them, image data is real-time images taken by onboard cameras, which are used for target recognition, environmental perception, terrain detection, etc.; lidar data is three-dimensional terrain maps or obstacle information generated by lidar, which provides spatial structure information for low-altitude aircraft and supports path adjustment and safe flight; infrared sensors obtain temperature data, which are used to determine heat sources or monitor ambient temperature changes; wind speed data is the local wind speed measured by onboard meteorological sensors.

[0075] Airspace operation status: Real-time acquisition of low-altitude airspace operation traffic conditions, including the location, speed, mission status, etc. of other nearby aircraft, to ensure the safe flight and dynamic avoidance of low-altitude aircraft in the airspace. Airspace operation status airspace Contains the status information of other aircraft, which is used for airspace flow control and conflict avoidance, expressed as:

[0076] R airspace ={Position other ,Velocity other ,Mission other}

[0077] Among them, Position other Refers to the current position of other aircraft in the low-altitude airspace; Velocity other is the current velocity component of other aircraft; Mission other Refers to the type of mission currently being performed by surrounding aircraft, which is used by the system to determine the flight intention of the aircraft.

[0078] Real-time environmental information: Real-time environmental data obtained through ground-based and airborne sensors is updated as the environment changes dynamically. Including wind speed, temperature, humidity, air pressure, etc., to ensure that low-altitude aircraft can adapt to the environment while being able to respond quickly to sudden meteorological changes. Real-time environmental information real-time The expression is:

[0079] E real-time ={E wind ,E temperature ,E humidity ,E pressure}

[0080] Among them, E wind The real-time ambient wind speed is used to monitor the wind impact in the current area; E temperature is the temperature information of the flight area; E humidity is the humidity information of the flight area; E pressure It is the air pressure information of the environment.

[0081] like Figure 2 As shown, the differentiated communication strategy based on cluster type design flexibly adjusts the data transmission mode according to the cluster type of low-altitude aircraft (stand-alone mode, same-task cluster mode, cooperative cluster mode and non-cooperative cluster mode), optimizes the use of communication resources and reduces data redundancy. Specifically expressed as:

[0082] 1. Single-machine mode: Low-altitude aircraft perform tasks independently without the need to coordinate with other low-altitude aircraft, so the data transmission path is relatively simple. Each low-altitude aircraft transmits its mission data and status information directly to the nearest regional base station in the form of unicast, which is suitable for scenarios with small data volume and simple communication paths. Suppose the mission data of a single low-altitude aircraft is D single , the unicast transmission path is P single :

[0083] P single :D single →Nearest ground station

[0084] Among them, the communication overhead C in stand-alone mode is single Mainly depends on the distance d single , which can be expressed as:

[0085] C single =f(d single )

[0086] 2. Same-task cluster mode: Applicable to scenarios where multiple low-altitude aircraft collaborate to perform the same task. In order to reduce communication load and data redundancy, low-altitude aircraft in the cluster use multicast and ad hoc networking for communication. Select several reporting nodes in the cluster. It is used to collect data within the cluster and perform data compression and aggregation to finally generate a joint message U task And upload to the regional base station.

[0087] Let the data of each node in the cluster be D i , the reporting node will report the data of all nodes Summarize and compress to generate a joint message U task , expressed as:

[0088]

[0089] Where f(g) is the data compression and aggregation function. The communication overhead of the reporting node C report It can be expressed as:

[0090] C report =g(|S task |,d report )

[0091] Where \S task | is the number of nodes in the cluster, d report To report the distance from the node to the regional base station.

[0092] 3. Collaborative cluster mode: Applicable to scenarios where multiple low-altitude aircraft collaborate to complete complex tasks, such as those that require frequent data sharing, path planning collaboration, and task allocation. Low-altitude aircraft in this mode use multi-hop communication to transmit data and forward data through neighboring nodes to achieve efficient task information sharing. At the same time, priority is set for uploading critical mission data to ensure that time-sensitive data can be transmitted quickly.

[0093] Assume that the key task data in the cooperative cluster is D key , the node set in the cooperative cluster is S coop , the neighbor node is N(i),

[0094]

[0095] Where g(g) represents the priority transmission strategy. Multi-hop transmission path P coop It can be expressed as:

[0096] P coop :D key →N(i)→Regional base station

[0097] Communication overhead of cooperative clusters C coop Depends on the number of nodes in the cluster|S coop | and the number of nodes n in the multi-hop communication path hops :

[0098] C coop =h(|S coop |,n hops )

[0099] 4. Non-cooperative cluster mode: Applicable to low-altitude aircraft clusters performing different tasks in the same airspace. These low-altitude aircraft only share necessary motion data. In this mode, each low-altitude aircraft node communicates independently with the regional base station and is allocated an independent communication frequency band or channel to avoid interference between clusters and ensure the independence of each task and data security.

[0100] Suppose each node of the non-cooperative cluster is S non-coop, whose task data is The transmission path is

[0101] →Nearest ground station

[0102] The total communication cost of non-cooperative clusters C non-coop It is expressed as the sum of the independent communication overhead of each node:

[0103]

[0104] Where f(d i ) represents the transmission distance d from a single node to the base station i The communication overhead generated.

[0105] The data preprocessing module includes multi-source data fusion, data anonymization, and decentralized federated learning model aggregation; multi-source data fusion is used to integrate the topological and spatiotemporal information of heterogeneous data; data anonymization and sharding are used to protect data privacy; decentralized federated learning model aggregation is used to achieve cross-party collaborative modeling without sharing original data. Specifically, it includes:

[0106] Multi-source data fusion: Multi-source data fusion aims to integrate heterogeneous data of different sources and types, unify static and dynamic data, topology and spatiotemporal information into a standardized data format, ensure data consistency and improve the accuracy of system analysis.

[0107] The heterogeneous data is defined as: Assume that the data sources in the low-altitude airspace include the static data set D dynamic ={D UAV ,D sensor ,D airspace}, and dynamic data set D dynamic ={D UAV ,D sensor ,D airspace};

[0108] The topological integration is to use the topological integration function T for spatially distributed multi-source data. topo (·) Map the data at different locations to a unified spatial coordinate system to form a spatial information matrix M covering the entire domain space :

[0109] M space =T topo (D static ,D dynamic )

[0110] The spatiotemporal information integration: the dynamic data that changes with time is integrated into the topological structure in time series to form a spatiotemporal data set M spatio-temp. Let the time dimension be t, then the space-time integration function T spatio-temp Expressed as:

[0111] M spatio-temp =T spatio-temp (M space ,t)

[0112] Among them, T spatio-temp (·) is used to ensure the consistency of spatial and temporal information. The data after spatiotemporal integration can be used as the input of the global model to support the intelligent scheduling of low-altitude aircraft and airspace resource optimization in time-varying environments.

[0113] Data anonymization and sharding: Data anonymization and sharding are used to protect data privacy and ensure that the original data is not leaked when data is exchanged between different participants. The data sharding divides the data of each participant into multiple shards according to content or spatial area. Suppose the data set of each participant is D i , whose fragment is D i ={D i1 ,D i2 ,...,D in}; The data anonymization is performed on each shard data D ij Anonymization is performed by adding differential privacy noise ∈ ij Achieve data privacy protection. Anonymize data D i ' j The expression is:

[0114] D i ' j =D ij +∈ ij ,∈ ij ~Laplace(0,δ)

[0115] Among them, δ is the differential privacy parameter, which controls the intensity of noise.

[0116] Decentralized federated learning model aggregation: Decentralized federated learning model aggregation is used to achieve collaborative model updates between different participants, ensuring that the joint training of the global model is completed without leaving the local data. Specifically, it includes:

[0117] Local model training: Each participant i has a local dataset D i ′ to train its model parameters θ i , get the updated local model parameters Assume the loss function is L i (θ), then the local model update can be expressed as:

[0118]

[0119] Where η is the learning rate, is the gradient of the loss on the local data.

[0120] Decentralized model aggregation: Decentralized federated learning achieves parameter aggregation through model parameter exchange between adjacent participants. Suppose the neighbor node set of participant i is Then the model parameters after decentralized aggregation are for:

[0121]

[0122] Among them, w ij is the weight, satisfying Through multiple rounds of aggregation, the model parameters of all nodes gradually converge, and finally a globally consistent model θ is obtained. global .

[0123] Global model output: After completing multiple rounds of decentralized aggregation, the generated global model θ global It can be expressed as:

[0124]

[0125] Among them, α i is the data volume weight of the participants, satisfying The global model aggregates the local models of each participant to form a joint model that contains data features from multiple parties, thereby achieving cross-party collaborative modeling.

[0126] The analysis and application modules include aircraft analysis, airspace analysis, environmental analysis and digital twin submodules; aircraft analysis is used for flight safety assessment and path optimization, airspace analysis is used for airspace traffic management and segment speed control strategy, environmental analysis is used for wind field and meteorological condition prediction, and digital twin submodule is used to build a virtual airspace environment and support airspace simulation and visualization. Specifically include:

[0127] The aircraft analysis module evaluates the flight status and optimizes the flight path to ensure flight safety and minimize energy consumption during mission execution. The main functions of this module are as follows:

[0128] Flight safety assessment: Real-time monitoring of the flight status (position, speed, attitude, etc.) of low-altitude aircraft to ensure that it flies within a safe range. Suppose the state set of low-altitude aircraft is S UAV ={Position, Velocity, Orientation}, flight safety assessment model F safe Determine whether the flight meets safety conditions:

[0129]

[0130] Path optimization: Optimize the path according to the mission requirements and environmental conditions, with the goal of minimizing energy consumption and flight risk. Let the path planning function be p(x), and the path optimization objective function be:

[0131]

[0132] Among them, c risk (x) and c energy (x) represent the risk and energy cost in the path respectively. The optimal configuration of the flight path is achieved by solving the optimization problem.

[0133] like Figure 3 As shown, the airspace analysis module analyzes the traffic conditions in the low-altitude airspace, and through traffic management and segment speed control strategies, it realizes the reasonable allocation of airspace resources and ensures the orderly operation of different aircraft in the same airspace. The main functions of this module are as follows:

[0134] Airspace flow management: Airspace flow management aims to optimize the flow distribution in the airspace, avoid route congestion, and ensure a safe distance between aircraft. Suppose the set of aircraft in the airspace is The traffic management objectives are:

[0135]

[0136] Among them, u(R i ) represents the efficiency of resource utilization, v(R j ) represents the resource conflict cost, and this objective function is used to achieve the rational allocation of resources in the airspace.

[0137] Segment speed control strategy: Based on the real-time airspace environment and mission requirements, set the optimal speed of the aircraft in each segment to avoid congestion or collision risks. Speed ​​control strategy function V control Expressed as:

[0138]

[0139] Among them, Distance to_other Indicates the distance between the current aircraft and other aircraft, Mission priority Indicates the task priority.

[0140] The environmental analysis module monitors and predicts the wind field and meteorological conditions in the airspace in real time, helping the aircraft to make necessary flight adjustments to ensure flight safety and mission stability. The main functions of this module are as follows:

[0141] Wind field prediction: Based on historical wind field data and real-time meteorological data, the wind speed and direction in low-altitude airspace are predicted to ensure the stable flight of the aircraft under different wind conditions. predBased on time t, it can be expressed as:

[0142] W pred (t+1)=f(W(t),Δt)

[0143] Among them, W(t) represents the current wind field state, and Δt is the time step.

[0144] Meteorological condition analysis: By analyzing environmental variables such as temperature, humidity, and air pressure, the weather change trend is predicted and early warning information is provided to the aircraft. Meteorological condition prediction model E pred It is expressed as:

[0145] E pred (t+1)=g(E(t),Δt)

[0146] Among them, E(t) represents the current meteorological conditions.

[0147] The digital twin module builds a virtual airspace environment to achieve simulation and visualization support for low-altitude airspace, helping managers to intuitively understand airspace conditions and provide decision support. The main functions of this module are as follows:

[0148] Virtual airspace environment construction: Based on multi-source data fusion and federated learning model output, a virtual airspace environment model M is constructed. twin , used to simulate the airspace conditions for aircraft operations:

[0149] M twin =h(D fused ,θ global )

[0150] Among them, h(·) is the virtual environment generation function, D fused is the fused multi-source data, θ global The global model output by federated learning.

[0151] Simulation and visualization: Through the virtual environment simulation, the real-time operation status of the aircraft and the airspace traffic situation are displayed to facilitate the real-time monitoring and management of low-altitude airspace. sim It is expressed as:

[0152] V sim =Visualize(M twin ,S UAV )

[0153] This simulation module helps managers predict and optimize task execution in a virtual environment.

[0154] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. Low-altitude industry chain information intelligent collaboration and airspace management platform based on federated learning, characterized by: It includes a data acquisition module, a data preprocessing module and an analysis application module which are connected in sequence; The data acquisition module is used to collect static and dynamic data in the low-altitude airspace, and adopts differentiated transmission strategies according to the cluster type to optimize communication efficiency and resource utilization; The data preprocessing module achieves data consistency integration, privacy protection and collaborative modeling through multi-source data fusion, data anonymization and sharding processing, and decentralized federated learning model aggregation; The analysis application module is used to perform intelligent analysis of aircraft, airspace and environment, including aircraft safety assessment, path optimization, airspace traffic management, environmental condition prediction and digital twin simulation, to achieve intelligent support for low-altitude airspace management and low-altitude aircraft mission scheduling.

2. The low-altitude industrial chain information intelligent collaboration and airspace management platform based on federated learning according to claim 1 is characterized by: The data acquisition module obtains static topology data and dynamic topology data from the environment. The static topology data is basic information containing spatial topology distribution that does not change with time, and is used for long-term planning and airspace layout. The dynamic topology data is real-time data that changes with mission execution and airspace environment changes, and is used for real-time monitoring, path optimization and risk avoidance of low-altitude aircraft.

3. The low-altitude industrial chain information intelligent collaboration and airspace management platform based on federated learning according to claim 1 is characterized by: The data acquisition module adopts differentiated transmission strategies according to the cluster type to optimize communication efficiency and resource utilization, specifically including: Stand-alone mode: The low-altitude aircraft transmits its mission data and status information directly to the nearest regional base station via unicast; Same-task cluster mode: low-altitude aircraft use multicast and ad hoc networks for communication. Several reporting nodes are selected in the cluster to collect data within the cluster and perform data compression and aggregation, and finally generate a joint message and upload it to the regional base station; Cooperative cluster mode: Low-altitude aircraft use multi-hop communication to transmit data, forward data through neighboring nodes, and achieve efficient task information sharing. At the same time, priority is set for uploading of critical task data to ensure that time-sensitive data can be transmitted quickly; Non-cooperative cluster mode: Low-altitude aircraft nodes communicate independently with regional base stations and are allocated independent communication frequency bands or channels to avoid interference between clusters and ensure the independence of each task and data security.

4. The low-altitude industrial chain information intelligent collaboration and airspace management platform based on federated learning according to claim 1 is characterized by: In the data preprocessing module, multi-source data fusion is used to integrate the topological and spatiotemporal information of heterogeneous data; data anonymization and sharding are used to protect data privacy; Decentralized federated learning model aggregation is used to achieve cross-party collaborative modeling without sharing original data, including: Multi-source data fusion: Multi-source data fusion integrates heterogeneous data from different sources and types, unifying static and dynamic data, topology and spatiotemporal information into a standardized data format; The heterogeneous data is defined as follows: assuming that the data sources in the low-altitude airspace include a static data set and a dynamic data set; Topological integration: For multi-source data distributed in space, a topological integration function is used to map data at different locations into a unified spatial coordinate system to form a spatial information matrix covering the entire area. Spatiotemporal information integration integrates dynamic data that changes over time into a topological structure in time series to form a spatiotemporal data set. The spatiotemporal integrated data is used as the input of the global model to support the intelligent scheduling of low-altitude aircraft and optimization of airspace resources in a time-varying environment. Data sharding divides the data of each participant into multiple shards according to content or spatial area. Data anonymization anonymizes the data of each shard and implements data privacy protection by adding differential privacy noise. Decentralized federated learning model aggregation realizes collaborative model updates among different participants, ensuring that the joint training of the global model is completed without leaving the local data; specifically, it includes: Local model training: Each participant trains its model parameters on the local data set to obtain updated local model parameters; Decentralized model aggregation: Decentralized federated learning achieves parameter aggregation through model parameter exchange between adjacent participants; Global model output: After completing multiple rounds of decentralized aggregation, a global model is generated. The global model aggregates the local models of each participant to form a joint model that contains the data features of multiple parties, thereby realizing cross-party collaborative modeling.

5. The low-altitude industrial chain information intelligent collaboration and airspace management platform based on federated learning according to claim 1 is characterized by: The analysis application modules include aircraft analysis, airspace analysis, environmental analysis and digital twin sub-modules; Aircraft analysis is used for flight safety assessment and path optimization, airspace analysis is used for airspace traffic management and segment speed control strategy, environmental analysis is used for wind field and meteorological condition prediction, and the digital twin submodule is used to build a virtual airspace environment to support airspace simulation and visualization; specifically, it includes: The aircraft analysis module evaluates the flight status and optimizes the flight path to ensure flight safety and minimize energy consumption during mission execution, including: Flight safety assessment: Real-time monitoring of the flight status of low-altitude aircraft to ensure that they fly within a safe range; Path optimization: Optimize the path according to mission requirements and environmental conditions, with the goal of minimizing energy consumption and flight risks; The airspace analysis module analyzes the traffic conditions in the low-altitude airspace, and realizes the reasonable allocation of airspace resources through traffic management and segment speed control strategies to ensure the orderly operation of different aircraft in the same airspace, including: Traffic management: optimize airspace traffic distribution, avoid route congestion, and ensure safe distance between aircraft; Segment speed control strategy: Based on the real-time airspace environment and mission requirements, set the optimal speed of the aircraft in each segment to avoid congestion or collision risks The environmental analysis module monitors and predicts the wind field and meteorological conditions in the airspace in real time, helping the aircraft to make necessary flight adjustments to ensure flight safety and mission stability, including: Wind field prediction: Use historical wind field data and real-time meteorological data to predict the wind speed and direction in low-altitude airspace to ensure stable flight of the aircraft under different wind conditions; Meteorological condition analysis: by analyzing environmental variables, predicting weather trends and providing early warning information for aircraft; The digital twin submodule realizes simulation and visualization support for low-altitude airspace by building a virtual airspace environment, so as to help managers intuitively understand the airspace status and provide decision support, including: Virtual airspace environment construction: Based on multi-source data fusion and federated learning model output, a virtual airspace environment model is constructed to simulate the airspace conditions of aircraft operations; Simulation and visualization: Through simulation in a virtual environment, the real-time operating status of the aircraft and airspace traffic conditions can be displayed, facilitating real-time monitoring and management of low-altitude airspace.

Images