Low-altitude airspace multi-source data fusion congestion prediction and dynamic scheduling system

Abstract

The application discloses a low-altitude airspace multi-source data fusion congestion prediction and dynamic scheduling system, which solves the problems of incomplete data fusion, low congestion prediction accuracy, lack of coordination and real-time in existing low-altitude airspace traffic management. The system includes a perception layer, a data fusion layer, a prediction layer, a decision-making and scheduling layer, a control execution layer, and a blockchain storage layer. The perception layer collects multi-source data, the data fusion layer realizes data fusion through an improved deep belief network, the prediction layer adopts a mixed model of LSTM and graph neural network to output congestion prediction results, the decision-making and scheduling layer generates a dynamic scheduling scheme based on a multi-objective optimization algorithm, the control execution layer executes the scheme and feeds back data, and the blockchain storage layer guarantees data security and privacy. The application improves the accuracy of congestion prediction and the adaptability of the scheduling scheme, and realizes efficient, safe and coordinated operation of the low-altitude airspace.

Description

Technical Field

[0001] This application relates to the field of low-altitude airspace traffic management technology, specifically to a congestion prediction and dynamic scheduling system based on multi-source data fusion in low-altitude airspace. Background Technology

[0002] With the rapid development of the low-altitude economy, the number of aircraft in low-altitude airspace has surged, encompassing various types such as drones, light helicopters, and general aviation aircraft, resulting in explosive growth in low-altitude traffic flow. As a crucial transportation resource, the efficient and safe management of low-altitude airspace directly impacts the sustainable development of the low-altitude economy.

[0003] Existing low-altitude airspace traffic management technologies have several shortcomings: First, data collection sources are limited, relying heavily on data reported by aircraft or from ground-based fixed sensors, lacking a comprehensive understanding of the overall airspace environment and dynamic traffic flow, resulting in insufficient data integrity. Second, data fusion methods are simplistic, often employing traditional weighted averages or simple splicing, making it difficult to effectively uncover deep correlations between multi-source data, affecting the accuracy of subsequent predictions and scheduling. Third, congestion prediction models often focus on temporal characteristics, ignoring the relationships between aircraft, facilities, and airspace nodes, leading to low prediction accuracy and insufficient lead time. Fourth, scheduling schemes are mostly local optimizations of a single airspace, lacking cross-airspace coordination mechanisms, easily causing airspace resource conflicts, and exhibiting poor adaptability in extreme scenarios. Fifth, data storage security and privacy protection are insufficient; multi-source data involves multiple participants, and the risk of data leakage or tampering affects scheduling reliability.

[0004] Therefore, there is an urgent need for a system that integrates multi-source data, accurately predicts congestion, enables collaborative dynamic scheduling, and ensures data security, in order to overcome the shortcomings of existing technologies. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a congestion prediction and dynamic scheduling system for low-altitude airspace based on multi-source data fusion, so as to solve the problems mentioned in the background technology.

[0006] According to one aspect of this application, a congestion prediction and dynamic scheduling system for low-altitude airspace multi-source data fusion includes a perception layer, a data fusion layer, a prediction layer, a decision scheduling layer, a control execution layer, and a blockchain storage layer, with each layer communicating with each other in sequence. The perception layer is used to collect low-altitude airspace multi-source data; the data fusion layer is used to fuse the multi-source data; the prediction layer is used to predict congestion based on the fused data; the decision scheduling layer is used to generate a dynamic scheduling scheme based on the prediction results; the control execution layer is used to execute the dynamic scheduling scheme and feed back execution data; and the blockchain storage layer is used to store data from each layer and provide a secure access mechanism.

[0007] Preferably, the perception layer includes a UAV perception module, an aircraft data acquisition module, a meteorological data acquisition module, a ground facility data acquisition module, and an airspace management data acquisition module. The UAV perception module is equipped with a high-definition camera, millimeter-wave radar, and miniature meteorological sensors. It covers the target low-altitude airspace according to a preset cruise path and collects the aircraft's position, speed, traffic density, and environmental parameters within the airspace. The aircraft data acquisition module collects its own flight parameters through airborne sensors. The meteorological data acquisition module obtains real-time meteorological data of the target airspace. The ground facility data acquisition module collects operational status data of airport runways, navigation equipment, and communication base stations. The airspace management data acquisition module obtains information on airspace planning, no-fly zones, and restricted areas.

[0008] Preferably, the data fusion layer includes a data preprocessing unit and an improved deep belief network fusion unit. The data preprocessing unit cleans, normalizes, and removes outliers from the multi-source data. The improved deep belief network fusion unit extracts features from the preprocessed multi-source data and performs layered fusion by adding an attention mechanism to a deep belief network, outputting fused feature data of a unified dimension.

[0009] Preferably, the prediction layer includes a temporal feature extraction unit, a spatial association modeling unit, and a congestion level determination unit. The temporal feature extraction unit uses an LSTM network to extract temporal features from the fused feature data. The spatial association modeling unit uses a graph neural network to model the association relationships between aircraft, facilities, and spatial nodes within the airspace. The congestion level determination unit combines the temporal features and spatial association features, and outputs the congestion prediction value and congestion level for a specific future period through a preset threshold system.

[0010] Preferably, the decision-making and scheduling layer includes a scheduling scheme generation unit, a multi-objective optimization unit, and a collaborative scheduling unit. The scheduling scheme generation unit selects a basic scheduling strategy from a preset strategy library based on the congestion level and formulates an initial scheduling scheme. The multi-objective optimization unit uses a hybrid algorithm that integrates reinforcement learning, genetic algorithm, and multi-objective particle swarm optimization to optimize the initial scheduling scheme. The collaborative scheduling unit performs cross-airspace collaborative scheduling for congestion situations in multiple adjacent airspaces to avoid airspace resource conflicts.

[0011] Preferably, the optimization objectives of the multi-objective optimization unit include flight safety, traffic efficiency, resource utilization, and carbon emission control. The hybrid algorithm adjusts the scheduling strategy through reinforcement learning feedback, achieves strategy diversity through genetic algorithm, and finds the optimal solution for multi-objective balance through multi-objective particle swarm optimization algorithm.

[0012] Preferably, the control execution layer includes an instruction issuing unit, an execution monitoring unit, and a feedback unit. The instruction issuing unit issues the dynamic scheduling plan to the aircraft, the airport control center, and the airspace management department through a high-speed communication module. The execution monitoring unit monitors the execution status of the plan in real time. The feedback unit feeds back the execution data and real-time airspace status data to the data fusion layer and the decision scheduling layer to form a closed-loop control.

[0013] Preferably, the blockchain storage layer includes a data encryption unit, a smart contract unit, and a privacy protection unit. The data encryption unit uses the SHA-256 algorithm to hash and encrypt the stored data. The smart contract unit presets data access permissions and scheduling policy execution rules, and automatically triggers data verification and policy updates. The privacy protection unit uses differential privacy technology to add noise of a corresponding magnitude according to the data sensitivity level to protect the privacy and security of multi-source data.

[0014] Preferably, the decision scheduling layer further includes an extreme scenario adaptation unit, which retrieves historical extreme scenario data from the blockchain storage layer, simulates the execution effect of the scheduling scheme under extreme weather and equipment failure, and adaptively adjusts the dynamic scheduling scheme to improve the system robustness.

[0015] A method for congestion prediction and dynamic scheduling based on multi-source data fusion in low-altitude airspace includes the following steps:

[0016] Step 1: The perception layer collects multi-source data from the low-altitude airspace and transmits the multi-source data to the data fusion layer;

[0017] Step 2: The data fusion layer preprocesses and fuses the multi-source data, outputs fused feature data, and transmits it to the prediction layer;

[0018] Step 3: The prediction layer extracts the temporal features and spatial correlation features of the fused feature data, outputs the congestion prediction value and congestion level, and transmits them to the decision scheduling layer;

[0019] Step 4: The decision-making and scheduling layer generates an initial scheduling plan based on the congestion level, and outputs a dynamic scheduling plan after multi-objective optimization and cross-space collaborative scheduling;

[0020] Step 5: The control execution layer issues dynamic scheduling plans and monitors the execution status, and feeds back execution data to the blockchain storage layer and decision scheduling layer;

[0021] Step 6: The blockchain storage layer stores data from each stage. The smart contract unit determines whether the scheduling plan needs to be adjusted based on the feedback data. If so, it triggers the decision scheduling layer to re-optimize the plan.

[0022] The advantages of this application compared to existing technologies are:

[0023] 1. By combining the UAV perception module with other data acquisition modules, comprehensive coverage of airspace, aircraft, environment, facilities, and management data is achieved, compensating for the shortcomings of single data sources and improving data integrity. An improved deep belief network with added attention mechanism effectively mines deep correlations between multi-source data, outputting high-quality fused feature data to lay the foundation for accurate prediction. Combining the temporal feature extraction capabilities of LSTM and the airspace correlation modeling capabilities of graph neural networks, both time and spatial dimensions are considered, improving congestion prediction accuracy and lead time. A multi-objective optimization algorithm achieves a balance between safety, efficiency, resources, and carbon emissions; cross-airspace collaborative scheduling avoids resource conflicts; and extreme scenario adaptation enhances system robustness. Blockchain technology ensures data immutability and traceability, while differential privacy technology protects the data privacy of participants, improving system reliability and trustworthiness. Real-time feedback of execution data and dynamic adjustment of scheduling schemes adapt to dynamic changes in low-altitude traffic flow, improving scheduling adaptability. Attached Figure Description

[0024] Figure 1 is a schematic diagram of the hierarchical architecture of the system of the present invention;

[0025] Figure 2 is a schematic diagram of the composition of the perception layer module of the system of the present invention;

[0026] Figure 3 is a schematic diagram of the prediction layer workflow of the system of the present invention;

[0027] Figure 4 is a schematic diagram of the operation flow of the system of the present invention. Detailed Implementation

[0028] To make the content of this application easier to understand, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0029] like Figures 1-4 As shown, a low-altitude airspace multi-source data fusion congestion prediction and dynamic scheduling system includes a perception layer, a data fusion layer, a prediction layer, a decision scheduling layer, a control execution layer, and a blockchain storage layer. Each layer is connected in sequence to form a closed-loop management system for the entire process.

[0030] The perception layer is used to collect multi-source data in low-altitude airspace to achieve comprehensive perception of airspace status. It includes UAV perception module, aircraft data acquisition module, meteorological data acquisition module, ground facility data acquisition module, and airspace management data acquisition module.

[0031] The UAV's perception module is equipped with a high-definition camera, millimeter-wave radar, and miniature weather sensor, covering the target's low-altitude airspace according to a preset cruise path. The cruise path is planned using airspace geographic information and historical congestion hotspots, employing an "S"-shaped fixed-point cruise mode to ensure no blind spots. The high-definition camera captures the aircraft's external features and position distribution, the millimeter-wave radar achieves high-precision distance and speed measurements, and the miniature weather sensor collects environmental parameters such as visibility and wind speed within the airspace. An embedded computing platform performs preliminary data processing, filtering out obviously noisy data through thresholding to reduce the burden on subsequent data transmission.

[0032] The aircraft data acquisition module collects parameters such as altitude, heading, remaining range, and flight attitude of the aircraft through onboard GPS, inertial navigation sensors, and other equipment. The data acquisition frequency is adapted to the flight status, increasing to 20Hz during high-speed flight and maintaining 10Hz during low-speed cruise.

[0033] The meteorological data acquisition module connects to the API interface of regional meteorological stations to obtain real-time meteorological data such as precipitation, visibility, and turbulence intensity in the target airspace. The data update cycle is 5 minutes, which is shortened to 1 minute under extreme weather conditions, providing environmental reference for congestion prediction and scheduling scheme formulation.

[0034] The ground infrastructure data acquisition module collects data such as runway occupancy status, navigation equipment operating parameters, and communication base station coverage through sensors deployed at airports and navigation stations, and monitors the ground infrastructure's ability to support low-altitude flight in real time.

[0035] The airspace management data acquisition module obtains management data such as airspace planning schemes, no-fly zone boundary coordinates, and restricted zone usage periods from the airspace management system to ensure that the scheduling scheme complies with airspace management regulations.

[0036] The data fusion layer is used to fuse multi-source data, uncover deep data correlations, and provide high-quality data support for subsequent predictions. It includes a data preprocessing unit and an improved deep belief network fusion unit.

[0037] The data preprocessing unit adopts a three-level processing flow: First, outliers in multi-source data are identified and removed using the Isolation Forest algorithm. This algorithm detects outliers based on differences in data density and is suitable for unstructured multi-source data. Second, the min-max normalization algorithm maps data of different magnitudes to the same interval, eliminating the impact of differences in dimensions on the fusion effect. Finally, a linear interpolation algorithm is used to fill in missing data to ensure data integrity.

[0038] The improved deep belief network fusion unit adds an attention mechanism to the traditional deep belief network to enhance the weight allocation of key data features. This unit categorizes preprocessed multi-source data into four types: aircraft data, environmental data, facility data, and airspace management data, which are then input into the input layer of the improved deep belief network. Unsupervised feature learning is performed using a multi-layer restricted Boltzmann machine. The attention mechanism dynamically adjusts the weights based on the data's contribution to congestion prediction. Finally, the output layer outputs fused feature data with a unified dimension of 128, balancing feature expressiveness and computational efficiency.

[0039] The prediction layer is used to predict congestion based on the fused data, taking into account both temporal and spatial correlation features to improve prediction accuracy. It includes a temporal feature extraction unit, a spatial correlation modeling unit, and a congestion level determination unit.

[0040] The temporal feature extraction unit employs an LSTM network, which effectively captures long-term temporal dependencies in the fused feature data through gating mechanisms including input gates, forget gates, and output gates. The network has two hidden layers, each with 64 neurons, and a time step of 12, corresponding to 60 minutes of historical data input, enabling it to fully learn the traffic variation patterns at different times.

[0041] The airspace association modeling unit uses a graph neural network to construct an airspace association model, treating aircraft, ground facilities, and airspace nodes as graph nodes, and flight paths, communication links, and distance relationships between nodes as edges. Airspace association features are extracted through graph convolution operations to uncover the mutual influences between aircraft and facilities, and between aircraft themselves, thus overcoming the shortcomings of traditional models that only focus on temporal features.

[0042] The congestion level determination unit combines temporal and spatial correlation features, inputting both types of features into a pre-defined fully connected layer and outputting a congestion prediction value for a specific future time period. A pre-defined four-level threshold system is used: mild congestion corresponds to a flow rate of no more than 80 flights per hour; moderate congestion corresponds to a flow rate of more than 80 flights per hour but no more than 120 flights per hour; severe congestion corresponds to a flow rate of more than 120 flights per hour but no more than 160 flights per hour; and extreme congestion corresponds to a flow rate of more than 160 flights per hour. The corresponding congestion level is output based on the predicted value.

[0043] The decision-making and scheduling layer is used to generate dynamic scheduling schemes based on prediction results, realize multi-objective optimization and cross-space domain collaboration, and includes a scheduling scheme generation unit, a multi-objective optimization unit, a collaborative scheduling unit, and an extreme scenario adaptation unit.

[0044] The scheduling scheme generation unit selects basic scheduling strategies from a preset strategy library based on the congestion level. The preset strategy library contains 20 basic strategies, including route adjustment, takeoff and landing sequence optimization, airspace resource allocation, and flight speed restrictions. Corresponding strategy combinations are matched to different congestion levels; for example, a flight speed optimization strategy is used for light congestion, while a combination strategy of route adjustment and staggered takeoff and landing is used for heavy congestion to formulate an initial scheduling scheme.

[0045] The multi-objective optimization unit employs a hybrid algorithm integrating reinforcement learning, genetic algorithms, and multi-objective particle swarm optimization. Optimization objectives include flight safety, traffic efficiency, resource utilization, and carbon emission control. The reinforcement learning agent interacts with the simulated airspace environment, using safety risk reduction, delay reduction, resource utilization improvement, and carbon emission reduction as reward signals to dynamically adjust the scheduling strategy. The genetic algorithm ensures strategy diversity through selection, crossover, and mutation operations, with a population size of 50 and 30 iterations. The multi-objective particle swarm optimization algorithm finds the optimal solution for multi-objective balance through particle position updates, with a particle count of 40, ensuring the scheduling scheme achieves optimal balance across multiple objective dimensions.

[0046] The collaborative scheduling unit establishes an airspace correlation matrix to quantify the mutual influence coefficient of traffic flow between multiple adjacent airspaces, addressing congestion situations. Based on this correlation matrix, cross-airspace route coordination and staggered takeoff and landing times are implemented to avoid simultaneous traffic peaks in adjacent airspaces and reduce airspace resource conflicts.

[0047] The extreme scenario adaptation unit retrieves historical extreme scenario data from the blockchain storage layer, including operational data from scenarios such as extreme rainstorms, strong winds, and navigation equipment failures. It simulates the execution effect of the current scheduling scheme under extreme scenarios and statistically analyzes the proportion of cases with serious operational problems. If the proportion exceeds a preset threshold of 15%, successful coping strategies are extracted from historical cases and adaptively adjusted in conjunction with the key parameters of the current scheme to improve the system's robustness under extreme scenarios.

[0048] The control execution layer is used to execute dynamic scheduling schemes and feed back execution data to form closed-loop control. It includes an instruction issuance unit, an execution monitoring unit, and a feedback unit.

[0049] The command issuance unit employs 5G and V2X high-speed communication modules to ensure low-latency transmission of scheduling commands, with transmission latency controlled within 100 milliseconds. Dynamic scheduling schemes are encapsulated in a standardized format and issued to aircraft, airport control centers, and airspace management departments, supporting multiple communication protocols adapted to different devices.

[0050] The execution monitoring unit uses the UAV sensing module and ground sensors to monitor the aircraft's execution of the scheduling plan in real time, collecting data such as actual flight trajectory, take-off and landing time, and changes in airspace traffic. The data collection frequency is 10Hz to ensure the real-time nature and accuracy of the monitoring data.

[0051] The feedback unit transmits execution data and real-time spatial status data to the data fusion layer and the decision-scheduling layer, forming a closed-loop control. A preset execution deviation threshold of 10% is set. If the deviation between the actual data and the predicted data exceeds this threshold, the decision-scheduling layer is triggered to re-optimize the scheduling scheme to ensure that the scheduling effect meets expectations.

[0052] The blockchain storage layer is used to store data from various layers and provide a secure access mechanism to ensure data immutability and privacy. It includes data encryption units, smart contract units, and privacy protection units.

[0053] The data encryption unit uses the SHA-256 algorithm to hash and encrypt the stored data, generating a unique hash value. Any modification to the data will cause the hash value to change, ensuring that the data is tamper-proof. Simultaneously, end-to-end encryption is performed on transmitted data to prevent data theft during transmission.

[0054] The smart contract unit pre-sets data access permission rules and scheduling strategy execution rules. Data access permissions are allocated according to the type of participant, with airspace management departments having the highest access permissions, and aircraft operating companies only being able to access their own relevant data; the scheduling strategy execution rules clearly define the execution time limit and feedback requirements after the plan is issued, and an early warning mechanism is automatically triggered if the plan is not executed within the time limit to ensure the effective implementation of the scheduling plan.

[0055] The privacy protection unit employs differential privacy technology, determining the noise level based on the data's sensitivity level. Data is categorized into high, medium, and low sensitivity levels: commercial aircraft operation data is classified as high sensitivity with an added noise level of 0.05; meteorological data is classified as low sensitivity with an added noise level of 0.01; and the remaining data is classified as medium sensitivity with an added noise level of 0.03. This noise-adding process perturbs the original data, protecting the privacy of participating parties without affecting data usability.

[0056] This invention also provides a method for congestion prediction and dynamic scheduling based on multi-source data fusion in low-altitude airspace, applied to the aforementioned system, comprising the following steps:

[0057] Step 1: Each module in the perception layer collects multi-source data of the low-altitude airspace at a preset collection frequency. After the embedded computing platform of the UAV perception module performs preliminary noise reduction processing on the collected data, it is transmitted to the data fusion layer along with the data collected by other modules.

[0058] Step 2: The preprocessing unit of the data fusion layer performs outlier removal, normalization, and missing value imputation on the multi-source data. The improved deep belief network fusion unit performs feature extraction and layered fusion on the preprocessed data, outputting fused feature data, which is then transmitted to the prediction layer.

[0059] Step 3: The temporal feature extraction unit of the prediction layer extracts the temporal features of the fused feature data through the LSTM network, the spatial domain association modeling unit constructs the spatial domain association graph through the graph neural network and extracts the spatial domain association features, and the congestion level determination unit combines the two types of features to output the congestion prediction value and congestion level for a specific period in the future, which is then transmitted to the decision scheduling layer.

[0060] Step 4: The scheduling scheme generation unit of the decision scheduling layer selects the basic scheduling strategy based on the congestion level and formulates the initial scheduling scheme. The multi-objective optimization unit optimizes the initial scheme, the collaborative scheduling unit performs cross-space collaborative adjustment, and the extreme scenario adaptation unit performs extreme scenario adaptability verification on the scheme. Finally, the dynamic scheduling scheme is output.

[0061] Step 5: The command issuing unit of the control execution layer issues the dynamic scheduling plan to each execution entity through the high-speed communication module. The execution monitoring unit monitors the execution status of the plan in real time. The feedback unit feeds back the execution data and real-time airspace status data to the blockchain storage layer and the decision scheduling layer.

[0062] Step 6: The blockchain storage layer stores data from each stage. The smart contract unit determines whether the execution deviation exceeds the preset threshold based on the feedback data. If it does, the decision scheduling layer is triggered to re-optimize the plan, forming a closed-loop control.

[0063] The above embodiments are only used to illustrate the technical solutions of the embodiments of this application, and are not intended to limit them. Although the embodiments of this application have been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features, without departing from the spirit and scope defined by the claims of this application.

Claims

1. A congestion prediction and dynamic scheduling system for low-altitude airspace based on multi-source data fusion, characterized in that, It includes a perception layer, a data fusion layer, a prediction layer, a decision scheduling layer, a control execution layer, and a blockchain storage layer, with each layer communicating with each other in sequence. The perception layer is used to collect multi-source data in low-altitude airspace. The data fusion layer is used to fuse the multi-source data. The prediction layer is used to predict congestion based on the fused data. The decision scheduling layer is used to generate a dynamic scheduling scheme based on the prediction results. The control execution layer is used to execute the dynamic scheduling scheme and feed back the execution data. The blockchain storage layer is used to store the data of each layer and provide a secure access mechanism.

2. The low-altitude airspace multi-source data fusion congestion prediction and dynamic scheduling system according to claim 1, characterized in that, The perception layer includes a UAV perception module, an aircraft data acquisition module, a meteorological data acquisition module, a ground facility data acquisition module, and an airspace management data acquisition module. The UAV perception module is equipped with a high-definition camera, millimeter-wave radar, and miniature meteorological sensors. It covers the target low-altitude airspace according to a preset cruise path and collects the aircraft's position, speed, traffic density, and environmental parameters within the airspace. The aircraft data acquisition module collects its own flight parameters through airborne sensors. The meteorological data acquisition module obtains real-time meteorological data of the target airspace. The ground facility data acquisition module collects operational status data of airport runways, navigation equipment, and communication base stations. The airspace management data acquisition module obtains information on airspace planning, no-fly zones, and restricted areas.

3. The low-altitude airspace multi-source data fusion congestion prediction and dynamic scheduling system according to claim 1, characterized in that, The data fusion layer includes a data preprocessing unit and an improved deep belief network fusion unit. The data preprocessing unit cleans, normalizes, and removes outliers from the multi-source data. The improved deep belief network fusion unit extracts features and performs layered fusion on the preprocessed multi-source data through a deep belief network with an added attention mechanism, outputting fused feature data with a unified dimension.

4. The low-altitude airspace multi-source data fusion congestion prediction and dynamic scheduling system according to claim 1, characterized in that, The prediction layer includes a temporal feature extraction unit, a spatial association modeling unit, and a congestion level determination unit. The temporal feature extraction unit uses an LSTM network to extract temporal features from the fused feature data. The spatial association modeling unit uses a graph neural network to model the association relationships between aircraft, facilities, and spatial nodes within the airspace. The congestion level determination unit combines the temporal features and spatial association features, and outputs the congestion prediction value and congestion level for a specific future period through a preset threshold system.

5. The low-altitude airspace multi-source data fusion congestion prediction and dynamic scheduling system according to claim 1, characterized in that, The decision-making and scheduling layer includes a scheduling scheme generation unit, a multi-objective optimization unit, and a collaborative scheduling unit. The scheduling scheme generation unit selects a basic scheduling strategy from a preset strategy library based on the congestion level and formulates an initial scheduling scheme. The multi-objective optimization unit uses a hybrid algorithm that integrates reinforcement learning, genetic algorithm, and multi-objective particle swarm optimization to optimize the initial scheduling scheme. The collaborative scheduling unit performs cross-airspace collaborative scheduling for congestion situations in multiple adjacent airspaces to avoid airspace resource conflicts.

6. The low-altitude airspace multi-source data fusion congestion prediction and dynamic scheduling system according to claim 5, characterized in that, The optimization objectives of the multi-objective optimization unit include flight safety, traffic efficiency, resource utilization, and carbon emission control. The hybrid algorithm adjusts the scheduling strategy through reinforcement learning feedback, achieves strategy diversity through genetic algorithm, and finds the optimal solution for multi-objective balance through multi-objective particle swarm optimization algorithm.

7. The low-altitude airspace multi-source data fusion congestion prediction and dynamic scheduling system according to claim 1, characterized in that, The control execution layer includes an instruction issuing unit, an execution monitoring unit, and a feedback unit. The instruction issuing unit issues dynamic scheduling plans to aircraft, airport control centers, and airspace management departments through a high-speed communication module. The execution monitoring unit monitors the execution status of the plan in real time. The feedback unit feeds back execution data and real-time airspace status data to the data fusion layer and the decision scheduling layer to form a closed-loop control.

8. The low-altitude airspace multi-source data fusion congestion prediction and dynamic scheduling system according to claim 1, characterized in that, The blockchain storage layer includes a data encryption unit, a smart contract unit, and a privacy protection unit. The data encryption unit uses the SHA-256 algorithm to hash and encrypt the stored data. The smart contract unit presets data access permissions and scheduling policy execution rules, and automatically triggers data verification and policy updates. The privacy protection unit uses differential privacy technology to add noise of a corresponding magnitude according to the data sensitivity level to protect the privacy and security of multi-source data.

9. The low-altitude airspace multi-source data fusion congestion prediction and dynamic scheduling system according to claim 1, characterized in that, The decision-making and scheduling layer also includes an extreme scenario adaptation unit. This unit retrieves historical extreme scenario data from the blockchain storage layer, simulates the execution effect of the scheduling scheme under extreme weather and equipment failure, and makes adaptive adjustments to the dynamic scheduling scheme to improve the system robustness.

10. A method for congestion prediction and dynamic scheduling based on multi-source data fusion in low-altitude airspace, characterized in that, Applied to the system according to any one of claims 1-9, the method includes the following steps: Step 1: The perception layer collects multi-source data from the low-altitude airspace and transmits the multi-source data to the data fusion layer; Step 2: The data fusion layer preprocesses and fuses the multi-source data, outputs fused feature data, and transmits it to the prediction layer; Step 3: The prediction layer extracts the temporal features and spatial correlation features of the fused feature data, outputs the congestion prediction value and congestion level, and transmits them to the decision scheduling layer; Step 4: The decision-making and scheduling layer generates an initial scheduling plan based on the congestion level, and outputs a dynamic scheduling plan after multi-objective optimization and cross-space collaborative scheduling; Step 5: The control execution layer issues dynamic scheduling plans and monitors the execution status, and feeds back execution data to the blockchain storage layer and decision scheduling layer; Step 6: The blockchain storage layer stores data from each stage. The smart contract unit determines whether the scheduling plan needs to be adjusted based on the feedback data. If so, it triggers the decision scheduling layer to re-optimize the plan.

Images