A low-altitude economy operation situation three-dimensional visualization and data mapping method based on digital twinning

Abstract

The application discloses a low-altitude economy operation situation three-dimensional visualization and data mapping method based on digital twinning, relates to the technical field of digital twinning and low-altitude airspace intelligent management, and comprises the following steps: collecting multi-dimensional operation data of a low-altitude aircraft to construct a digital twinning body, and deducing a future situation based on a multi-agent simulation engine combined with rules and reinforcement learning; mapping the deduced result to a visual parameter through a confidence attenuation function to realize uncertain coding expression; combining eye movement data and interactive behavior to calculate a cognitive load index and highlight the situation; constructing a stress state index through a physiological signal and triggering authority control and scene load reduction; and realizing dispatcher state visualization and task handover in a multi-seat collaborative scene. The application can improve the low-altitude situation prediction capability and visual expression effect, realize man-machine collaborative optimization and safe scheduling.

Description

Technical Field

[0001] This invention relates to the field of digital twin and low-altitude airspace intelligent management technology, and in particular to a three-dimensional visualization and data mapping method for low-altitude economic operation status based on digital twin. Background Technology

[0002] With the rapid development of the low-altitude economy, application scenarios such as unmanned aerial vehicle (UAV) logistics, urban air traffic (UAM), emergency inspection, and refined airspace management are constantly emerging. Low-altitude airspace is gradually evolving from traditional low-density, low-frequency flight activities to a complex operational mode characterized by high density, high dynamics, and multiple concurrent tasks. Against this backdrop, 3D visualization technology based on digital twins is gradually becoming an important technical means for low-altitude operational situational awareness and decision support. Existing technologies typically construct a mapping relationship between physical airspace and virtual space, combined with multi-source sensor data, to achieve 3D reconstruction and visualization of aircraft position, trajectory, and operational status. However, existing methods primarily focus on real-time situational awareness, and still have significant shortcomings in future situational projection capabilities, multi-agent collaborative decision-making modeling, and human-machine interaction and cognitive adaptation mechanisms. On the one hand, traditional digital twin systems lack the ability to dynamically model the complex interactive behaviors of multiple low-altitude aircraft, making it difficult to support high-precision predictions and uncertainty representation of future situations. On the other hand, existing 3D visualization results are usually presented statically or in a single dimension, failing to visually encode key information such as prediction confidence and risk level, making it difficult for dispatchers to quickly identify potential risks. Furthermore, existing systems generally ignore changes in the cognitive and physiological states of dispatchers under high-load scenarios, lacking human factors engineering-based adaptive interface adjustments and risk intervention mechanisms, which can easily lead to cognitive overload, decision-making delays, and even operational errors in high-density operation scenarios.

[0003] CN121460095A discloses a method and device for visualized management of emergency operations based on digital twins. By constructing a 3D twin scene corresponding to the physical emergency room space, and combining multi-source data acquisition and a data-driven model, it achieves patient flow heatmap identification, resource utilization analysis, and predictive scheduling, thereby improving the visualization and decision-making efficiency of emergency operations. While this method achieves dynamic visualized management driven by digital twins in medical scenarios, its core focus is on resource scheduling and traffic prediction within a single scene. It does not involve collaborative deduction mechanisms for complex behaviors among multiple agents, and particularly lacks quantitative expression and visual encoding of future situation uncertainties. Furthermore, this method does not consider the cognitive load changes of dispatchers in high-information-density environments, nor does it introduce dynamic interface adjustment strategies based on human-computer interaction feedback, making it difficult to apply to complex scenarios in low-altitude economies where multiple aircraft operate concurrently and risks evolve rapidly.

[0004] CN111833426B discloses a 3D visualization method based on digital twins. This method collects sensor data, performs classification, statistical analysis, and predictive analysis, generates virtual digital modules using a data registration pool mechanism, and then renders them uniformly in a 3D scene to construct a virtual IoT digital twin system. This method has certain advantages in data organization and 3D visualization, enabling the virtual mapping and display of the operating status of IoT devices. However, it primarily focuses on data-driven static or near-real-time displays, lacking the ability to simulate and extrapolate multi-agent scenarios over future time series, and failing to construct decision-making models based on reinforcement learning or rule fusion to support the prediction of complex situational evolution. Furthermore, its 3D visualization expression does not involve key mechanisms such as confidence decay and uncertainty mapping, making it difficult to provide clear risk levels and distinct information layers for scheduling decisions. In addition, this method does not involve the perception and feedback control of the operator's cognitive and physiological states, lacking human-driven system adaptive adjustment capabilities. Summary of the Invention

[0005] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this section, the abstract and title of the invention. Such simplifications or omissions shall not be used to limit the scope of the present invention.

[0006] In view of the problems of insufficient future situation prediction capability, lack of multi-agent collaborative modeling and imperfect human-machine cognitive adaptation mechanism in existing low-altitude operational situation visualization technology, this invention is proposed.

[0007] Therefore, the problem to be solved by this invention is how to achieve high-precision prediction and simulation of the operational status of multiple low-altitude aircraft, visualization of uncertainty information, and adaptive interactive optimization based on the cognitive state of dispatchers.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a method for three-dimensional visualization and data mapping of low-altitude economic operation status based on digital twins, which includes: Real-time flight status data of all aircraft in operation in the low-altitude airspace are collected, and an initial state set of digital twins of each aircraft is constructed. The low-altitude operational status is then simulated in real time through a multi-agent simulation engine that uses a hybrid decision-making model of rules and reinforcement learning, generating a simulated status sequence. Based on the inferred situation sequence, the pre-constructed confidence decay function is mapped to the rendering transparency and boundary blur parameters of the corresponding twin, forming an inferred visualization frame stream carrying uncertainty coding, and is rendered and output in a split-screen overlay manner with the current real-time three-dimensional situation scene. The system collects the visual dwell time and saccade frequency of the dispatcher, calculates the real-time cognitive load index of the dispatcher by combining the hot zone distribution of interactive behavior, triggers the key situation highlighting algorithm, aggregates low-risk aircraft twins into cluster icons, and pushes high-risk aircraft twins to the center of the field of vision with magnification and highlighted boundaries. Collect dispatchers' heart rate variability and skin conductance values ​​to construct a stress state index for dispatchers. Lock high-risk operation permissions through a twin system, switch the 3D scene to low-density rendering mode, and simultaneously trigger multi-channel alarm push. The cognitive load index and the stress state index are used as dynamic attributes of the commander's twin. The status of each dispatcher is visualized in a multi-seat collaborative 3D scenario using color-coded annotations. Task handover suggestions are generated and mapped to the task allocation data layer of the 3D situation scenario.

[0009] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention collects real-time situational data of low-altitude airspace vehicles across the entire domain and constructs an initial state set of a digital twin. Combined with a rule-based and reinforcement learning hybrid decision-making model, it drives a multi-agent simulation engine for ultra-real-time inference, achieving a forward-looking prediction of future operational situations and overcoming the limitation of traditional systems that can only present the current state. By mapping the confidence decay function to the rendering transparency and boundary blurring parameters of the twin, the uncertainty of the inference is encoded visually and presented synchronously with the real-time scene in a split-screen overlay format. This allows dispatchers to intuitively perceive the credibility of the predicted information while obtaining it, effectively avoiding the risk of misjudgment due to over-reliance on the inference results. Furthermore, by collecting eye-tracking features and interaction hotspot distribution in real time to calculate the cognitive load index, it adaptively triggers a key situation highlighting algorithm, thereby enhancing the predictability of low-altitude airspace operations. The system aggregates and reduces noise from high-risk aircraft and prioritizes pushing high-risk aircraft to the center of the field of view, achieving a dynamic match between information density and the dispatcher's cognitive capacity. This fundamentally alleviates the information overload problem in high-density, low-altitude scenarios. Simultaneously, by collecting heart rate variability and skin conductance values ​​to construct a stress state index, it automatically locks high-risk operation permissions and switches to a low-density rendering mode when the dispatcher's physiological stress exceeds the threshold. This embeds a human factor safety protection mechanism into the system's underlying layer, proactively preventing safety accidents caused by misoperation under high pressure. Furthermore, the cognitive load index and stress state index are visualized as dynamic attributes of the commander's twin in multi-seat collaborative scenarios using color gradations. Task handover suggestions are automatically generated and mapped to the task allocation data layer, enabling intelligent collaborative scheduling between seats and effectively improving the overall resilience and risk resistance of the multi-seat joint operation system. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a scene diagram for a method of 3D visualization and data mapping of low-altitude economic operation status based on digital twins. Detailed Implementation

[0011] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0012] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort should fall within the scope of protection of this invention.

[0013] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0014] As mentioned in the background section, existing technologies struggle to balance situational prediction accuracy, information delivery efficiency, and human-machine collaboration capabilities in complex and dynamic scenarios. To address these issues, this invention provides a method for three-dimensional visualization and data mapping of low-altitude economic operation situations based on digital twins.

[0015] Reference Figure 1 , Figure 1 This is a flowchart illustrating a method for three-dimensional visualization and data mapping of low-altitude economic operation status based on digital twins, according to an embodiment of the present invention. Figure 1 As shown, a method for three-dimensional visualization and data mapping of low-altitude economic operation status based on digital twins includes: S1: Collect real-time flight status data of all aircraft in the low-altitude airspace, construct the initial state set of digital twins of each aircraft, and use a multi-agent simulation engine with a hybrid decision-making model of rules and reinforcement learning to perform ultra-real-time simulation of the low-altitude operational status and generate a simulation status sequence. S1.1: By deploying an ADS-B ground receiving station network, Doppler radar array, and 5G base station direction finding module in the low-altitude airspace, real-time flight status data of all aircraft in flight in the low-altitude airspace is collected synchronously, and the real-time flight status data is timestamped to generate a sequence of original state frames of multi-source synchronous aircraft. Preferably, the ADS-B ground receiving station network, Doppler radar array, and 5G base station direction finding module each have their own data refresh cycle; using the local UTC clock as a unified reference, a linear interpolation method is used to align the sensor data in the original state frame sequence of the multi-source synchronous aircraft to the same timestamp grid, with the time alignment error not exceeding a preset time error threshold; real-time flight situation data includes real-time position, speed, heading, and mission status data; mission status data includes the aircraft's unique identifier, mission type label, payload status flag, and pre-declared route number; the mission type label ranges from four categories: logistics delivery, emergency rescue, inspection operations, and manned travel, and is declared by the aircraft to the low-altitude operation management platform before takeoff and written into the airborne broadcast message; the preset time error threshold is determined based on the real-time requirements of the multi-sensor fusion system, and is usually set to 1 / 10 of the sensor cycle with the shortest data refresh cycle. For a typical hybrid sensing network composed of an ADS-B system (1Hz refresh rate), a radar array (2Hz refresh rate), and a 5G base station (10Hz refresh rate), the preset time error threshold is set to 10 milliseconds to ensure the real-time requirements of situational awareness; S1.2: Perform coordinate system transformation on the data of each aircraft in the original state frame sequence of the multi-source synchronous aircraft to generate a unified coordinate system aircraft state frame sequence; Specifically, using a preset geodetic coordinate system as the reference, the ADS-B output coordinates and the local plane coordinates output by the 5G base station direction-finding module are transformed to the preset geodetic coordinate system through a seven-parameter Helmholtz transformation and a Gauss-Kruger forward calculation, respectively. The preset geodetic coordinate system adopts the CGCS2000 geodetic coordinate system. The seven-parameter Helmholtz transformation process adopts a three-dimensional spatial similarity transformation method, using three translation parameters, three rotation parameters, and one scale factor to transform the three-dimensional coordinates in the ADS-B WGS-84 coordinate system. Convert to 3D coordinates in CGCS2000 coordinate system The conversion formula is: in, , , ε, ψ, and ω are translation parameters, while ω, ψ, and ω are rotation parameters. These parameters are obtained by calculating the coordinate correspondence of national surveying and mapping benchmark stations, serving as scale factors. The Gauss-Kruger forward calculation converts the local plane coordinates (x,y) output by the 5G base station direction finding module into latitude and longitude coordinates (B,L) in the CGCS2000 geodetic coordinate system. Specifically, the plane coordinates (x,y) are first calculated into the longitude deviation l and latitude B under the Gauss projection using the inverse formula, and then the longitude deviation l is added to the central longitude L0 through iterative calculation to obtain the longitude L. Furthermore, in response to the non-line-of-sight errors in the direction finding data of 5G base stations in densely built urban areas, the occlusion probability of the line-of-sight path from each base station to the aircraft is calculated based on the pre-constructed three-dimensional occlusion model of urban building clusters. Direction finding data with an occlusion probability greater than a preset occlusion threshold are marked as non-line-of-sight suspected error frames. The position components in the non-line-of-sight suspected error frames are corrected by applying an NLOS error compensation vector and then incorporated into the unified coordinate system aircraft state frame sequence. It should be noted that the preset occlusion threshold is calibrated based on the statistical density of typical buildings in the city; the NLOS error compensation vector is pre-calculated by the three-dimensional occlusion model of the urban building cluster combined with the ray tracing algorithm, and stored in the error compensation lookup table with the grid number where the aircraft is located as the index. When querying, the corresponding NLOS error compensation vector can be obtained directly through the grid number index. Furthermore, the construction of a 3D occlusion model for urban building complexes is as follows: Based on the building outline data and height information provided by the government planning department, combined with satellite imagery and lidar point cloud data, a LOD2 level (including roof structure) 3D model of urban buildings is generated using 3D modeling software; the building model is discretized into a voxel mesh using an octree spatial index structure, with the voxel size set to 5m × 5m × 5m; for each 5G base station location, the number of voxels traversed by rays between it and surrounding spatial grid points is pre-calculated, and an electromagnetic wave propagation attenuation model is established based on the intersection of rays and building voxels; the attenuation coefficient from the base station to each spatial grid point is compared with the theoretical value of the standard free space propagation model, and grid points with a difference exceeding a preset attenuation threshold are marked as NLOS areas, with the corresponding attenuation difference serving as the basic data for NLOS error compensation; S1.3: Based on the unified coordinate system aircraft state frame sequence, construct the initial state of the digital twin for each aircraft, and aggregate the initial states of the digital twins of all aircraft in flight in the low-altitude airspace into a set of initial states of the digital twins. Preferably, the initial state of the digital twin is stored in a structured format, with stored fields including the aircraft's unique identifier, three-dimensional coordinates in a unified coordinate system, velocity components, heading angle, airframe type parameters, and mission status data. The airframe type parameters include four categories: UAV, eVTOL, helicopter, and fixed-wing aircraft, and are obtained by querying the aircraft registration information database associated with the aircraft's unique identifier. The initial state set of the digital twin is stored in an in-memory database with the aircraft's unique identifier as the primary key, supporting millisecond-level random access. Specifically, the horizontal accuracy of the position vectors in the initial state set of the digital twin is assessed using circular error. If the circular error of an aircraft exceeds a preset accuracy threshold, the state confidence flag of this aircraft's digital twin is set to low confidence, and its state is downweighted in subsequent simulations. The preset accuracy threshold is determined based on the low-altitude flight safety interval standard. For general urban low-altitude flight environments, the preset accuracy threshold is set to 10 meters, meaning that when the circular error of the position data exceeds 10 meters, the data is judged as low-quality data and needs to be downweighted. For special areas such as the vicinity of airports and critical airspaces such as hospital emergency access routes, the preset accuracy threshold is tightened to 5 meters to ensure high safety requirements. S1.4: A multi-agent simulation engine driven by a set of initial states of digital twins, based on a hybrid decision-making model of rules and reinforcement learning. The hybrid decision-making model of rules and reinforcement learning includes a rule sub-module and a reinforcement learning sub-module. The reinforcement learning sub-module is responsible for outputting the continuous action vectors of each aircraft agent within the constraints of the rule sub-module. The rule sub-module is responsible for executing air traffic separation rules, no-fly zone boundary constraints, and priority avoidance rules. It should be noted that the air traffic separation rules refer to the following: the horizontal separation within the same airspace layer should be no less than 50 meters, and the vertical separation between different airspace layers should be no less than 30 meters; for emergency rescue aircraft, other aircraft must maintain a horizontal separation of more than 100 meters; for manned aircraft, drones must maintain a horizontal separation of more than 150 meters; no-fly zone boundary constraints mean that all aircraft are prohibited from entering permanent no-fly zones designated by the government, including areas around military facilities, important government institutions, airport airspace, and areas around nuclear facilities; for temporary no-fly zones, such as areas over large event venues and emergency response sites, the system will automatically plan detour routes for relevant aircraft; for height-restricted areas, such as areas around high-rise building complexes, aircraft are required to maintain a safe vertical separation of no less than 50 meters from the top of the buildings; priority avoidance rules mean that aircraft priority is determined according to the order of emergency rescue, passenger transport, logistics distribution, and inspection operations, with high-priority aircraft having route priority and low-priority aircraft needing to actively give way; between aircraft of the same priority, potential conflicts are handled according to the "right-hand avoidance" principle. Furthermore, the reinforcement learning submodule uses a near-end policy optimization algorithm for pre-training, and the training environment is a simulated airspace scenario constructed based on historical low-altitude operational data; the state space is the relative position and velocity vector of all neighboring aircraft within a preset perception radius centered on the current aircraft, the action space is a two-dimensional continuous space composed of heading angle increment and velocity increment, and the reward function is a linear weighted sum of task completion reward, interval maintenance reward and conflict penalty term. It should be noted that the Proximal Policy Optimization (PPO) algorithm training process adopts an Actor-Critic network architecture: the Actor network maps the state space to a probability distribution in the action space, and the Critic network evaluates the value function of the current state; the training process is carried out in a simulated urban low-altitude environment containing multiple randomly generated aircraft, each with a random task with independent start and end points; the training iteration adopts a mini-batch update method, collecting 1000 interaction samples per batch, with a learning rate of 0.0003 and a discount factor of 0.99; to avoid policy collapse, a trust region constraint is used to limit the KL divergence between the old and new policies within a preset threshold; the training convergence criterion is that the average reward change rate over 100 consecutive episodes is less than 1%; Furthermore, the collaboration between the rule submodule and the reinforcement learning submodule is as follows: the rule submodule first performs compliance verification on the continuous action vectors output by the reinforcement learning submodule. If the continuous action vectors cause the aircraft to enter the no-fly zone or violate the interval rules, the rule submodule will forcibly correct the continuous action vectors to the most recent compliant action that satisfies the constraints. The corrected action vectors will be used as the final action to be executed at this time step and input to the next state update step. S1.5: Taking the state of each aircraft digital twin in the initial state set of the digital twin as the starting node, the multi-agent simulation engine is driven to perform parallel state updates of all aircraft agents in the low-altitude airspace at a preset simulation step size, and gradually extrapolate to the future preset time period. The snapshots of the global aircraft state output by each extrapolation step are arranged in chronological order to generate the extrapolation situation sequence. Preferably, the multi-agent simulation engine is based on a GPU parallel computing architecture and uses a parallel computing kernel to perform the state updates of all aircraft agents in parallel. In scenarios where the number of aircraft does not exceed the preset concurrency scale, the total time taken to complete all simulation frames meets the timeliness requirement that the ratio of simulation speed to actual time is not less than the preset super real-time multiple. Specifically, for each frame of the global aircraft state snapshot in the simulation situation sequence, the simulation confidence of each aircraft agent state is calculated according to the exponential decay function, with the time difference between the current frame simulation time and the time of acquisition of the initial state set of the digital twin as the independent variable. This confidence is then written as an additional attribute field into the aircraft state record of the corresponding frame in the simulation situation sequence for use in the transparency mapping of visualization rendering in subsequent steps. The simulation confidence decreases monotonically with the increase of simulation time.

[0016] It should be noted that the decay coefficient of the exponential decay function is calibrated based on the historical simulation accuracy; the initial value and lower limit constraint of the simulation confidence are preset according to the minimum requirements of the simulation results credibility in the actual operation scenario, so as to ensure that the last frame of the simulation situation sequence still has reference value. S2: Based on the inferred situation sequence, the pre-constructed confidence decay function is mapped to the rendering transparency and boundary blur parameters of the corresponding twin, forming an inferred visualization frame stream carrying uncertainty coding, and is rendered and output in a split-screen overlay manner with the current real-time three-dimensional situation scene; S2.1: Based on the inference confidence carried by the aircraft state records of each frame in the inference situation sequence, a confidence decay function is constructed. The confidence decay function takes the time difference between the inference time and the acquisition time of the initial state set of the digital twin as the independent variable, and the output value range is constrained to the inference confidence scalar between the preset confidence lower limit and the initial confidence. Furthermore, the confidence decay function adopts an exponential decay form, and the decay coefficient is fitted and calibrated by the historical simulation error statistics. For each aircraft state record in each frame of the simulation situation sequence, the time difference corresponding to the simulation time is substituted into the confidence decay function to calculate the simulation confidence scalar of the aircraft at the simulation time, and the simulation confidence scalar is used as the unified input for subsequent rendering parameter mapping. It should be noted that for aircraft whose state confidence flag is set to low confidence in the initial state set of the digital twin, the inference confidence scalar is multiplied by a preset weighting coefficient based on the result calculated by substituting into the confidence decay function. The preset weighting coefficient is less than 1 to reflect the transmission effect of the initial state data quality on the reliability of the inference result. The preset weighting coefficient is determined based on the correlation analysis between historical data quality and inference accuracy, and is usually set to 0.7. This value is obtained by statistically analyzing the deviation between the inference results of a large number of historical low-quality data and the actual flight trajectory, and can reasonably reflect the attenuation effect of the low confidence of the initial state on the reliability of subsequent inferences. For special cases, such as aircraft that have completely lost ADS-B signals and rely solely on radar or 5G positioning, the preset weighting coefficient is further reduced to 0.5 to more conservatively express the uncertainty of the inference. S2.2: The inference confidence scalar is converted into a rendering transparency parameter through a preset transparency mapping function, and then converted into a boundary fuzziness parameter through a preset fuzziness mapping function. The rendering transparency parameter and the boundary fuzziness parameter together constitute the uncertainty coding parameter pair of this aircraft twin at the corresponding inference time. Furthermore, the preset transparency mapping function is a monotonically increasing piecewise linear function. If the inference confidence scalar is high, the rendering transparency parameter approaches the upper limit of opacity; if the inference confidence scalar drops to the preset lower limit of confidence, the rendering transparency parameter converges to the preset lower limit of transparency, so as to ensure that the aircraft twin in the final frame of the inference still has visual recognizability.

[0017] Specifically, the preset fuzzy mapping function is a monotonically decreasing piecewise linear function. If the inference confidence scalar is high, the boundary fuzzy parameter approaches zero, meaning the boundaries of the aircraft twin are clear. If the inference confidence scalar decreases, the boundary fuzzy parameter increases, and the boundary contour of the 3D model of the corresponding aircraft twin is processed by a Gaussian fuzzy kernel during rendering. The radius of the fuzzy kernel is positively correlated with the boundary fuzzy parameter. It should be noted that the segmentation nodes and slopes of the preset transparency mapping function and the preset fuzzy mapping function are calibrated based on actual visual perception experiments of dispatchers, in order to achieve a balance between the visual saliency of uncertain coding and the overall readability of the scene. S2.3: Based on the uncertainty coding parameter pair, perform three-dimensional rendering on each frame of the global aircraft state snapshot in the inferred situation sequence, and render and output a single frame inferred visualization image; Furthermore, the 3D models of each aircraft twin are configured with alpha channel values ​​based on corresponding rendering transparency parameters, and Gaussian blur kernels are applied to the boundary contours of the 3D models based on corresponding boundary blur parameters. The 3D rendering process is executed in the GPU rendering pipeline, and the rendering flow includes a geometric transformation stage, a rasterization stage, and a fragment shading stage. In the fragment shading stage, the rendering transparency parameters are written as alpha blending coefficients to the fragment output, and the boundary blur parameters are passed to the post-processing blur shader to perform convolution blur processing on the edge pixel regions of the 3D models of the aircraft twins. The 3D models of the aircraft twins are configured according to the aircraft body type parameters. The rendering uses low-polygon models from a pre-built 3D aircraft model library. This library contains simplified mesh models of various typical aircraft types, such as quadcopter, hexacopter, and octocopter standard models for multi-rotor UAVs; distributed electric propulsion configuration models for eVTOLs; single-rotor and dual-rotor models for helicopters; and conventional and flying-wing models for fixed-wing aircraft. Each model undergoes polygon reduction processing to ensure that the number of faces in a single model does not exceed 1,000, thus guaranteeing rendering performance in large-scale scenes. The model materials use PBS (physically based shading) materials, which support ambient occlusion and metallicity / roughness parameter adjustment to improve visual realism. Preferably, for aircraft twins labeled as emergency rescue in the simulation situation sequence, a preset high-brightness tone is superimposed on the uncertainty coding parameter pair to ensure that the emergency rescue aircraft twins maintain visual priority in the simulation visualization image and do not lose visual salience due to the increase in transparency caused by the decrease in simulation confidence. It should be noted that the convolution range of the Gaussian blur kernel only applies to the boundary contour pixel area of ​​the 3D model of the aircraft twin, and does not affect the rendering of the internal filling area of ​​the aircraft twin, so as to preserve the shape recognition of the aircraft twin while conveying the uncertainty of position. S2.4: Combine the single-frame simulation visualization images of all frames in the simulation situation sequence in a time sequence to generate a simulation visualization frame stream, and present the dynamic change process of the low-altitude operation situation in a time sequence continuous playback mode. Furthermore, the frame rate of the simulation visualization frame stream corresponds to the simulation step size of the simulation situation sequence. After the simulation visualization frame stream is generated, it is cached in the video memory frame buffer. The playback control module reads the single-frame simulation visualization image carrying uncertainty coding sequentially from the video memory frame buffer and outputs it to the rendering and compositing module at a time interval of the preset simulation step size of the simulation situation sequence. When the simulation situation sequence is triggered to be re-simulated due to a new round of update of the initial state set of the digital twin, the old simulation visualization frame stream in the video memory frame buffer is overwritten and replaced by the newly generated simulation visualization frame stream. It should be noted that the playback rate of the simulation visualization frame stream allows the scheduler to adjust the speed through the interactive interface, covering the range from slow motion to fast motion, in order to meet the scheduler's differentiated needs for detailed review and rapid preview of the simulation situation. S2.5: Obtain the real-time three-dimensional situation scene of the current low-altitude airspace. The real-time three-dimensional situation scene is driven by the latest frame of the aircraft state frame sequence in the unified coordinate system. The twins of each aircraft are rendered in a three-dimensional space scene with a pre-built three-dimensional city base map as the background, and the real-time three-dimensional situation rendering frame is output. Specifically, the rendering transparency parameter of each aircraft twin in the real-time 3D situation scene is set to the upper limit of opacity, and the boundary blur parameter is set to zero, so as to visually distinguish it from the aircraft twins carrying uncertainty coding in the simulation visualization frame stream; the pre-built 3D city base map includes building outline models, road network models and no-fly zone boundary voxel models, which are loaded into the video memory in an offline pre-rendering manner. The preferred method for constructing the pre-built 3D city base map is as follows: The building outline model is automatically generated based on the 2D building outlines and elevation data from municipal planning data, using LOD (Level of Detail) hierarchical management. Near-field areas use a LOD2 model containing textures and building details, while distant areas use a simplified LOD1 model. The road network model extracts road centerlines from Open Street Map (OSM) data, combines road width attributes to construct a 3D road grid, and assigns different materials according to road level. The no-fly zone boundary voxel model converts the boundary data of various no-fly zones into a semi-transparent voxel set. Different types of no-fly zones use different color codes, such as red for permanent no-fly zones, orange for temporary no-fly zones, and yellow for height-restricted zones. The 3D base map is stored in blocks based on city blocks, using a quadtree index structure and supporting a dynamic loading mechanism with view frustum pruning to ensure that only geographic data within the current field of view is rendered. It should be noted that the refresh frequency of the real-time 3D situation rendering frame is consistent with the refresh cycle of the timestamp grid of the unified coordinate system aircraft state frame sequence to ensure the time synchronization between the real-time 3D situation scene and the actual physical airspace state. S2.6: The simulation visualization frame stream and the real-time 3D situation rendering frame input rendering synthesis module are rendered and output on the same display interface using a split-screen overlay method; Furthermore, the main display area of ​​the display interface renders real-time 3D situation rendering frames, while the auxiliary display area renders the current playback frame of the deduction visualization frame stream. The main display area and the auxiliary display area share the view parameters and spatial coordinate system of the pre-built 3D city base map. Preferably, the viewing angle parameters of the main display area and the auxiliary display area include camera position, pitch angle and zoom ratio. The viewing angle parameters of the two areas are synchronized in real time. The viewing angle interaction operation performed by the scheduler on any area is simultaneously applied to the other area to ensure the intuitive comparison and readability of the main display area and the auxiliary display area under the same spatial view. Furthermore, a rendering time marker bar is superimposed on the screen split line area at the junction of the main display area and the auxiliary display area. The time marker bar displays the time difference between the time of the current playback frame in the auxiliary display area and the current real time in real time in the form of text, so that the dispatcher can intuitively perceive the time offset between the time of the simulation situation and the real-time situation. It should be noted that the split-screen ratio can be dynamically adjusted by the scheduler by dragging the split-screen line. The split-screen ratio adjustment operation does not trigger the re-rendering of the inference visualization frame stream or the real-time 3D situation rendering frame. It only modifies the viewport clipping parameters of the rendering and compositing module to reduce the interaction response latency. S3: Collect the dispatcher's visual dwell time and eye saccade frequency, combine the interactive behavior hot zone distribution to calculate the dispatcher's real-time cognitive load index, trigger the key situation highlighting algorithm, aggregate low-risk aircraft twins into cluster icons, and push high-risk aircraft twins to the center of the field of vision with magnification and highlighted boundaries. S3.1: Using an eye tracker deployed at the dispatcher's workstation, raw eye movement data of the dispatcher is continuously collected at a preset sampling frequency. The raw eye movement data is then filtered and denoised to generate a clean eye movement data sequence. Specifically, the filtering and denoising process removes abnormal sampling points caused by blinking and loss of eye tracker signals. This process combines median filtering with linear interpolation: a sliding window median filter is applied to the pupil center coordinate sequence to filter out abnormal points within the window that deviate from the median value by more than a preset coordinate deviation threshold; the missing sampling point intervals after filtering are then filled in using linear interpolation with the pupil center coordinates of adjacent valid sampling points; the same processing procedure is applied to the pupil diameter sequence to generate a clean eye tracker data sequence. It should be noted that the raw eye-tracking data includes the pupil center coordinate sequence, pupil diameter sequence, and eye-tracking timestamp sequence; the preset sampling frequency is set according to the eye tracker hardware specifications to ensure complete capture of saccade events; the preset coordinate deviation threshold is calibrated according to the upper limit of the normal eye movement speed of the dispatcher, and sampling points exceeding this threshold are judged as invalid points; S3.2: Based on the eye movement cleaning data sequence, the velocity threshold classification algorithm is used to classify eye movement events and generate an eye movement event sequence; Furthermore, the classification results include: identifying consecutive sampling point segments where the displacement velocity of the pupil center coordinates between adjacent sampling points is lower than a preset gaze velocity threshold as gaze events; and identifying sampling points where the displacement velocity is higher than the preset gaze velocity threshold as saccade events. The preset gaze velocity threshold is determined based on the typical velocity boundary between gaze and saccade in human eye physiology, which is usually between 20 and 40 degrees / second. This method takes the median value of 30 degrees / second as the classification standard. This threshold can effectively distinguish between the two basic eye movement modes of human eye: gaze persistence and rapid movement (saccade). For special display environments, such as ultra-wide screens or high-resolution displays, the preset gaze velocity threshold can be adjusted proportionally according to pixel density and viewing distance to adapt to eye movement characteristics under different hardware configurations. Preferably, the visual dwell time of a fixation event is taken as the time span of the continuous sampling point segment corresponding to the fixation event; the saccade amplitude of an eye movement event is taken as the Euclidean distance between the centroids of the pupil centers of two adjacent fixation events; for fixation events in the eye movement event sequence whose duration is lower than the preset minimum fixation duration threshold, they are merged into adjacent saccade events and are not included in the statistics of valid fixation events. Furthermore, the eye movement event sequence is statistically analyzed using a preset sliding time window. Within each sliding time window, the average visual dwell time of the effective fixation event and the frequency of saccade events per unit time are calculated. The frequency of saccade events per unit time is the saccade frequency. The average visual dwell time and saccade frequency are used as the eye movement feature parameter pair at the current moment for subsequent cognitive load index calculation. It should be noted that the length of the preset sliding time window is set according to the time scale characteristics of the cognitive load response, and the window step size is aligned with the preset sampling frequency to ensure the real-time update continuity of the eye movement feature parameter pairs. S3.3: Obtain the scheduler's interaction behavior records on the split-screen overlay rendering output interface, divide the display interface into a uniform grid with a preset spatial resolution, count the cumulative number of interaction events in each grid unit within a preset statistical time window, and generate an interaction behavior hotspot distribution matrix. Specifically, the interaction behavior records include mouse click coordinate sequences, click timestamp sequences, and scroll wheel operation event sequences; the cumulative number of interaction events in each grid cell of the interaction behavior hotspot distribution matrix is ​​smoothed using a two-dimensional Gaussian kernel to eliminate sparse noise in the hotspot distribution matrix caused by pixel-level discreteness of click coordinates, generating a smooth interaction behavior hotspot distribution matrix; the smooth interaction behavior hotspot distribution matrix is ​​then normalized to a preset numerical range by rows and columns to obtain a normalized interaction hotspot intensity matrix; It should be noted that the preset statistical time window and the preset sliding time window use the same time length to ensure the alignment consistency between the normalized interaction heat zone intensity matrix and the eye-tracking feature parameter pair in the time dimension; the preset spatial resolution is set according to the display interface resolution and the typical interaction accuracy of the scheduler. S3.4: Calculate the real-time cognitive load index of the dispatcher based on eye-tracking feature parameter pairs and the normalized interaction hot zone intensity matrix; Preferably, the cognitive load index is obtained by weighted linear fusion of the average visual dwell time, saccade frequency, and entropy value of the normalized interaction heatmap intensity matrix; the entropy value of the normalized interaction heatmap intensity matrix is ​​calculated according to the information entropy formula, and the higher the entropy value, the more dispersed the distribution of interaction behavior, and the more fragmented the scheduler's attention allocation; the cognitive load index is calculated as the sum of the products of each feature component and its corresponding weight coefficient, where the sum of the weights of the average visual dwell time, saccade frequency, and entropy value of the normalized interaction heatmap intensity matrix is ​​constrained by a preset normalization constant; the output value range of the cognitive load index is normalized to a preset standard interval so as to make a unified dimension comparison with the subsequent preset cognitive load threshold; the cognitive load index is continuously refreshed with an update cycle of the same size as the preset sliding time window step; Furthermore, the weight coefficients were calibrated by regression analysis of historical dispatcher eye-tracking data and subjective cognitive load scale scores. The calibration process used the least squares method for fitting, with the goal of minimizing the mean square error between the cognitive load index and the subjective scale scores. S3.5: Trigger the critical situation highlighting algorithm to obtain low-risk aircraft twins and high-risk aircraft twins, and perform spatial aggregation and field-of-view center magnification and highlight boundary rendering on the obtained low-risk aircraft twins and high-risk aircraft twins respectively to generate critical situation highlighting rendering frames. Furthermore, the cognitive load index is compared with the preset cognitive load threshold. If the cognitive load index is greater than or equal to the preset cognitive load threshold, the critical situation highlighting algorithm is triggered to assess the risk level of all aircraft twins in the real-time 3D situation scene and calculate the flight risk score of each aircraft twin. If the cognitive load index is less than the preset cognitive load threshold, the real-time 3D situation scene maintains the standard rendering state and no subsequent highlighting processing is performed. It should be noted that the preset cognitive load threshold is determined based on a psychological load test experiment of professional dispatchers, and the value is 0.75 within the standardized interval [0,1]. This value corresponds to the medium to high intensity workload level reported by dispatchers and is the upper limit threshold for maintaining reliable decision-making ability. In special circumstances, such as during emergency response or large-scale collaborative tasks, the preset cognitive load threshold can be temporarily adjusted to 0.85, allowing dispatchers to withstand a higher cognitive load for a short period of time. Specifically, the flight risk score of each aircraft twin is calculated based on three indicators: probability of flight conflict, probability of intrusion into no-fly zone, and yaw deviation. The three indicators are multiplied by their respective weight coefficients and then summed. The sum of the weight coefficients is constrained to 1, and the accident contribution rate of each risk type in the low-altitude operation safety specifications is statistically calibrated. Preferably, the probability of flight conflict is obtained by comparing the predicted value of the aircraft spacing in the frame corresponding to the current simulation time in the simulation situation sequence with the preset safety interval standard; the probability of no-fly zone intrusion is estimated by the shortest crossing time between the current trajectory vector of the aircraft twin and the voxel model of the no-fly zone boundary; the yaw deviation is the lateral distance between the current position vector of the aircraft twin and the standard route corresponding to the pre-declared route number, and is normalized to a preset numerical range before being included in the weighted summation calculation.

[0018] Furthermore, the flight risk score is divided into two categories based on a preset risk classification threshold: if the flight risk score of the aircraft twin is greater than or equal to the preset risk classification threshold, the aircraft twin is classified into the high-risk aircraft twin set; if the flight risk score is less than the preset risk classification threshold, the aircraft twin is classified into the low-risk aircraft twin set. The preset risk classification threshold is set according to the risk level classification standard in the Low Altitude Operation Safety Specification.

[0019] Furthermore, for aircraft twins in the low-risk aircraft twin set, a hierarchical clustering algorithm based on geospatial proximity is used for spatial aggregation. Adjacent low-risk aircraft twins with a spatial distance less than a preset aggregation distance threshold are merged into cluster icons to reduce the visual interface occupation of low-risk aircraft twins and free up the cognitive resources of dispatchers. Specifically, the hierarchical clustering algorithm adopts a cohesive bottom-up strategy. It constructs a distance matrix using the horizontal components of the position vectors of each aircraft twin. Aircraft twin pairs in the distance matrix that are less than a preset aggregation distance threshold are merged sequentially until the minimum distance between all remaining classes is greater than or equal to the preset aggregation distance threshold. After aggregation, the rendering anchor point coordinates of each cluster are taken as the arithmetic mean of the position vectors of all member aircraft twins in that cluster. The cluster icon is represented by a circular symbol, and the number of low-risk aircraft twins contained in the circular symbol is marked inside.

[0020] It should be noted that when the spatial distance between a certain aircraft twin in the low-risk aircraft twin set and any member in the high-risk aircraft twin set is less than the preset proximity warning distance, the aircraft twin will not participate in the cluster icon aggregation and will be retained in the rendering as an independent twin to avoid obscuring potential risk-related information due to aggregation operations. Furthermore, for the aircraft twins in the high-risk aircraft twin set, they are sorted from high to low according to their flight risk scores. The three-dimensional models of each high-risk aircraft twin are rendered in the center of the field of view of the real-time three-dimensional situation scene at a preset magnification ratio. High-brightness boundary rendering effects are superimposed on the boundary contours of the three-dimensional models of each high-risk aircraft twin to generate key situation highlighting rendering frames. The key situation highlighting rendering frames replace the real-time three-dimensional situation rendering frames and are output to the main display area of ​​the split-screen superimposed rendering output interface. Furthermore, the preset magnification ratio is calibrated based on the flight risk score and visual saliency perception experimental data. The higher the flight risk score, the larger the corresponding magnification ratio. The upper limit of the magnification ratio is constrained to not occlude the rendering area of ​​adjacent high-risk aircraft twins. The highlight boundary rendering effect is completed by overlaying the preset highlight color and preset boundary width on the edge pixels of the aircraft twin 3D model during the fragment shading stage. The preset highlight color is mapped to different hues according to the flight risk score range to distinguish high-risk aircraft twins with different risk levels in the same field of view. Specifically, when the cognitive load index falls below the preset cognitive load threshold, the critical situation highlighting algorithm stops triggering, the real-time 3D situation scene is restored to the standard rendering state before the critical situation highlighting algorithm is triggered, the aggregated cluster icons are expanded and restored to the independent rendering form of the corresponding low-risk aircraft twins, the magnified rendering and highlighting boundary effect of the high-risk aircraft twins are simultaneously canceled, and the interface rendering parameters return to the standard situation display configuration. S4: Collect the dispatcher's heart rate variability and skin conductance values, construct the dispatcher's stress state index, lock high-risk operation permissions through the twin system, switch the 3D scene to low-density rendering mode, and simultaneously trigger multi-channel alarm push. S4.1: By wearing a wearable device on the dispatcher's wrist, the dispatcher continuously collects the photoplethysmography pulse wave signal and the raw skin conductance signal at a preset physiological sampling frequency and performs preprocessing to generate a heartbeat interval sequence and skin conductance cleaning signal; Further, the preprocessing includes: bandpass filtering of the photoplethysmography (PPG) signal to extract the peak interval sequence of adjacent heartbeats; low-pass filtering of the raw skin conductance signal to remove high-frequency motion artifacts; the passband frequency range of the bandpass filtering is set according to the frequency range corresponding to the normal human heart rate to filter out baseline drift and high-frequency noise; the extraction of the peak interval sequence of adjacent heartbeats adopts a peak detection algorithm to locate the maximum point of the filtered PPG signal within a sliding window, and the timestamp difference sequence of adjacent maximum points is used as the heartbeat interval sequence; the cutoff frequency of the low-pass filtering is set according to the physiological response bandwidth of the skin conductance signal.

[0021] It should be noted that the wearable device maintains a real-time connection with the data acquisition terminal at the dispatching position via a wireless communication protocol, and the data transmission delay does not exceed the preset communication delay threshold to ensure the timeliness of the heartbeat interval sequence and skin conductance cleaning signal; the preset physiological sampling frequency is set according to the minimum requirements of time resolution for heart rate variability analysis. S4.2: Extract heart rate variability feature parameters from the heart rate interval sequence to obtain the heart rate variability feature vector; Preferably, the heart rate variability characteristic parameters include the root mean square of the difference between adjacent heartbeat intervals, the standard deviation of the heartbeat interval, and the power ratio of low frequency to high frequency. These heart rate variability characteristic parameters are combined into a heart rate variability feature vector. The root mean square of the difference between adjacent heartbeat intervals is the arithmetic square root of the mean of the squares of all adjacent heartbeat interval differences within a preset heart rate analysis time window. The standard deviation of the heartbeat interval is the standard deviation of all heartbeat interval values ​​relative to the mean within the preset heart rate analysis time window. The power ratio of low frequency to high frequency is calculated by performing a fast Fourier transform on the heartbeat interval sequence to obtain the power spectral density, and then integrating the power spectral densities of the low-frequency band and the high-frequency band respectively and taking the ratio. Furthermore, for intervals in the heartbeat interval sequence that deviate from the mean of adjacent heartbeat intervals by more than a preset multiple for ectopic heartbeat determination, they are identified as ectopic heartbeat artifacts and removed. The missing intervals after removal are filled in by linear interpolation and then substituted into the above heart rate variability feature parameter calculation process to eliminate the interference of ectopic heartbeats on the heart rate variability feature vector. It should be noted that the length of the preset heart rate analysis time window is set according to the minimum requirement for the data segment length in the frequency domain analysis of heart rate variability. The window is updated in a rolling manner with a preset step size to ensure the continuous real-time output of the heart rate variability feature vector. The frequency boundaries between the low-frequency band and the high-frequency band are set according to the physiological frequency band definition of the autonomic nervous system regulation. S4.3: Extract skin conductivity feature parameters based on skin conductivity cleaning signals to obtain skin conductivity feature vectors; Specifically, the skin conductance characteristic parameters include the average skin conductance level, the peak frequency of the skin conductance response, and the average amplitude of the peak frequency of the skin conductance response. These skin conductance characteristic parameters are combined into a skin conductance feature vector. The average skin conductance level is taken as the arithmetic mean of the skin conductance cleaning signal within the corresponding time interval, reflecting the baseline of the dispatcher's continuous tension level. The peak frequency of the skin conductance response is obtained by detecting positive abrupt change points in the first-order difference sequence of the skin conductance cleaning signal, and the number of abrupt change points per unit time is the peak frequency of the skin conductance response. The average amplitude of the peak frequency of the skin conductance response is taken as the average increase in signal amplitude relative to the local baseline at each abrupt change point. It should be noted that the criteria for determining a positive mutation point are that the first-order difference value exceeds the preset skin conductance mutation threshold and the duration is not shorter than the preset minimum response duration, so as to eliminate the interference of signal spikes on the peak frequency statistics of skin conductance response. The preset skin conductance mutation threshold is set according to the typical amplitude range of skin conductance response in healthy adults. S4.4: Concatenate the heart rate variability feature vector and the skin conductance feature vector into a stress physiological feature vector, input it into the pre-trained stress state regression model, and output the scheduler's stress state index; Furthermore, the stress state index is normalized to a preset standard interval. The stress state regression model is constructed using the support vector regression algorithm. The training dataset consists of stress physiological feature vectors and corresponding subjective stress scale scores collected synchronously by multiple schedulers in a simulated low-temperature air conditioning stress experiment. The subjective stress scale score is used as the regression target, and the stress physiological feature vector is used as the input feature. The model is trained after cross-validation to optimize the kernel function parameters. The output of the stress state regression model is normalized to a preset standard interval by a linear mapping and then used as the stress state index output. It should be noted that the stress state index is continuously refreshed with the same update cycle as the preset heart rate analysis time window and is aligned with the cognitive load index using the same time benchmark to ensure that the stress state index and the cognitive load index can be jointly analyzed in time synchronization in subsequent steps. Before deployment, the stress state regression model is individually calibrated to account for the physiological baseline differences of different dispatchers. The calibration process involves collecting the stress physiological feature vectors of each dispatcher in the resting state as individual baseline offsets and applying baseline offset compensation to the stress physiological feature vectors at the model input to reduce the impact of physiological differences between individuals on the accuracy of stress state index assessment. S4.5: The stress state index is compared with the preset fatigue threshold. When the stress state index exceeds the preset fatigue threshold, a permission lock command is sent to the twin system permission management module, whereby the permission management module locks the high-risk operation permissions of the current dispatcher's seat. Furthermore, the scope of high-risk operation permissions lockout includes the execution permissions of mandatory route modification commands, temporary lifting of no-fly zone commands, and emergency landing commands for aircraft. Under the permission lockout state, the operator's interface disables the interactive controls corresponding to the above-mentioned high-risk operation permissions. The disabling is manifested by the corresponding controls becoming unclickable and having a visual lockout status mark superimposed. The twin system permission management module simultaneously sends a permission transfer notification to the multi-seat collaborative management module, which then temporarily transfers the high-risk operation permissions to the standby dispatcher seat that is in normal status according to the preset seat priority rules. It should be noted that the preset fatigue threshold is set based on statistical data of the stress physiological characteristics of dispatchers after long-term high-intensity work, and is verified in conjunction with the minimum standard requirements for the dispatcher's working state in the low-altitude operation safety specifications; the condition for unlocking the permission lock status is that the stress state index is continuously lower than the preset fatigue threshold for a preset recovery duration, in order to prevent frequent switching of the permission lock status due to short-term fluctuations in the stress state index. S4.6: When the stress state index exceeds the preset fatigue threshold, switch both the real-time 3D situation scene and the simulation visualization frame stream to low-density rendering mode. Specifically, in low-density rendering mode, the risk level assessment of all aircraft twins in the current real-time 3D situation scene is re-executed according to the method in step S3.5. If the key situation highlighting algorithm in step S3.5 is already triggered, the already generated low-risk aircraft twin set and high-risk aircraft twin set are directly reused. If the key situation highlighting algorithm is not triggered, when low-density rendering mode is started, all aircraft twins in the scene are classified once according to the same flight risk score calculation logic and preset risk classification threshold. This classification is only used for the rendering classification in this step and does not trigger the field magnification and highlight boundary rendering process of S3.5. Furthermore, the rendering rules for the low-density rendering mode are as follows: all aircraft twins in the low-risk aircraft twin set are rendered in the form of cluster icons, and the cluster icon generation rules are consistent with the S3.5 hierarchical clustering algorithm; aircraft twins in the high-risk aircraft twin set only retain position markers and flight risk score numerical labels, hiding the geometric details of the 3D model; the pre-built city 3D base map is switched to a simplified base map that only retains the wireframe rendering of the no-fly zone boundary voxel model and the polyline rendering of the road network model, and the geometric rendering channel of the building outline model is closed; Furthermore, the reduced geometric rendering complexity shortens the GPU rendering pipeline's single-frame latency, resulting in a higher frame rate in low-density rendering mode compared to standard rendering mode. This helps reduce system rendering latency under stress. The switching process for low-density rendering mode uses a linear interpolation gradient with a preset number of transition frames. The display density of each rendering element transitions frame by frame from standard rendering state to low-density rendering state according to a linear interpolation ratio, avoiding sudden visual interference to the scheduler from instantaneous rendering mode switching. During recovery, a reverse transition is achieved using the same number of linear interpolation frames.

[0022] S4.7: When switching to low-density rendering mode, trigger multi-channel alarm push, which is output in coordination through visual alarm channel, auditory alarm channel and tactile alarm channel; Preferably, the visual alarm channel overlays and renders a stress state alarm banner at the top of the main display area of ​​the split-screen overlay rendering output interface; the rendered content of the stress state alarm banner includes the numerical display of the current stress state index, a high-risk operation permission lock status mark, and a seat number label indicating the target seat whose permissions have been transferred; the pitch and duration of the preset alarm prompt tone are set according to the degree to which the stress state index exceeds the preset fatigue threshold, with the larger the degree of exceedance, the higher the pitch and the longer the duration; the auditory alarm channel plays the preset alarm prompt tone through the audio output device of the scheduling seat, and the tactile alarm channel outputs a preset vibration mode through the vibration module of the wearable device; the vibration frequency of the preset vibration mode corresponds to the grading rules of the preset alarm prompt tone; It should be noted that the multi-channel alarm push will automatically stop after the stress state index drops below the preset fatigue threshold. The stress state alarm banner of the visual alarm channel will exit rendering with a fade-out animation. The auditory alarm channel and the tactile alarm channel will be silent simultaneously. After the alarm stops, the low-density rendering mode will be restored to the standard rendering mode with the same preset transition frame number linear interpolation gradient as when switching. S5: The cognitive load index and stress state index are used as dynamic attributes of the commander twin. The status of each dispatcher is visualized in the multi-seat collaborative 3D scene using color-gradient annotation. Task handover suggestions are generated and mapped to the task allocation data layer of the 3D situation scene. S5.1: Construct a corresponding commander twin for each dispatcher's seat in the dispatch center; Specifically, the static attributes of the commander twin include the seat number, the dispatcher's unique identifier, and the airspace responsibility partition number corresponding to the seat; the dynamic attributes of the commander twin include the real-time updated cognitive load index and stress state index. The commander twins of all seats are aggregated into a commander twin set, which is stored in an in-memory database with the seat number as the primary key. Furthermore, the airspace responsibility zone number corresponds to the low-altitude airspace sub-area that the dispatcher is responsible for supervising under the current shift schedule. The boundary coordinates of the low-altitude airspace sub-area are stored as a sequence of polygon vertices in a preset geodetic coordinate system and maintain a spatial correspondence with the airspace zone layer in the pre-constructed urban 3D base map. The dynamic attributes of the commander twin are updated synchronously with the same update cycle as the cognitive load index and stress state index to ensure the real-time status data of each position in the commander twin set. Furthermore, when the dispatch center performs a shift change operation, the dispatcher's unique identifier and corresponding personalized calibration parameters of the commander's twin are updated synchronously with the shift change event. The individual baseline offset of the stress state regression model is also switched to the baseline offset corresponding to the new on-duty dispatcher to ensure the individual adaptability of the stress state index assessment after the shift change. S5.2: Construct a multi-seat collaborative 3D scene and overlay a seat status visualization layer on top of the real-time 3D situation scene; Specifically, the seat status visualization layer renders the airspace responsibility area of ​​each seat as a semi-transparent polygon over the airspace area corresponding to the pre-built city 3D base map. The fill color level of the semi-transparent polygon is generated by the dynamic attribute mapping of the commander twin corresponding to the seat. Furthermore, the mapping rule for filling color levels is as follows: the weighted average of the cognitive load index and the stress state index is used as the comprehensive state score, and the comprehensive state score is mapped to the preset color level range; Furthermore, the transparency of the semi-transparent polygon in the seat status visualization layer is fixed to the preset overlay transparency to ensure that the color fill of the semi-transparent polygon does not visually obscure the rendering content of the aircraft twin in the real-time 3D situation scene below; the boundary outline of the semi-transparent polygon is rendered with a solid line stroke, and the stroke color and the current color value of the fill color level maintain a correspondence of consistent hue and increased brightness. It should be noted that the comprehensive status score is calculated by weighting the cognitive load index and its corresponding weight coefficient, and the stress state index and its corresponding weight coefficient; the preset color range consists of a continuous gradient spectrum from green, yellow, orange to red; the lower the comprehensive status score, the closer the corresponding color range is to the green end, and the higher the comprehensive status score, the closer the corresponding color range is to the red end; the weight coefficients of the cognitive load index and the stress state index in the calculation of the comprehensive status score are set according to the relative importance assessment of the two types of risks, cognitive overload and physiological stress, in the low-altitude operation safety specifications, and the sum of the two weight coefficients is constrained by a preset normalization constant; S5.3: Overlay and render the commander twin status label of each seat at the geometric centroid of the semi-transparent polygon of the airspace responsibility area of ​​each seat in the multi-seat collaborative 3D scene. Preferably, the display content of the commander twin status label includes the seat number, the dispatcher's unique identifier, the current value of the cognitive load index, the current value of the stress state index, and the color block corresponding to the comprehensive status score. The commander twin status label updates the rendered content synchronously with the refresh cycle of the dynamic attributes. Specifically, the commander's twin status label is rendered in a bulletin board style with a fixed orientation and viewing angle, ensuring that the text content of the commander's twin status label is readable from any tilt and rotation angle on the main control display of the dispatch center. The values ​​of the cognitive load index and stress state index are presented in the commander's twin status label with progress bar graphics and text, and the fill ratio of the progress bar corresponds linearly to the normalized value of the corresponding index. Furthermore, when either the cognitive load index or the stress state index of a certain seat exceeds its respective preset threshold, the commander's twin status label applies a flashing rendering effect to the progress bar corresponding to the threshold-exceeding index. The flashing frequency is set in stages according to the magnitude of exceeding the threshold. For every 0.1 increase in the magnitude of exceeding the threshold, the flashing frequency increases by 0.5Hz, with a base flashing frequency of 1Hz, so as to visually highlight abnormal seats in a multi-seat collaborative 3D scene. S5.4: Calculate the duration of stress state index for each position in the commander twin set, record the fatigue duration value of each position, and generate a suggestion to trigger task handover. Furthermore, the fatigue duration timer is compared with a preset fatigue duration threshold. When the fatigue duration timer for any seat exceeds the preset fatigue duration threshold, the task handover suggestion generation process is triggered. The fatigue duration timer starts timing when the stress state index first exceeds the preset fatigue threshold, and the timing accuracy is consistent with the update cycle of the stress state index. When the stress state index briefly drops below the preset fatigue threshold during the timing period, if the duration of the drop is less than the preset short-term recovery tolerance time, the fatigue duration timer is not reset and continues to accumulate. If the duration of the drop is greater than the preset short-term recovery tolerance time, it is determined that the stress state of that seat has returned to normal, and the fatigue duration timer is reset to zero. It should be noted that the preset fatigue duration threshold is set according to the upper limit requirement for the continuous high-stress operation time of dispatchers in the low-altitude operation safety specifications; the preset short-term recovery tolerance time is set according to the statistical inertial time constant of the stress physiological response, so as to avoid the accidental reset of the fatigue duration time value due to short-term fluctuations in physiological signals. S5.5: Generate the airspace responsibility partition number of the fatigued seat in the process based on the task handover suggestion, filter the candidate receiving seats, and generate a priority list of candidate receiving seats; Specifically, a set of candidate receiving positions is selected from the set of commander twins whose stress state index is lower than the preset fatigue threshold and whose cognitive load index is lower than the preset cognitive load threshold. An airspace adjacency constraint is added, that is, the airspace responsibility area of ​​the candidate receiving position and the airspace responsibility area of ​​the fatigued position have a shared boundary or an adjacency relationship within a preset buffer distance (set to 5 kilometers) in geographical space. Furthermore, the candidate receiving seats are sorted from low to high according to their comprehensive status scores to generate a priority list of candidate receiving seats. The seat ranked first is the recommended primary receiving seat, and the remaining seats are selected as alternative receiving seats in order. When the candidate receiving seat set is empty, that is, when all other seats in the dispatch center are in a state of high stress or high cognitive load, the task handover suggestion generation process outputs an overload alarm for all personnel. The overload alarm information is pushed to the dispatch center supervisor's display terminal through the visual alarm channel and the auditory alarm channel in the multi-channel alarm push, and the dispatch center supervisor intervenes to handle the situation. S5.6: Based on the candidate receiving seat priority list, generate a task handover suggestion data packet and map it to the task allocation data layer of the three-dimensional situation scene; Preferably, the task handover suggestion data package includes the seat number of the fatigue seat, the boundary coordinate sequence corresponding to the airspace responsibility partition number, the flight risk score list of all currently airborne aircraft twins in the airspace, the seat number of the recommended primary receiving seat and the priority list of candidate receiving seats, and maps the task handover suggestion data package to the task allocation data layer of the three-dimensional situation scene; Furthermore, the task allocation data layer is an independent rendering layer superimposed on the multi-seat collaborative 3D scene. After the task handover suggestion data packet is mapped to the task allocation data layer, a task handover marker is superimposed and rendered on the semi-transparent polygon of the airspace responsibility partition of the fatigue seat. The task handover marker is represented by a dynamic arrow graphic. The starting point of the arrow is the geometric centroid coordinate of the airspace responsibility partition of the fatigue seat, and the ending point of the arrow is the geometric centroid coordinate of the airspace responsibility partition of the recommended primary receiving seat. The arrow line is rendered as a dashed line and a flowing animation effect is applied to convey the handover direction information. Specifically, the task allocation data layer synchronously overlays and renders a task handover and acceptance prompt box below the status label of the commander twin at the recommended primary receiving position. The displayed content includes the responsibility zone number of the airspace to be handed over, the number of aircraft to be handed over, and the statistics of the number of twins of high-risk aircraft in the flight risk score list, so that the dispatcher at the recommended primary receiving position can conduct a load pre-assessment before confirming acceptance. Furthermore, the task handover suggestion data packet is synchronously pushed to the operation terminal of the recommended primary receiving position through the internal communication bus of the dispatch center; the dispatcher of the recommended primary receiving position responds through the confirmation interaction control on the operation terminal. When the confirmation response signal returns to the twin system permission management module, the permission management module transfers the supervision permission corresponding to the fatigue position's airspace responsibility partition to the recommended primary receiving position, synchronously updates the airspace responsibility partition number field of the two positions in the commander twin set, and switches the dynamic dashed arrow marker in the task allocation data layer to the handover completion status marker. Furthermore, the handover completion status marker replaces the dynamic dashed arrow with a static solid arrow graphic, the arrow color is switched to the preset completion hue, and a handover completion timestamp is superimposed and rendered in the center of the arrow; the semi-transparent polygonal boundary outline of the spatial responsibility partition of the fatigued seat in the multi-seat collaborative 3D scene is switched to dashed rendering to intuitively indicate that the responsibility handover status of the partition has been completed. It should be noted that if the recommended primary receiving seat dispatcher does not return a confirmation response signal within the preset response waiting time (set to 1 minute), the permission management module will automatically push the task handover suggestion data packet to the second-ranked candidate receiving seat in the candidate receiving seat priority list, and poll in sequence until a confirmation response signal is obtained or all seats in the candidate receiving seat priority list do not respond. In the case of no response, the overload alarm process for all personnel will be triggered.

[0023] In summary, this invention collects real-time situational data of aircraft across the entire low-altitude airspace and constructs an initial state set of digital twins. It then combines a rule-based and reinforcement learning hybrid decision-making model to drive a multi-agent simulation engine for ultra-real-time simulation, achieving a forward-looking prediction of future operational situations. This overcomes the limitation of traditional systems that can only present the current state. By mapping the confidence decay function to the rendering transparency and boundary blurring parameters of the twin, the uncertainty of the simulation is visually encoded and presented synchronously with the real-time scene in a split-screen overlay format. This allows dispatchers to intuitively perceive the credibility of the predicted information while acquiring it, effectively avoiding the risk of misjudgment due to over-reliance on the simulation results. Furthermore, by collecting eye-tracking features and interaction hotspot distribution in real time to calculate the cognitive load index, and adaptively triggering a key situational highlighting algorithm, low-risk aircraft are aggregated. Noise reduction and high-risk aircraft are prioritized for placement in the center of the field of view, achieving dynamic matching between information density and dispatcher cognitive capacity, fundamentally alleviating information overload in high-density low-altitude scenarios. Simultaneously, by collecting heart rate variability and skin conductance values ​​to construct a stress state index, high-risk operation permissions are automatically locked and a low-density rendering mode is switched when the dispatcher's physiological stress exceeds the threshold. This embeds a human factor safety protection mechanism into the system's underlying layer, proactively preventing safety accidents caused by misoperation under high pressure. Furthermore, the cognitive load index and stress state index are visualized as dynamic attributes of the commander's twin in multi-seat collaborative scenarios using color gradations, and task handover suggestions are automatically generated and mapped to the task allocation data layer, enabling intelligent collaborative scheduling between seats and effectively improving the overall resilience and risk resistance of the multi-seat joint operation system.

[0024] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for three-dimensional visualization and data mapping of low-altitude economic operation status based on digital twins, characterized in that, include: Real-time flight status data of all aircraft in operation in the low-altitude airspace are collected, and an initial state set of digital twins of each aircraft is constructed. The low-altitude operational status is then simulated in real time through a multi-agent simulation engine that uses a hybrid decision-making model of rules and reinforcement learning, generating a simulated status sequence. Based on the inferred situation sequence, the pre-constructed confidence decay function is mapped to the rendering transparency and boundary blur parameters of the corresponding twin, forming an inferred visualization frame stream carrying uncertainty coding, and is rendered and output in a split-screen overlay manner with the current real-time three-dimensional situation scene. The system collects the visual dwell time and saccade frequency of the dispatcher, calculates the real-time cognitive load index of the dispatcher by combining the hot zone distribution of interactive behavior, triggers the key situation highlighting algorithm, aggregates low-risk aircraft twins into cluster icons, and pushes high-risk aircraft twins to the center of the field of vision with magnification and highlighted boundaries. Collect dispatchers' heart rate variability and skin conductance values ​​to construct a stress state index for dispatchers. Lock high-risk operation permissions through a twin system, switch the 3D scene to low-density rendering mode, and simultaneously trigger multi-channel alarm push. The cognitive load index and the stress state index are used as dynamic attributes of the commander's twin. The status of each dispatcher is visualized in a multi-seat collaborative 3D scenario using color-coded annotations. Task handover suggestions are generated and mapped to the task allocation data layer of the 3D situation scenario.

2. The method for three-dimensional visualization and data mapping of low-altitude economic operation status based on digital twin as described in claim 1, characterized in that, The inference visualization frame stream includes: Based on the inference confidence carried by the aircraft state records of each frame in the inference situation sequence, a confidence decay function is constructed. The confidence decay function takes the time difference between the inference time and the acquisition time of the initial state set of the digital twin as the independent variable, and the output value range is constrained to the inference confidence scalar between the preset confidence lower limit and the initial confidence. The inference confidence scalar is converted into a rendering transparency parameter through a preset transparency mapping function, and then into a boundary fuzziness parameter through a preset fuzziness mapping function. The rendering transparency parameter and the boundary fuzziness parameter together constitute the uncertainty coding parameter pair of this aircraft twin at the corresponding inference time. Based on the uncertainty coding parameter pair, three-dimensional rendering is performed on each frame of the global aircraft state snapshot in the inferred situation sequence, and a single-frame inferred visualization image is rendered and output. The single-frame simulation visualization images of all frames in the simulation situation sequence are combined in time sequence to generate a simulation visualization frame stream, and the simulation dynamic changes of the low-altitude operation situation are presented in a time sequence continuous playback mode.

3. The method for three-dimensional visualization and data mapping of low-altitude economic operation status based on digital twin as described in claim 2, characterized in that, The generation of the simulated situation sequence includes: Starting with the state of each aircraft digital twin in the initial state set of the digital twin, the multi-agent simulation engine is driven to update the state of all aircraft agents in the low-altitude airspace in parallel with a preset simulation step size, and gradually extrapolate to a preset future time period. The snapshots of the global aircraft state output by each extrapolation step are arranged in chronological order to generate a extrapolation situation sequence.

4. The method for three-dimensional visualization and data mapping of low-altitude economic operation status based on digital twin as described in claim 3, characterized in that, The initial state set of the digital twin includes: Real-time flight status data of all aircraft in flight within the low-altitude airspace is collected, and the real-time flight status data is timestamped to generate a sequence of original state frames for multi-source synchronous aircraft. The coordinate system is transformed into a unified coordinate system for each aircraft data in the original state frame sequence of the multi-source synchronous aircraft. Based on the unified coordinate system aircraft state frame sequence, a digital twin initial state is constructed for each aircraft, and the digital twin initial states of all aircraft in flight in the low-altitude airspace are aggregated into a digital twin initial state set.

5. The method for three-dimensional visualization and data mapping of low-altitude economic operation status based on digital twin as described in claim 4, characterized in that, The initial state set of the digital twin drives a multi-agent simulation engine based on a hybrid decision-making model of rules and reinforcement learning, wherein the hybrid decision-making model of rules and reinforcement learning includes a rule sub-module and a reinforcement learning sub-module; The reinforcement learning submodule is responsible for outputting the continuous action vectors of each aircraft agent within the constraints of the rules submodule; the rules submodule is responsible for executing air traffic separation rules, no-fly zone boundary constraints, and priority avoidance rules.

6. The method for three-dimensional visualization and data mapping of low-altitude economic operation status based on digital twin as described in claim 1, characterized in that, The real-time cognitive load index includes: An eye tracker deployed at the dispatcher's workstation continuously collects raw eye movement data from the dispatcher at a preset sampling frequency. The raw eye movement data is then filtered and denoised to generate a clean eye movement data sequence. Based on the eye movement cleaning data sequence, an eye movement event classification algorithm is used to classify eye movement events and generate an eye movement event sequence. The system obtains the scheduler's interaction behavior records on the split-screen overlay rendering output interface, divides the display interface into a uniform grid with a preset spatial resolution, counts the cumulative number of interaction events in each grid unit within a preset statistical time window, and generates an interaction behavior hotspot distribution matrix. Based on the eye-tracking feature parameter pairs and the normalized interaction heat zone intensity matrix, the real-time cognitive load index of the scheduler is calculated.

7. The method for three-dimensional visualization and data mapping of low-altitude economic operation status based on digital twin as described in claim 6, characterized in that, Comparing the cognitive load index with a preset cognitive load threshold includes: If the cognitive load index is greater than or equal to the preset cognitive load threshold, the key situation highlighting algorithm is triggered to assess the risk level of all aircraft twins in the real-time three-dimensional situation scene and calculate the flight risk score of each aircraft twin. If the cognitive load index is less than the preset cognitive load threshold, the real-time 3D situation scene will maintain a standard rendering state.

8. The method for three-dimensional visualization and data mapping of low-altitude economic operation status based on digital twin as described in claim 7, characterized in that, The flight risk score is classified into two categories based on a preset risk classification threshold, including: If the flight risk score of the aircraft twin is greater than or equal to the preset risk classification threshold, then the aircraft twin will be classified into the high-risk aircraft twin set. If the flight risk score is less than the preset risk classification threshold, the aircraft twin will be classified into the low-risk aircraft twin set.

9. The method for three-dimensional visualization and data mapping of low-altitude economic operation status based on digital twin as described in claim 8, characterized in that, For the aircraft twins in the set of low-risk aircraft twins, a hierarchical clustering algorithm based on geospatial proximity is used for spatial aggregation. Adjacent low-risk aircraft twins with a spatial distance less than a preset aggregation distance threshold are merged into cluster icons to reduce the visual interface occupation of low-risk aircraft twins and free up the cognitive resources of the dispatcher.

10. The method for three-dimensional visualization and data mapping of low-altitude economic operation status based on digital twin as described in claim 8, characterized in that, For the aircraft twins in the set of high-risk aircraft twins, they are sorted from high to low according to their flight risk scores. The three-dimensional models of each high-risk aircraft twin are rendered in the center of the field of view of the real-time three-dimensional situation scene at a preset magnification ratio. Highlighted boundary rendering effects are superimposed on the boundary contours of the three-dimensional models of each high-risk aircraft twin to generate key situation highlighting rendering frames.

Images