An air space network flight simulation system

Abstract

The application relates to the technical field of unmanned aerial vehicle air traffic management, and discloses an airspace network flight simulation system, which comprises a multi-source data fusion module, a height section graph generation module, an algorithm selection module, a path planning module, a cooperative scheduling module and a visualization module, and a simulation verification module. The multi-source data fusion module dynamically integrates geographic information data by adopting a time-weighted fusion algorithm. The height section graph generation module adaptively adjusts a grid resolution and introduces a safety buffer zone to generate an obstacle distribution graph. The algorithm selection module intelligently matches an optimal path algorithm based on a machine learning prediction model. The path planning module generates a high-quality trajectory by adopting a hierarchical strategy in combination with global planning and local optimization. The cooperative scheduling module constructs a distributed architecture and guarantees the safe cooperation of multiple machines by means of predictive conflict detection and negotiation solution strategies. The visualization module provides an augmented reality interface. The simulation verification module verifies in a digital twin environment. The application provides a reliable solution for urban low-altitude traffic management.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) air traffic management technology, and more specifically, to an airspace network flight simulation system. Background Technology

[0002] With the rapid development of drone technology and the promotion of low-altitude economic policies, drones are increasingly being used in urban logistics, emergency rescue, and infrastructure inspection. Especially in core urban areas, multiple logistics companies simultaneously operate drone delivery services, creating a high-density demand for airspace usage. However, the urban environment is characterized by dense buildings, complex aerial obstacles, and numerous no-fly zones, placing extremely high demands on drone path planning and flight safety.

[0003] Current UAV air traffic management systems primarily rely on traditional 3D path planning technology, but these technologies exhibit numerous problems when facing complex urban environments. Existing UAV path planning systems directly search for paths in 3D space. When the obstacle density in the environment is high, the search space grows cubically, leading to excessively high computational complexity and failing to meet real-time requirements. In high-density urban flight scenarios, it is necessary to plan non-conflicting paths for multiple UAVs simultaneously. Existing technologies typically employ sequential processing, where time complexity increases linearly with the number of UAVs, severely impacting system practicality.

[0004] Existing systems lack effective multi-algorithm integration mechanisms, making it impossible to select the optimal algorithm for different scenarios, thus limiting the system's versatility and efficiency. Traditional systems offer limited visualization capabilities, making it difficult for users to intuitively understand path planning results, especially in complex 3D environments, hindering the effective verification of path safety and rationality. Furthermore, existing technologies lack comprehensive simulation verification capabilities, failing to adequately test the feasibility of path planning before actual deployment, posing significant safety risks. Therefore, there is an urgent need to develop an efficient, safe, and reliable airspace network flight simulation system to address these technical challenges. Summary of the Invention

[0005] This invention provides an airspace network flight simulation system that solves the technical problems of traditional UAV path planning calculations being complex, having poor real-time performance, and being difficult to coordinate with multiple UAVs.

[0006] This invention provides an airspace network flight simulation system, comprising: The multi-source data fusion module uses a time-weighted fusion algorithm to intelligently fuse multi-source data and output a geographic feature database. The height section map generation module receives a geographic feature database, adaptively adjusts the raster resolution according to environmental complexity, and generates a multi-level obstacle distribution map. The algorithm selection module establishes an environmental feature extraction model based on the obstacle distribution map and automatically selects the optimal path planning algorithm. The path planning module adopts a hierarchical planning strategy based on the selected path planning algorithm, generates a global path framework on a coarse-grained cross-section map, performs local path optimization within the flight corridor, generates a Pareto optimal solution using multi-objective optimization, and outputs the flight trajectory. The collaborative scheduling module receives the flight trajectory and establishes a distributed collaborative architecture. It selects a leader node based on the node priority calculation formula, identifies potential conflicts using a predictive conflict detection algorithm, and executes avoidance actions through a collaborative scheduling solution strategy. The visualization module uses adaptive rendering technology based on the collaborative scheduling results, and introduces augmented reality technology to overlay virtual trajectories; The simulation verification module receives visualized data to build a digital twin environment and introduces artificial intelligence for intelligent verification and system optimization.

[0007] In a preferred embodiment, the multi-source data fusion module includes: The data preprocessing layer standardizes the format of data from different sources, converts all spatial data into the WGS84 geographic coordinate system, and converts elevation data into an altitude benchmark. The time-weighted fusion layer uses a time-weighted fusion algorithm to intelligently fuse multi-source data. The algorithm assigns time weights based on the timeliness of the data. The time weights are calculated using an exponential decay function, and the weight values ​​decrease exponentially as the time difference increases. The time decay constant is determined based on the data type. A smaller decay constant is set for rapidly changing dynamic data, and a larger decay constant is set for relatively stable static data. The data validation layer establishes a data consistency verification mechanism, identifies inconsistent data sources through cross-validation algorithms, and employs a majority voting mechanism for conflicting data.

[0008] In a preferred embodiment, the height section map generation module includes: The environmental complexity assessment unit establishes an environmental complexity assessment model, quantifies environmental complexity by analyzing indicators such as building density, height variation gradient, and obstacle distribution uniformity, and classifies the environment into three levels: simple, medium, and complex based on the assessment results. The adaptive raster resolution determination unit adjusts the raster resolution according to the complexity of the environment. It uses high-resolution raster in complex areas with dense buildings, low-resolution raster in simple areas with sparse buildings, and medium-resolution raster in medium-complexity areas. It establishes resolution transition areas to avoid path discontinuity caused by sudden resolution changes. The multi-level section generation unit uses multi-level section technology to generate obstacle distribution maps at different height levels and introduces the concept of a safety buffer zone, setting a safety buffer distance around the obstacles.

[0009] In a preferred embodiment, the algorithm selection module includes: The environmental feature extraction unit establishes an environmental feature extraction model, automatically analyzes the key feature parameters of the current environment, and combines obstacle density, spatial connectivity, path complexity, and dynamic feature parameters into a four-dimensional feature vector. The algorithm performance prediction unit constructs an algorithm performance prediction model based on machine learning, trains the performance prediction model using the random forest algorithm, uses environmental feature vectors as input to predict the performance of each algorithm, and establishes an algorithm applicability scoring mechanism that comprehensively considers prediction performance, computing resource status and task priority to calculate the applicability score. The intelligent algorithm selection unit designs an intelligent algorithm selection strategy, automatically selects the optimal algorithm based on task characteristics and environmental conditions, establishes an algorithm combination strategy, and runs multiple algorithms simultaneously for particularly complex scenarios and selects the optimal result.

[0010] In a preferred embodiment, the path planning module includes: The global path planning unit adopts a hierarchical planning strategy to obtain flight corridors by performing global path planning on a coarse-grained cross-sectional map. An improved A-Star algorithm is used for global search. The heuristic function adopts a weighted combination of Euclidean distance and Manhattan distance to establish the concept of flight corridors and expand to both sides on the basis of global path to form a search space. The local path optimization unit performs local path optimization on the fine-grained cross-sectional map to generate a detailed flight trajectory. The Theta algorithm is used for local optimization to generate a smoother trajectory. A path smoothing mechanism is established, and the path is smoothed using a cubic spline interpolation algorithm. The multi-objective optimization unit establishes a multi-objective optimization evaluation system that comprehensively considers multiple objectives such as path length, flight time, energy consumption, and safety, and uses a non-dominated sorting genetic algorithm to generate a Pareto optimal solution set.

[0011] In a preferred embodiment, the cooperative scheduling module includes: The distributed collaborative architecture establishment unit establishes a distributed collaborative scheduling architecture. Each UAV acts as an intelligent node with local decision-making capabilities, while also accepting coordination and management from the central scheduling system. A hierarchical communication protocol is designed, and a dynamic allocation mechanism for node roles is established. The predictive conflict detection unit establishes a four-dimensional conflict detection model, adding a time dimension to the three-dimensional spatial coordinates. The trajectory prediction adopts a kinematic model, predicting the future position based on the current position, velocity, and acceleration, and using a cylindrical protection zone model to determine conflicts. The negotiation-based conflict resolution unit adopts a negotiation-based conflict resolution strategy. When a conflict is detected, the relevant UAVs negotiate through communication to determine the optimal avoidance plan. Multiple avoidance strategies are designed, including altitude adjustment, speed adjustment, path deviation, and waiting strategies.

[0012] In a preferred embodiment, the visualization module includes: The adaptive rendering optimization unit uses adaptive rendering technology to dynamically adjust the rendering quality based on device performance and network conditions. It establishes a device performance evaluation model and evaluates the device level by detecting the device's processor performance, graphics processing capability, and memory capacity parameters. It designs three rendering levels: high quality, medium quality, and low quality. A multi-level visualization system unit is established to support seamless switching from macro-level city-level views to micro-level single-machine trajectory views. The view hierarchy structure is designed to achieve smooth switching animations between different levels and dynamically adjust information density according to the view hierarchy. The augmented reality overlay unit introduces augmented reality technology to overlay virtual flight trajectories onto the real environment. It adopts feature point-based tracking technology to achieve virtual-real registration and designs an intelligent overlay strategy to adjust the display effect of virtual objects according to ambient lighting and background complexity.

[0013] In a preferred embodiment, the simulation verification module includes: The high-fidelity environment modeling unit constructs a high-fidelity digital twin environment, establishes a detailed three-dimensional geographic model, integrates real-time weather data, wind field information, and dynamic environmental elements, and uses a physics engine to simulate real aerodynamic effects. A multi-level simulation verification unit is established to create a multi-level simulation verification system, including single-machine simulation verification, multi-machine collaborative simulation verification, and system-level simulation verification. A simulation state management mechanism is established, and a hierarchical experimental design method is adopted. The intelligent verification and optimization unit introduces artificial intelligence technology for intelligent verification and system optimization, establishes an intelligent analysis pipeline, uses machine learning algorithms to identify abnormal patterns in simulation, and establishes a comparative analysis model between simulation data and real flight data.

[0014] In a preferred embodiment, the intelligent verification and optimization unit further includes: The abnormal behavior recognition subunit uses deep learning algorithms to automatically identify abnormal behaviors and potential risks during the simulation process, including trajectory anomalies, performance degradation, and safety hazards. The automatic optimization subunit automatically adjusts algorithm parameters, optimizes path planning strategies, and improves collaborative scheduling schemes based on verification results, while maintaining historical optimization records to track optimization effects. The credibility assessment subunit comprehensively considers simulation accuracy, verification coverage, and consistency verification factors to provide a credibility score for the simulation results. It also establishes a feedback mechanism to feed back the verification and optimization results to the preceding steps, forming a closed-loop optimization process.

[0015] A computer-readable storage medium for storing computer-readable instructions that, when read by a computer, enable the execution of the aforementioned airspace network flight simulation system.

[0016] The beneficial effects of this invention are as follows: The innovative height-section plot technique significantly reduces computational complexity. Traditional 3D path planning has a cubic computational complexity, while this invention reduces it to a quadratic complexity, shortening path planning time in complex urban environments from minutes to seconds, improving computational efficiency by more than tenfold. This dimensionality reduction not only maintains the accuracy of path planning but also improves the system's real-time response capability, meeting the needs of rapid path planning in dynamic environments. The multi-algorithm integration framework and intelligent selection mechanism significantly improve the system's environmental adaptability and planning quality. The system can automatically select the optimal algorithm based on environmental characteristics, achieving best performance in environments of varying complexity and greatly increasing the success rate of path planning. Compared to single-algorithm systems, this invention improves path quality, computational efficiency, and environmental adaptability, providing flexible solutions for different application scenarios. Attached Figure Description

[0017] Figure 1 This is a block diagram of an airspace network flight simulation system according to the present invention; Figure 2 This is a comparison chart of the multi-source data fusion effects of the present invention; Figure 3 This is a performance comparison chart of the path planning algorithm of this invention; Figure 4 This is a diagram analyzing the multi-machine collaborative scheduling effect of the present invention. Detailed Implementation

[0018] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, some features described in the examples may be combined in other examples.

[0019] At least one embodiment of the present invention discloses an airspace network flight simulation system, such as Figure 1 As shown, it includes: The multi-source data fusion module uses a time-weighted fusion algorithm to intelligently fuse multi-source data and output a geographic feature database. Specifically, the following steps are included: Data preprocessing involves inputting satellite remote sensing data, UAV aerial photography data, ground sensor data, building vector data, and terrain elevation data. The system establishes a multi-layered data fusion architecture, including a data preprocessing layer, a data fusion layer, and a data verification layer. In the data preprocessing layer, data from different sources are standardized in format, converting all spatial data to the WGS84 geographic coordinate system, and elevation data to an altitude benchmark. For satellite remote sensing data, geometric and radiometric corrections are performed to eliminate atmospheric effects and sensor errors. Geometric correction uses a multinomial transformation model, establishing a mapping relationship between pixel coordinates and geographic coordinates through ground control points. Radiometric correction uses an atmospheric transmission model to eliminate the effects of atmospheric scattering and absorption. For UAV aerial photography data, a high-precision digital surface model is generated through feature point matching and stereo vision technology. Feature point matching uses a scale-invariant feature transformation algorithm to extract key points, and stereo vision technology calculates depth information using binocular parallax. For ground sensor data, noise filtering and outlier detection are performed. Noise filtering uses a Kalman filter algorithm, and outlier detection uses a three-standard-deviation criterion based on statistical distribution to ensure data quality.

[0020] Output standardized multi-source geographic information datasets (including building height vector data, terrain elevation raster data, obstacle location coordinate data) and data quality assessment reports (including spatial accuracy indicators, temporal consistency indicators, and data integrity indicators). Time-weighted fusion is implemented by taking a standardized multi-source geographic information dataset, data timestamps, data accuracy identifiers, and reliability scores as input. A time-weighted fusion algorithm is used to intelligently fuse the multi-source data. The algorithm assigns time weights based on the timeliness of the data. These time weights are calculated using an exponential decay function, with the weight value decreasing exponentially as the time difference increases. The time decay constant is determined based on the data type; a smaller decay constant is set for rapidly changing dynamic data, and a larger decay constant is set for relatively stable static data. Accuracy weights are also assigned based on data accuracy. Accuracy weights are inversely proportional to the standard error of the data; high-precision data has a smaller standard error and therefore receives a higher weight. A data consistency verification mechanism is established. A cross-validation algorithm is used to identify inconsistent data sources. Cross-validation employs a leave-one-out method, using each data source as the validation set and the remaining data sources as the training set. Prediction errors are calculated to assess data consistency. For conflicting data, a majority voting mechanism is used: if a majority of data sources point to the same result, that result is adopted; otherwise, an expert system is used for adjudication.

[0021] Outputs include a fusion weight matrix (containing time weight coefficients and accuracy weight coefficients for each data source), a consistency verification report (containing data conflict identification results and fusion accuracy evaluation indicators), and a geographic feature database in a unified coordinate system. Dynamic update is triggered by inputting the fusion weight matrix, real-time monitoring data, historical benchmark data, and change threshold parameters; Establish a dynamic update triggering mechanism based on change detection. The system continuously monitors data changes in key areas, and automatically triggers a local data update process when a building height change exceeds a set threshold or a new obstacle is detected. An incremental update strategy is adopted, dividing the update area into several grid cells, and only reprocessing the grid cells that have changed, avoiding redundant calculations of the entire dataset. An update priority queue is established to determine the update order based on the importance and degree of change of the area, prioritizing data updates for high-risk and frequently used areas.

[0022] Output an updated area identification map (including change detection results and updated area boundary coordinates), a priority update queue (including update task sequence and timestamp information), and an incremental update data package (including newly added building height data, modified terrain elevation data, and deleted obstacle information). The height section map generation module receives a geographic feature database, adaptively adjusts the raster resolution according to environmental complexity, and generates a multi-level obstacle distribution map. Specifically, the following steps are included: Environmental complexity assessment involves inputting fused geographic information data, updated regional identification maps, flight area boundaries, and mission type parameters. An environmental complexity assessment model is established, quantifying environmental complexity through analysis of indicators such as building density, altitude gradient, and obstacle distribution uniformity. Building density is calculated as the proportion of building area per unit area, using a grid statistical method. The study area is divided into regular grids, and the ratio of building coverage area to the total grid area within each grid is calculated. Altitude gradient is obtained by calculating the standard deviation of altitude differences between adjacent areas. A sliding window technique is used, establishing a 3x3 neighborhood window centered on each grid, and calculating the standard deviation of altitude values ​​within the window as the altitude gradient for that grid. Obstacle distribution uniformity is assessed using spatial autocorrelation analysis, with the Moran's index measuring spatial correlation. A Moran's index close to zero indicates a random distribution, close to positive one indicates a clustered distribution, and close to negative one indicates a discrete distribution. Based on the assessment results, the environment is divided into three levels: simple, medium, and complex. Simple environment has a building density of less than 30% and a height change gradient of less than 10 meters. Complex environment has a building density of more than 70% or a height change gradient of more than 30 meters. Medium environment is in between. This provides a basis for subsequent adaptive processing.

[0023] Output an environmental complexity level map (including building density distribution data, height change gradient distribution data, and obstacle distribution uniformity index) and a complexity assessment parameter table (including the complexity level, quantification score value, and classification threshold parameter for each grid cell). The adaptive raster resolution is determined by taking into account the input environment complexity level map, computational resource constraints, and accuracy requirements. The raster resolution is adaptively adjusted based on environmental complexity. In densely built, complex areas, a high-resolution raster with a grid size of one to two meters is used to ensure accurate identification of small obstacles. In sparsely built, simple areas, a low-resolution raster with a grid size of five to ten meters is used to reduce computation and improve processing efficiency. In moderately complex areas, a moderate-resolution raster with a grid size of three to four meters is used to balance accuracy and efficiency. Resolution transition regions are established, using a gradual transition at the boundaries of different resolution regions to avoid path discontinuities caused by abrupt resolution changes.

[0024] Output adaptive raster resolution distribution map (including raster size parameters and resolution level labels for each region), resolution transition region labels (including transition region boundary coordinates and gradient function parameters), and raster grid coordinate index table. Multi-layered section generation, input adaptive grid resolution distribution map, cruise altitude, and safety buffer distance processing: Multi-layered sectioning technology is used to generate obstacle distribution maps at different height levels. The main section height is set to a predetermined cruise altitude, typically 50 to 100 meters above the ground. Backup section heights are set 10 meters above and 20 meters below the main section height, providing alternative solutions for emergency obstacle avoidance and altitude adjustment. For each grid cell, the maximum obstacle height at that location is calculated, including terrain height, building height, and other artificial obstacle heights. A safety buffer concept is introduced, setting a safety buffer distance around obstacles, typically two to three times the size of the UAV, to ensure flight safety. Grids exceeding the section height plus the safety buffer distance are marked as obstacle grids and assigned a value of one; the remaining grids are marked as passable grids and assigned a value of zero.

[0025] Output a multi-level obstacle distribution matrix (including binary raster data of each height level, obstacle height numerical matrix, and accessibility identification matrix) and a safety buffer zone identification map (including buffer zone boundary coordinates, safety distance parameters, and risk level distribution). The algorithm selection module establishes an environmental feature extraction model based on the obstacle distribution map and automatically selects the optimal path planning algorithm. Specifically, the following steps are included: Environmental feature extraction involves inputting a multi-level obstacle distribution matrix, an environmental complexity level map, historical flight data, and mission requirement parameters. An environmental feature extraction model is then established to automatically analyze key characteristic parameters of the current environment. Obstacle density is obtained by calculating the proportion of obstacle grids to the total number of grids in the cross-sectional image; obstacle density equals the number of obstacle grids divided by the total number of grids. Spatial connectivity is assessed by analyzing the number and size distribution of connected components in passable areas. A connected component labeling algorithm is used to identify all connected regions; the connectivity index is the ratio of the area of ​​the largest connected component to the total passable area. Path complexity is measured by calculating the ratio of the shortest straight-line distance from the starting point to the end point to the actual feasible path length; a ratio closer to one indicates a more direct path, while a smaller ratio indicates a more complex path. Dynamics are assessed by analyzing the frequency and magnitude of environmental changes in historical data; the frequency of change is the number of times the environment is updated per unit time, and the magnitude of change is the root mean square error of the difference before and after the environmental change. A feature vector is established, combining obstacle density, spatial connectivity, path complexity, and dynamics into a four-dimensional feature vector. in, Represents the environmental feature vector; This represents the normalized obstacle density; This represents the normalized spatial connectivity. This represents the normalized path complexity. This represents the dynamic characteristics after normalization. obstacle density The value ranges from 0 to 1 and does not require normalization; spatial connectivity Calculated using the ratio of the areas of the largest connected components, no normalization is required; path complexity... Calculating the path length ratio requires a logarithmic transformation. Then normalize to the [0, 1] interval. Indicates the actual path length; Represents the shortest straight-line distance; dynamic characteristics By changing frequency and amplitude The weighted average is calculated using the maximum value normalization method: ,in and The weights are for frequency and amplitude, respectively. Indicates the maximum frequency of change; Output environment feature vector (a four-dimensional numerical vector containing normalized values ​​of obstacle density, spatial connectivity, path complexity, and dynamics, with a data size of approximately 32 bytes), feature importance ranking (including weight coefficients and impact factor scores for each feature parameter), and environment classification label (processing time less than 10 milliseconds). Algorithm performance prediction involves inputting environmental feature vectors, historical algorithm performance data, and computing resource status; and constructing a machine learning-based algorithm performance prediction model. Historical performance data of different algorithms under various environmental conditions is collected, including metrics such as computation time, path length, path smoothness, and success rate, and a performance database is established to store these historical records. The performance prediction model is trained using a random forest algorithm, which consists of multiple decision trees. Each decision tree uses a different subset of training samples and features, with the environmental feature vector as input, to predict the performance of each algorithm. The prediction results are the expected value and confidence interval of each performance metric. An algorithm suitability scoring mechanism is established, and the suitability score calculation formula is: in, The algorithm's applicability is represented by a comprehensive score. Indicates the first Standardized scores for each performance metric; Indicates the first Weighting coefficients for each indicator; This indicates the total number of evaluation indicators.

[0026] Calculation time index The score is calculated by comparing the actual calculation time with the baseline time, and then normalized using a reciprocal transformation to ensure that shorter times result in higher scores; path length metric The success rate metric is calculated by comparing the actual path length to the shortest path length, and then normalized using the same reciprocal transformation. Since it's already in percentage form, just divide by 100 to normalize; path smoothness index The standard deviation of the path curvature change is calculated, and normalization is performed using the reciprocal transform. The weighting coefficients are set according to the application scenario, and are typically [value missing]. , , , ,satisfy Constraints are set. A performance threshold is defined, which is determined based on historical statistical data, typically set to 80% of the historical average performance. Only algorithms with scores exceeding the threshold are included in the candidate algorithm set.

[0027] Output algorithm performance prediction results (including predicted computation time, path length, and success rate for each algorithm), applicability score ranking (including comprehensive algorithm applicability score and confidence interval parameters), and candidate algorithm set (including algorithm name, parameter configuration, and performance threshold). Intelligent algorithm selection involves inputting a set of candidate algorithms, task priority settings, and real-time requirements. An intelligent algorithm selection strategy is designed to automatically select the optimal algorithm based on task characteristics and environmental conditions. For tasks prioritizing optimality, the A* algorithm is preferentially selected in simple environments, guaranteeing the finding of the shortest path. For tasks prioritizing path smoothness, the Theta algorithm is selected in moderately complex environments, reducing path inflection points through arbitrary angle connections. For tasks prioritizing fast response, an improved fast randomized exploration tree algorithm is selected in complex dynamic environments, rapidly generating feasible paths. An algorithm combination strategy is established; for particularly complex scenarios, multiple algorithms are run simultaneously, and the optimal result is selected, improving planning quality through parallel computation. Output results include: optimal algorithm selection results (including selected algorithm name, algorithm parameter configuration, and heuristic function settings), algorithm combination strategy (including multi-algorithm collaboration scheme and result fusion weights), and parallel computation allocation scheme (including computational resource allocation and task scheduling strategy).

[0028] Optionally, in some embodiments, a path planning algorithm based on deep reinforcement learning is used instead of the traditional A* and RRT algorithms. Deep reinforcement learning algorithms learn optimal strategies through interaction with the environment, enabling them to handle more complex dynamic environments and uncertainties. This algorithm establishes an interaction model between the agent and the environment. The agent selects actions based on the current state, and the environment provides reward feedback based on the actions. The agent learns the optimal path planning strategy by maximizing accumulated rewards. The state space includes information such as the current position of the drone, the target position, the distribution of surrounding obstacles, and the positions of other drones. The action space includes basic actions such as forward, backward, left turn, right turn, ascent, and descent. A deep neural network is used as a function approximator. The network structure includes an input layer, multiple hidden layers, and an output layer. The input layer receives state information, the hidden layers use a combination of convolutional and fully connected layers, and the output layer outputs the action probability distribution or value function. The reward function is designed to comprehensively consider multiple objectives such as path length, flight time, safety, and energy consumption. A positive reward is given for reaching the target point, a negative reward for colliding with obstacles, and negative rewards for path length and flight time. Reward shaping techniques guide the agent to learn the desired behavioral pattern.

[0029] The path planning module adopts a hierarchical planning strategy based on the selected path planning algorithm, generates a global path framework on a coarse-grained cross-section map, performs local path optimization within the flight corridor, generates a Pareto optimal solution using multi-objective optimization, and outputs the flight trajectory. Specifically, the following steps are included: Global path planning is performed by taking a coarse-grained cross-section as input, the optimal algorithm selection result, the origin and destination coordinates, and global constraints. Based on the coarse-grained cross-section, the system performs data format conversion, transforming the raster data of the cross-section into the graph structure representation required by the path planning algorithm. Specifically, each raster cell is treated as a graph node, the connectivity between adjacent raster cells is treated as graph edges, and obstacle raster cells are marked as impassable nodes. Then, the heuristic function parameters of the A* algorithm are dynamically configured in conjunction with the algorithm selection result. Global path planning is performed on the coarse-grained cross-section to obtain a rough flight corridor. A hierarchical planning strategy is adopted, ignoring detailed obstacles, to quickly generate a global path framework in a simplified environment. An improved A* algorithm is used for global search, with the heuristic function employing a weighted combination of Euclidean distance and Manhattan distance, with the weights adaptively adjusted according to environmental characteristics. The concept of a flight corridor is established, extending a certain width to both sides of the global path to form a flight corridor region, providing search space for subsequent local optimization. An adaptive adjustment mechanism for corridor width is set up, reducing the corridor width in areas with dense obstacles and increasing the corridor width in open areas. To ensure data consistency with subsequent steps, the system keeps the coordinate system of the global path framework consistent with the coordinate system of the section view, and adds timestamps and height information to each path node to facilitate subsequent local optimization processing.

[0030] Output a global path framework (including path node coordinate sequence, path segment connection relationship, and total path length parameter), a flight corridor area (including corridor boundary coordinates, centerline trajectory, and available space range), and a corridor width distribution map (including width values ​​of each path segment and constraint boundary coordinates). For local path optimization, the system takes a multi-level obstacle distribution matrix, flight corridor region, and detailed constraints as input. It performs data fusion processing, spatially registering the multi-level obstacle distribution matrix with the flight corridor region to ensure consistency in coordinate systems and resolutions between the two data sources. Bilinear interpolation is used to unify data from different resolutions to the same raster scale. Then, local path optimization is performed within the flight corridor region to generate a detailed flight trajectory. A high-precision raster map is used to identify small obstacles and detailed features. The Theta algorithm is used for local optimization, allowing path connections at arbitrary angles and generating smoother trajectories. During optimization, the system continuously monitors the deviation of path points from the global path framework. When the deviation exceeds a preset threshold, a path correction mechanism is automatically triggered to ensure consistency between the local optimization results and the global planning. A path smoothing mechanism is established, using cubic spline interpolation to smooth the path, eliminating sharp inflection points and discontinuities. Dynamic constraints are set to ensure that the generated path meets the UAV's flight performance limits, including maximum turning radius, maximum rate of climb, and maximum rate of descent. To ensure the continuity of the processing, the system verifies the feasibility and safety of the path in each optimization iteration and feeds the verification results back to the next round of optimization.

[0031] Output the optimized detailed path (including 3D coordinate point sequence, path segment timestamps, velocity and acceleration parameters), path smoothness evaluation (including curvature change statistics, smoothness score, inflection point analysis results), and dynamic constraint verification report (including turning radius verification, climb rate verification, and speed limit verification). Multi-objective optimization involves inputting the optimized detailed path, path smoothness assessment, optimization objective weights, and performance evaluation criteria. A multi-objective optimization evaluation system is established, comprehensively considering multiple objectives such as path length, flight time, energy consumption, and safety. The path length objective is evaluated by calculating the total path length; the flight time objective considers the UAV's flight speed and acceleration / deceleration processes; the energy consumption objective is calculated based on the UAV's power consumption model; and the safety objective is evaluated by analyzing the minimum distance between the path and obstacles and the crossing of risky areas. A non-dominated sorting genetic algorithm is used to generate a Pareto optimal solution set, providing users with multiple optimization schemes. An adaptive adjustment mechanism for objective weights is established, dynamically adjusting the weights of each objective based on task type and user preferences; for example, emergency tasks prioritize flight time, while cargo transportation tasks prioritize energy consumption.

[0032] Output the optimal solution set (including multiple optimized path schemes, performance index values ​​of each scheme, and scheme selection suggestions), target weight configuration (including weight coefficients of each optimization target and weight adjustment strategies), and performance comparison analysis report (including path length comparison, flight time comparison, energy consumption comparison, and safety comparison).

[0033] The collaborative scheduling module receives the flight trajectory and establishes a distributed collaborative architecture. It selects a leader node based on the node priority calculation formula, identifies potential conflicts using a predictive conflict detection algorithm, and executes avoidance actions through a collaborative scheduling solution strategy. Specifically, the following steps are included: A distributed collaborative architecture is established. Inputs include the optimal path set, UAV network topology, communication protocol parameters, and scheduling strategy configuration. The system performs path compatibility analysis on the optimal path set, checks the possibility of spatiotemporal conflicts between UAV paths, and establishes a path dependency graph. Based on the path analysis results, a distributed collaborative scheduling architecture is established, with each UAV acting as an intelligent node, possessing local decision-making capabilities while simultaneously accepting coordination and management from the central scheduling system. The system maintains a global state information table, recording the current position, flight status, mission progress, and communication status of each UAV in real time, ensuring that collaborative decisions are based on the latest system state. A hierarchical communication protocol is designed, including vertical communication between UAVs and the central system and horizontal communication between UAVs. A dynamic node role allocation mechanism is established, dynamically allocating leader and follower node roles based on the UAV's position, mission priority, and communication capabilities. The node priority calculation formula is: in, Indicates the first The priority of each drone node; Indicates the first The remaining battery power of the drone; This indicates the total battery power of the drone; Indicates the first The distance from the drone to the center of the area; Indicates the first Communication quality score for each drone; , , These represent the power weight, location weight, and communication weight, respectively.

[0034] Power ratio Normalized to the [0,1] interval; distance reciprocal Standardization is required; calculate the maximum distance from all drones to the center of the area. Then, the reciprocal of the distance is standardized to ,in Minimum distance; Communication quality score Usually expressed as a percentage, it needs to be normalized by dividing by 100 to convert it to a value in the range [0,1]. Weighting coefficients , , satisfy The constraints.

[0035] The system incorporates communication redundancy and fault tolerance mechanisms, enabling it to automatically rebuild the communication network and ensure the continuity of coordinated scheduling when some communication links fail.

[0036] Output a distributed network topology diagram (including node connection matrix, network hierarchy, and topology stability parameters), communication protocol configuration (including communication frequency settings, data packet format, and transmission priority), and role allocation scheme (including master node identifier, slave node list, and role switching rules). Predictive collision detection takes as input a distributed network topology graph, communication protocol configuration, multi-drone flight trajectories, real-time position information, velocity vectors, and a prediction time window. Based on the established network topology graph, the system initializes a distributed collision detection network, with each drone node responsible for monitoring collisions within its neighborhood. The system maintains a dynamic trajectory cache, storing historical and predicted trajectory data for each drone, and periodically updates it according to the communication protocol configuration. A prediction-based collision detection algorithm is employed, detecting not only current collisions but also predicting potential collisions within future time windows. A four-dimensional collision detection model is established, adding a time dimension to the three-dimensional spatial coordinates to form a four-dimensional spatiotemporal trajectory. Within each detection cycle, the system first updates the state information of all drones and then performs trajectory prediction based on the latest state. The trajectory prediction uses a kinematic model. in, Represents the position vector at the predicted time; Represents the position vector at the current moment; Represents the velocity vector at the current moment; This represents the acceleration vector at the current moment; Indicates the prediction time interval.

[0037] For each UAV, the future trajectory is predicted based on its current position, velocity, and acceleration, taking into account flight dynamics constraints and control delays. A cylindrical protection zone model is used, establishing a cylindrical protection zone centered on the UAV. A conflict is defined when two protection zones overlap. A conflict severity assessment mechanism is established. in, A score indicating the severity of the conflict; Indicates the expected time of the conflict; Indicates the minimum separation distance; Indicates the magnitude of relative velocity; , , These represent time weight, distance weight, and speed weight, respectively.

[0038] Output conflict detection results (including conflict type classification, conflict location coordinates, and conflict time prediction), conflict severity rating (including severity score, risk level classification, and processing priority), and predicted trajectory data (including future trajectory coordinate sequence, velocity change prediction, and acceleration change prediction). The system employs a negotiated conflict resolution mechanism. Inputting conflict detection results, conflict severity ratings, drone priorities, and available avoidance strategies, the system immediately initiates the conflict resolution process upon receiving the detection results. It categorizes and prioritizes conflict events, ensuring high-risk conflicts are handled first. The system maintains a conflict resolution state machine to track the processing progress and resolution status of each conflict event, preventing duplicate or missed processing. A dynamic priority adjustment mechanism is established, dynamically calculating drone priorities based on factors such as mission urgency, flight progress, remaining battery power, and payload importance. A negotiated conflict resolution strategy is used; when a conflict is detected, relevant drones negotiate to determine the optimal avoidance plan. The system maintains a negotiation history during the negotiation process to avoid circular negotiations and deadlocks. Multiple avoidance strategies are designed, including altitude adjustment, speed adjustment, path deviation, and waiting strategies. The altitude adjustment strategy avoids the conflict by changing flight altitude; the speed adjustment strategy delays the conflict time by accelerating or decelerating; the path deviation strategy temporarily deviates from the original path to bypass the conflict area; and the waiting strategy hovers and waits for other drones to pass. An avoidance cost evaluation model is established, comprehensively considering factors such as time delay, increased energy consumption, and path deviation to select the optimal avoidance strategy. The system continuously monitors the avoidance effect during strategy execution and adjusts subsequent strategies based on actual performance. The avoidance cost calculation formula is as follows: in, Indicates the cost of avoidance; Indicates a time delay; This indicates an increase in energy consumption; Indicates the distance deviated from the path; , , These represent the weighting coefficients for time, energy consumption, and distance, respectively.

[0039] Output conflict solutions (including path adjustment strategies, time scheduling corrections, and priority reallocation), avoidance strategy selection (including avoidance action types, execution time arrangements, and coordination mechanisms), and cost assessment reports (including time cost analysis, energy cost analysis, and path deviation cost analysis).

[0040] Optionally, in some embodiments, a blockchain-based distributed collaborative scheduling mechanism replaces the traditional centralized scheduling method. Blockchain technology provides a decentralized trust mechanism, where each drone acts as a node in the blockchain network, reaching a consensus on scheduling decisions through a consensus algorithm. Smart contracts define collaborative rules and conflict resolution strategies, ensuring all participants adhere to a unified protocol. Consensus mechanisms such as proof-of-work or proof-of-stake are used to verify the validity of scheduling decisions and prevent interference from malicious nodes. A token incentive mechanism is designed to encourage drone nodes to actively participate in collaborative scheduling and information sharing, improving the overall network's collaborative efficiency.

[0041] The visualization module uses adaptive rendering technology based on the collaborative scheduling results, and introduces augmented reality technology to overlay virtual trajectories; Specifically, the following steps are included: Adaptive rendering optimization takes into account inputs such as the collaborative scheduling scheme (including flight plans for each drone, collaborative action sequences, and synchronization time points), conflict resolution methods (including path adjustment strategies, time scheduling corrections, and priority reallocation), avoidance strategy selection (including avoidance action types, execution time arrangements, and coordination mechanisms), device performance parameters, network bandwidth status, and user preference settings. The system integrates and processes multiple scheduling data points to establish a unified visual data model, converting the collaborative scheduling scheme, conflict resolution methods, and avoidance strategy selections into scene objects and animation sequences required for visual rendering. The system maintains a rendering state cache, tracking the current state and change trends of each visual object in real time to ensure that the rendered content remains synchronized with the actual flight status. Adaptive rendering technology is used to dynamically adjust rendering quality based on device performance and network conditions, ensuring a smooth user experience. A device performance evaluation model is established to assess device level by detecting parameters such as processor performance, graphics processing capabilities, and memory capacity. During rendering, the system continuously monitors frame rate and system load; when performance indicators fall below a threshold, it automatically reduces rendering quality or simplifies scene complexity. The comprehensive device performance score is calculated using the following formula: in, This indicates the overall performance score of the equipment; , , These represent standardized scores for the processor, graphics processor, and memory, respectively. , , These represent the weighting coefficients for the processor, graphics processor, and memory, respectively.

[0042] Processor performance rating The graphics processing unit (GPU) performance score is calculated by comparing the processor frequency to a reference frequency (typically 2.0 GHz) and then normalized using the sigmoid function to ensure the score falls within the range of [0, 1]. Memory capacity score is calculated by comparing the GPU benchmark score to a reference score, using the same sigmoid normalization method; Calculated by the ratio of memory size to a reference capacity (typically 4GB), and normalized using the sigmoid function. The weighting coefficients are set according to the characteristics of the rendering task, and are typically [value missing]. , , ,satisfy The constraints.

[0043] Design a multi-level rendering quality scheme, including high, medium, and low quality levels. Establish a dynamic quality adjustment mechanism to automatically adjust the rendering quality based on the real-time frame rate and system load, ensuring smooth system response.

[0044] Output adaptive rendering configuration (including rendering quality parameters, frame rate target settings, and resource allocation strategies), quality level allocation (including device performance level, corresponding rendering quality, and dynamic adjustment threshold), and performance monitoring report (including real-time frame rate statistics, system load analysis, and quality adjustment records). A multi-layered visualization system is implemented, accepting inputs such as adaptive rendering configurations, quality level assignments, 3D scene data, flight trajectory information, and user perspective parameters. This system supports seamless switching between macro-level city views and micro-level single-drone trajectory views. The city-level view displays the distribution of drones and overall traffic flow across the entire city area, using heatmaps and flow maps to show airspace usage density. The region-level view focuses on the details of a specific area, displaying the real-time location and flight trajectory of all drones within that area. The single-drone view tracks the flight process of a specific drone, displaying detailed flight parameters and status information. View transition animations are implemented, achieving a natural transition between different view levels through smooth zooming and panning animations. An adaptive information density adjustment mechanism is set up, hiding detailed information in the macro view to avoid visual clutter, and displaying detailed information in the micro view to meet the needs of precise analysis.

[0045] Output multi-level view configuration (including view hierarchy definition, display content specifications, and interactive operation mapping), switching animation scheme (including animation duration settings, transition effect parameters, and view transformation matrix), and information density distribution map (including information display rules for each level, density control parameters, and adaptive adjustment strategy). Augmented reality overlay is implemented, inputting multi-layered view configurations, switching animation schemes, information density distribution maps, real-world images, virtual flight trajectories, and camera calibration parameters. Augmented reality technology is introduced to overlay virtual flight trajectories and related information onto the real environment, providing a more intuitive spatial perception experience. A camera calibration and tracking system is established, using feature point detection and matching algorithms to determine the camera's position and attitude in real time. A combination of label-based tracking and label-free natural feature tracking methods is employed to improve tracking stability and accuracy. A virtual information overlay strategy is designed to overlay the flight trajectory as a 3D curve onto the real scene, using different colors to distinguish different flight stages. Key information labels are overlaid, including drone number, flight altitude, speed, and remaining battery power. An occlusion handling mechanism is established to correctly handle occlusion relationships when virtual objects are occluded by real objects, enhancing realism.

[0046] Output augmented reality overlay effect (including virtual object rendering parameters, overlay position coordinates, and visual effect configuration), tracking accuracy report (including position tracking error, pose tracking accuracy, and stability assessment), and occlusion handling solution (including depth detection algorithm, occlusion relationship judgment, and rendering order adjustment).

[0047] The simulation verification module receives visualized data to build a digital twin environment and introduces artificial intelligence for intelligent verification and system optimization. Specifically, the following steps are included: High-fidelity environment modeling involves inputting augmented reality overlay effects (including virtual object rendering parameters, overlay position coordinates, and visual effect configurations), tracking accuracy reports (including position tracking error, attitude tracking accuracy, and stability assessment), occlusion handling solutions (including depth detection algorithms, occlusion relationship judgment, and rendering order adjustment), real-world environmental data, physical parameters, and dynamic element information. Based on the visualization output, the system extracts key environmental parameters and object attributes as the foundational data for digital twin environment modeling. The system establishes an environmental data synchronization mechanism to ensure consistency between the digital twin environment and the visualization system's environmental representation, avoiding simulation deviations caused by data inconsistencies. A high-fidelity digital twin environment is constructed, accurately reproducing the physical characteristics and dynamic changes of the real world. A detailed 3D geographic model is established, including static elements such as terrain, buildings, vegetation, and roads, with model accuracy reaching the decimeter level. The system maintains model version control during the modeling process, tracks environmental change history, and supports retrospective and comparative analysis of environmental states. Dynamic environmental elements are integrated, including real-time weather data, wind field information, other aircraft dynamics, and ground traffic flow. A physical simulation engine is established to simulate the effects of various physical forces such as gravity, air resistance, and wind. The system continuously verifies the accuracy of the physical model during the simulation process and adjusts the model parameters based on actual observation data. The force analysis of the UAV adopts Newton's second law.

[0048] Computational fluid dynamics (CFD) methods are used to simulate complex wind field distributions, considering the impact of buildings on the wind field. Sensor simulation models are established to simulate the operating characteristics and error features of sensors such as GPS, IMU, cameras, and lidar. An environmental change simulation mechanism is implemented to simulate various scenarios including weather changes, emergencies, and equipment failures.

[0049] Output a high-fidelity digital twin environment (including 3D scene model, physical property parameters, dynamic element configuration), physical simulation parameters (including mechanical model parameters, environmental physical constants, simulation accuracy settings), and sensor model configuration (including sensor characteristic parameters, error model settings, and data output format). The system employs a multi-layered simulation verification approach. Inputting high-fidelity digital twin environment physical simulation parameters, sensor model configurations, flight plans, and UAV model parameters, the system initializes the simulation verification framework based on the established digital twin environment, ensuring complete compatibility between the simulation environment and the data models from previous steps. A simulation state management mechanism is established to maintain simulation progress, intermediate results, and abnormal states, supporting pause, resumption, and backtracking operations during the simulation process. A multi-layered simulation verification system is established, encompassing comprehensive verification from single-machine simulation to multi-machine collaborative simulation, and from normal to extreme conditions. Single-machine simulation verifies the path tracking performance, obstacle avoidance capabilities, and flight stability of a single UAV. Multi-machine simulation verifies the collaborative flight, conflict avoidance, and communication coordination capabilities of multiple UAVs. Data transfer interfaces are established between each simulation layer to ensure that simulation results from higher layers are correctly transmitted to lower layers. Normal condition simulation verifies system performance under standard weather and environmental conditions. Extreme condition simulation includes system performance under adverse conditions such as strong winds, rainfall, fog, and communication interference. A simulation experiment design methodology is established, systematically testing the effects of various parameter combinations and random factors through orthogonal experimental design and Monte Carlo methods. The system continuously collects performance data during simulation and analyzes the validity and reliability of the simulation results in real time. A performance evaluation index system is established, including path tracking accuracy, task completion rate, safety indicators, and efficiency indicators.

[0050] Output multi-level simulation results (including single-machine simulation data, multi-machine collaborative simulation data, and performance indicators at each level), performance evaluation reports (including path tracking accuracy, task completion rate, safety indicators, and efficiency indicators), and extreme condition test reports (including severe weather test results, communication interference test results, and equipment failure test results). Intelligent verification and optimization involves inputting multi-level simulation results, performance evaluation reports, extreme condition test reports, real flight data, and performance benchmarks. Upon receiving simulation verification results, the system performs data quality checks and integrity verification to ensure the validity and consistency of all simulation data. An intelligent analysis pipeline is established to perform correlation analysis on simulation results, performance reports, and test reports, identifying inherent connections and potential contradictions between data. Artificial intelligence technology is introduced for intelligent verification and system optimization. A comparative analysis model between simulation data and real flight data is established, using statistical analysis and machine learning methods to identify deviations and deficiencies in the simulation model. The system maintains the training state and knowledge base of the learning model, supporting incremental learning and model updates. Deep learning algorithms are used to automatically identify abnormal behaviors and potential risks during the simulation process, including trajectory anomalies, performance degradation, and safety hazards. An automatic optimization mechanism is established to automatically adjust algorithm parameters, optimize path planning strategies, and improve collaborative scheduling schemes based on verification results. During optimization, the system maintains historical optimization records, tracks optimization effects and parameter change trends, and avoids oscillations and divergences during optimization. A credibility assessment system is designed, comprehensively considering factors such as simulation accuracy, verification coverage, and consistency checks to provide a credibility score for the simulation results. The system establishes a feedback mechanism to return verification and optimization results to preceding steps, forming a closed-loop optimization process. The credibility assessment formula is as follows: in, This indicates the simulation credibility score; Indicates the simulation accuracy score; This indicates the verification coverage score; Indicates the consistency verification score; , , These represent the weighting coefficients corresponding to simulation credibility, simulation accuracy, and verification coverage, respectively.

[0051] Simulation accuracy rating The root mean square error is calculated by comparing the simulation results with the actual data, using the formula... Normalization is performed, where The maximum acceptable error; verify coverage score. The consistency verification score is calculated directly using the ratio of the number of scenarios covered by test cases to the total number of scenarios. Consistency analysis of multiple simulation results was performed, and the reciprocal of the coefficient of variation was used for scoring. ,in This is the coefficient of variation. The weighting coefficients are set according to the validation requirements, and are typically [value missing]. , , ,satisfy The constraints are defined. A continuous learning mechanism is established, enabling the system to learn from each simulation and actual flight, and continuously improve and optimize its performance.

[0052] Output intelligent verification report (including abnormal behavior identification results, risk assessment analysis, and verification conclusion summary), optimization suggestions (including parameter adjustment suggestions, strategy improvement schemes, and performance enhancement measures), credibility assessment results (including simulation accuracy score, verification coverage score, and consistency verification score), and continuous learning model (including learning algorithm configuration, knowledge update mechanism, and experience accumulation strategy).

[0053] In one embodiment of the present invention, an application example of an airspace network flight simulation system is provided: A 30-day field test was conducted in a science and technology park area of ​​a certain city. The test area covered approximately 25 square kilometers and included high-rise buildings ranging from 50 to 300 meters in height, with a building density of 75%. During the test, 15 multi-rotor drones were deployed to perform logistics tasks such as express delivery, food delivery, and medical supply delivery. The environmental parameters and system configuration information of the test area are shown in Table 1. Table 1: Environmental Parameters and System Configuration of the Test Area The implementation results of each step are statistically analyzed in Table 2: Table 2: Statistics on the Implementation Results of Each Step like Figure 2 As shown, this demonstrates the advantages of multi-source data fusion technology. The fused data outperforms single data sources in all dimensions, particularly in terms of accuracy and reliability. Satellite remote sensing data has advantages in coverage, UAV aerial photography data excels in accuracy, and ground sensor data leads in update frequency; the fused data combines the strengths of each data source, achieving a comprehensive improvement.

[0054] like Figure 3 As shown, the advantages of the algorithm of this invention compared to traditional algorithms are demonstrated. In environments of varying complexity, the height-section graph technique of this invention can significantly shorten path planning time. In extremely complex environments, the traditional A* algorithm requires five minutes to complete path planning, while the algorithm of this invention only takes thirty-five seconds, improving efficiency by more than eight times, fully verifying the advanced nature of the technical solution.

[0055] like Figure 4As shown, the system of this invention demonstrates its comprehensive performance in multi-machine collaborative scheduling. The system outperforms traditional systems in six dimensions: conflict detection rate, path optimization, communication efficiency, scheduling success rate, system stability, and resource utilization. In particular, it achieves a high conflict detection rate of 99.8% and a scheduling success rate of 98.5%, fully validating the effectiveness of the distributed collaborative architecture and predictive conflict detection algorithm.

[0056] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A space network flight simulation system, characterized in that, include: The multi-source data fusion module uses a time-weighted fusion algorithm to intelligently fuse multi-source data and output a geographic feature database. The height section map generation module receives a geographic feature database, adaptively adjusts the raster resolution according to environmental complexity, and generates a multi-level obstacle distribution map. The algorithm selection module establishes an environmental feature extraction model based on the obstacle distribution map and automatically selects the optimal path planning algorithm. The path planning module adopts a hierarchical planning strategy based on the selected path planning algorithm, generates a global path framework on a coarse-grained cross-section map, performs local path optimization within the flight corridor, generates a Pareto optimal solution using multi-objective optimization, and outputs the flight trajectory. The collaborative scheduling module receives the flight trajectory and establishes a distributed collaborative architecture. It selects a leader node based on the node priority calculation formula, identifies potential conflicts using a predictive conflict detection algorithm, and executes avoidance actions through a collaborative scheduling solution strategy. The visualization module uses adaptive rendering technology based on the results of collaborative scheduling, and introduces augmented reality technology to overlay virtual trajectories. The simulation verification module receives visualized data to build a digital twin environment and introduces artificial intelligence for intelligent verification and system optimization.

2. The airspace network flight simulation system according to claim 1, characterized in that, The multi-source data fusion module includes: The data preprocessing layer standardizes the format of data from different sources, converts all spatial data into the WGS84 geographic coordinate system, and converts elevation data into an altitude benchmark. The time-weighted fusion layer uses a time-weighted fusion algorithm to intelligently fuse multi-source data. The algorithm assigns time weights based on the timeliness of the data. The time weights are calculated using an exponential decay function, and the weight values ​​decrease exponentially as the time difference increases. The time decay constant is determined based on the data type. A smaller decay constant is set for rapidly changing dynamic data, and a larger decay constant is set for relatively stable static data. The data validation layer establishes a data consistency verification mechanism, identifies inconsistent data sources through cross-validation algorithms, and employs a majority voting mechanism for conflicting data.

3. The airspace network flight simulation system according to claim 1, characterized in that, The height section map generation module includes: The environmental complexity assessment unit establishes an environmental complexity assessment model, quantifies environmental complexity by analyzing indicators such as building density, height variation gradient, and obstacle distribution uniformity, and classifies the environment into three levels: simple, medium, and complex based on the assessment results. The adaptive raster resolution determination unit adjusts the raster resolution according to the complexity of the environment. It uses high-resolution raster in complex areas with dense buildings, low-resolution raster in simple areas with sparse buildings, and medium-resolution raster in medium-complexity areas. It establishes resolution transition areas to avoid path discontinuity caused by sudden resolution changes. The multi-level section generation unit uses multi-level section technology to generate obstacle distribution maps at different height levels and introduces the concept of a safety buffer zone, setting a safety buffer distance around the obstacles.

4. The airspace network flight simulation system according to claim 1, characterized in that, The algorithm selection module includes: The environmental feature extraction unit establishes an environmental feature extraction model, automatically analyzes the key feature parameters of the current environment, and combines obstacle density, spatial connectivity, path complexity, and dynamic feature parameters into a four-dimensional feature vector. The algorithm performance prediction unit constructs an algorithm performance prediction model based on machine learning, trains the performance prediction model using the random forest algorithm, uses environmental feature vectors as input to predict the performance of each algorithm, and establishes an algorithm applicability scoring mechanism that comprehensively considers prediction performance, computing resource status and task priority to calculate the applicability score. The intelligent algorithm selection unit designs an intelligent algorithm selection strategy, automatically selects the optimal algorithm based on task characteristics and environmental conditions, establishes an algorithm combination strategy, and runs multiple algorithms simultaneously for particularly complex scenarios and selects the optimal result.

5. The airspace network flight simulation system according to claim 1, characterized in that, The path planning module includes: The global path planning unit adopts a hierarchical planning strategy to obtain flight corridors by performing global path planning on a coarse-grained cross-sectional map. An improved A-Star algorithm is used for global search. The heuristic function adopts a weighted combination of Euclidean distance and Manhattan distance to establish the concept of flight corridors and expand to both sides on the basis of global path to form a search space. The local path optimization unit performs local path optimization on the fine-grained cross-sectional map to generate a detailed flight trajectory. The Theta algorithm is used for local optimization to generate a smoother trajectory. A path smoothing mechanism is established, and the path is smoothed using a cubic spline interpolation algorithm. The multi-objective optimization unit establishes a multi-objective optimization evaluation system that comprehensively considers multiple objectives such as path length, flight time, energy consumption, and safety, and uses a non-dominated sorting genetic algorithm to generate a Pareto optimal solution set.

6. The airspace network flight simulation system according to claim 1, characterized in that, The collaborative scheduling module includes: The distributed collaborative architecture establishment unit establishes a distributed collaborative scheduling architecture. Each UAV acts as an intelligent node with local decision-making capabilities, while also accepting coordination and management from the central scheduling system. A hierarchical communication protocol is designed, and a dynamic allocation mechanism for node roles is established. The predictive conflict detection unit establishes a four-dimensional conflict detection model, adding a time dimension to the three-dimensional spatial coordinates. The trajectory prediction adopts a kinematic model, predicting the future position based on the current position, velocity, and acceleration, and using a cylindrical protection zone model to determine conflicts. The negotiation-based conflict resolution unit adopts a negotiation-based conflict resolution strategy. When a conflict is detected, the relevant UAVs negotiate through communication to determine the optimal avoidance plan. Multiple avoidance strategies are designed, including altitude adjustment, speed adjustment, path deviation, and waiting strategies.

7. The airspace network flight simulation system according to claim 1, characterized in that, The visualization module includes: The adaptive rendering optimization unit uses adaptive rendering technology to dynamically adjust the rendering quality based on device performance and network conditions. It establishes a device performance evaluation model and evaluates the device level by detecting the device's processor performance, graphics processing capability, and memory capacity parameters. It designs three rendering levels: high quality, medium quality, and low quality. A multi-level visualization system unit is established to support seamless switching from macro-level city-level views to micro-level single-machine trajectory views. The view hierarchy structure is designed to achieve smooth switching animations between different levels and dynamically adjust information density according to the view hierarchy. The augmented reality overlay unit introduces augmented reality technology to overlay virtual flight trajectories onto the real environment. It adopts feature point-based tracking technology to achieve virtual-real registration and designs an intelligent overlay strategy to adjust the display effect of virtual objects according to ambient lighting and background complexity.

8. The airspace network flight simulation system according to claim 1, characterized in that, The simulation verification module includes: The high-fidelity environment modeling unit constructs a high-fidelity digital twin environment, establishes a detailed three-dimensional geographic model, integrates real-time weather data, wind field information, and dynamic environmental elements, and uses a physics engine to simulate real aerodynamic effects. A multi-level simulation verification unit is established to create a multi-level simulation verification system, including single-machine simulation verification, multi-machine collaborative simulation verification, and system-level simulation verification. A simulation state management mechanism is established, and a hierarchical experimental design method is adopted. The intelligent verification and optimization unit introduces artificial intelligence technology for intelligent verification and system optimization, establishes an intelligent analysis pipeline, uses machine learning algorithms to identify abnormal patterns in simulation, and establishes a comparative analysis model between simulation data and real flight data.

9. The airspace network flight simulation system according to claim 8, characterized in that, The intelligent verification and optimization unit also includes: The abnormal behavior recognition subunit uses deep learning algorithms to automatically identify abnormal behaviors and potential risks during the simulation process, including trajectory anomalies, performance degradation, and safety hazards. The automatic optimization subunit automatically adjusts algorithm parameters, optimizes path planning strategies, and improves collaborative scheduling schemes based on verification results, while maintaining historical optimization records to track optimization effects. The credibility assessment subunit comprehensively considers simulation accuracy, verification coverage, and consistency verification factors to provide a credibility score for the simulation results. It also establishes a feedback mechanism to feed back the verification and optimization results to the preceding steps, forming a closed-loop optimization process.

10. A computer-readable storage medium, characterized in that, It is used to store computer-readable instructions that, when read by a computer, enable the operation of an airspace network flight simulation system as described in any one of claims 1-9.

Images