A modeling method and system based on physical interfeed

Abstract

The application relates to a modeling method and system based on physical interaction, which comprises the following steps: obtaining a target city model, configuring attribute parameters for each landscape object, discretizing a simulated three-dimensional space into a plurality of semantic attribute grid units, mapping the attribute parameters of each landscape object to corresponding grid units through space intersection, obtaining effective physical attributes of each grid unit, configuring dynamic state variables for each grid unit, constructing an offline feature base library based on the effective physical attributes, interpolating a reference physical field from the offline feature base library according to real-time macro weather input, generating a landscape disturbance field according to the dynamic state variables and the effective physical attributes of each grid unit at the current time, determining a global coupled physical field based on the landscape disturbance field and the reference physical field, updating the dynamic state variables of each grid unit, and forming physical interaction coupling between the landscape and the weather. Thus, the physical fidelity is improved, real-time operation requirements are met, and good universality is achieved.

Description

Technical Field

[0001] This application mainly relates to the field of flight simulation technology, and in particular to a modeling method and system based on physical feedback. Background Technology

[0002] With the rapid development of the urban air mobility industry, electric vertical takeoff and landing (eVTOL) aircraft rely on urban buildings and streets to carry out low-altitude flights. The flight environment is a micro-meteorological environment of urban canyons with meter-level resolution, which puts extremely high requirements on the environmental simulation realism, real-time performance and physical interactivity of the simulator.

[0003] Traditional flight simulators were initially designed for large civil aircraft, with simulation scenarios primarily focused on high altitudes and airport runways. Meteorological and terrain models were independent of each other, only meeting macroscopic flight simulation requirements. However, during low-altitude near-city flights using eVTOL, building shapes, materials, temperatures, and airflows generate strong local wind fields, thermal circulation, eddies, wind shear, and other microscopic meteorological phenomena. Existing technologies cannot achieve two-way physical feedback between terrain and meteorology, resulting in problems such as insufficient simulation fidelity, poor real-time performance, disconnect between physical and visual scenes, and incompatibility with the scale of eVTOL low-altitude operations. Summary of the Invention

[0004] One objective of this application is to provide a modeling method and system based on physical feedback to address the problems of insufficient fidelity, poor real-time performance, separation of physical and visual scenes, and mismatch with the operational scale in existing technologies.

[0005] According to one aspect of this application, a physical feedback-based modeling method is provided for use in a flight simulator for an electric vertical takeoff and landing (EVTOL) aircraft. This modeling method includes: Obtain the collection of landscape objects of the target city model and configure attribute parameters for each landscape object; The simulated 3D space is discretized into multiple semantic attribute grid cells. By spatial intersection, the attribute parameters of each landscape object are mapped to the corresponding grid cell, the effective physical attributes of each grid cell are obtained, and dynamic state variables are configured for each grid cell. An offline feature base library is constructed based on the effective physical properties of the semantic attribute grid cells. A reference physical field is obtained by interpolation from the offline feature base library according to the real-time macro meteorological input. A landscape disturbance field is generated according to the dynamic state variables of each grid cell at the current time and the effective physical properties. The global coupled physical field is determined based on the landscape disturbance field and the reference physical field, and the dynamic state variables of each grid cell are updated according to the global coupled physical field to form a physical feedback coupling between the landscape and the weather.

[0006] Optionally, obtaining the set of landscape objects of the target city model includes: Acquire and fuse multi-source geospatial data of the target city to obtain fused data; The fused data is semantically segmented to output a set of structured landscape objects and generate a multi-level-of-detail model for each structured landscape object, wherein each structured landscape object is associated with triangular mesh geometry and semantic labels.

[0007] Optionally, the modeling method further includes: The real-time parameters of the electric vertical take-off and landing aircraft are obtained, and the real-time parameters are input into a preset disturbance calculation model. The disturbance calculation model outputs flow field disturbance parameters, and the aircraft disturbance field is constructed based on the flow field disturbance parameters. The perturbation field of the aircraft is superimposed on the local position of the aircraft in the global coupled physical field, and used as the input for the calculation of the aerodynamic load of the aircraft.

[0008] Optionally, the step of interpolating from the offline feature base library based on real-time macro-meteorological input to obtain the reference physical field includes: Obtain the current macro-meteorological input parameters, which include at least wind direction, wind speed, solar altitude angle, atmospheric stability, ambient temperature, and ambient relative humidity; Select multiple reference scene points adjacent to the macro-meteorological input parameters from the offline feature base library; Multidimensional linear interpolation is performed on the global basis physics field corresponding to the multiple adjacent reference scene points to obtain the reference physics field.

[0009] Optionally, constructing an offline feature base library based on the effective physical attributes of the semantic attribute grid cells includes: Construct a scene parameter space that covers a set range of meteorological parameters, and take discrete reference scene points within the scene parameter space; For each benchmark scene point, the effective physical properties of the semantic attribute grid unit are used as the surface boundary conditions. The multiphysics coupling control equations are solved simultaneously to obtain the global base physical field corresponding to the benchmark scene point. An offline feature base library is constructed based on the global base physical field corresponding to all benchmark scene points.

[0010] Optionally, the attribute parameters include aerodynamic attribute parameters, and the landscape disturbance field includes landscape-induced horizontal disturbance wind speed components, landscape-induced static pressure disturbance components, and landscape-induced additional turbulent kinetic energy components. Generating the landscape disturbance field based on the dynamic state variables of each grid cell at the current moment and the effective physical attributes includes: Obtain the dynamic state variables and effective physical properties of each grid cell at the current moment; The landscape-induced horizontal disturbance wind speed component is determined based on the preset mapping relationship between the geometric occlusion effect of landscape objects at each grid unit and the aerodynamic property parameters. The static pressure disturbance component induced by the landscape is determined based on air density, base flow wind speed, and synthesized wind speed after disturbance. The additional turbulent kinetic energy component induced by the landscape is determined based on the horizontal disturbance wind speed component induced by the landscape and the building turbulence generation coefficient associated with the urban morphological feature zone to which the grid cell belongs. The value of the building turbulence generation coefficient is determined by the urban morphological feature zone.

[0011] Optionally, the landscape disturbance field further includes a landscape-induced vertical thermal buoyancy disturbance wind speed component, a landscape-induced temperature disturbance component, and a landscape-induced humidity disturbance component. The generation of the landscape disturbance field further includes: Based on the local temperature difference between the surface temperature of each grid cell and the ambient reference temperature, the vertical thermal buoyancy disturbance wind speed component induced by the landscape is determined by combining the gravitational acceleration, thermal buoyancy correction coefficient and near-ground reference height, and the vertical airflow direction is determined based on the sign of the local temperature difference. An exponential decay factor is constructed based on the ground height of the grid unit and the thermal diffusion attenuation height, and the landscape-induced temperature disturbance component is determined based on the local temperature difference and the exponential decay factor. Based on the instantaneous surface water content and building shadow shading coefficient of each grid unit, a weighted difference calculation is performed using the corresponding first empirical correction coefficient and second empirical correction coefficient to determine the landscape-induced humidity disturbance component.

[0012] Optionally, the step of acquiring the real-time parameters of the electric vertical takeoff and landing aircraft, inputting the real-time parameters into a preset disturbance calculation model, having the disturbance calculation model output flow field disturbance parameters, and constructing the aircraft disturbance field based on the flow field disturbance parameters includes: The real-time parameters of the electric vertical takeoff and landing aircraft are obtained, wherein the real-time parameters include real-time spatial position, rotor induced velocity field and rotor speed vector; The real-time parameters are input into a preset rotor downwash disturbance calculation operator. The rotor downwash disturbance calculation operator outputs the three-dimensional disturbance wind speed component induced by the aircraft, the static pressure disturbance component induced by the aircraft, and the turbulent kinetic energy component induced by the aircraft, which together constitute the aircraft disturbance field. The rotor downwash disturbance calculation operator is a fast calculation operator pre-established based on the sheet theory and the finite-length vortex wake model.

[0013] Optionally, the attribute parameters include acoustic attribute parameters, the flight simulator includes a visual rendering channel, a flight dynamics calculation channel, and an acoustic simulation channel, and the modeling method further includes: The same global coupled physical field is synchronously output to the visual rendering channel, flight dynamics calculation channel and acoustic simulation channel; The rendering parameters of the visual scene are updated by the visual rendering channel based on the global coupled physical field; The flight dynamics solution channel extracts the global coupled physical field at the location of the electric vertical takeoff and landing aircraft and uses it as an aerodynamic load input term into the six-degree-of-freedom dynamic equations of the aircraft. The acoustic simulation channel determines the spatial environmental acoustic parameters of the listening point location based on the global coupled physical field and the acoustic attribute parameters.

[0014] Optionally, the modeling method further includes: Using the global coupled physical field as the only physical field input, and based on the corresponding physical quantities of the global coupled physical field, risk subfields are calculated and determined according to a preset flight safety threshold parameter set. The risk subfields include vertical airflow risk subfields, turbulent eddy current risk subfields, thermal airflow disturbance risk subfields, and sensor performance degradation risk subfields. The risk subfields are weighted and summed to obtain a comprehensive risk value, which is then mapped to different risk levels according to a preset grading threshold.

[0015] Optionally, the modeling method further includes: Based on the simulated viewpoint and line of sight, the comprehensive risk value is integrated by line of sight and attenuated to obtain the pixel risk superposition intensity. Based on the pixel risk overlay intensity, the warning color matching the risk level, and the layer transparency coefficient, a risk visualization visual image is generated by overlaying. Extract the risk level corresponding to the center of mass position of the electric vertical take-off and landing aircraft. When the risk level is equal to or higher than the preset alarm level, send an alarm trigger signal.

[0016] According to another aspect of this application, a physical feedback-based modeling system is also provided for use in an electric vertical takeoff and landing aircraft flight simulator. The modeling system includes a digital twin construction module, a physics field generation module, and a coupled solution module. The digital twin construction module is used to obtain a set of landscape objects of the target city model, configure attribute parameters for each landscape object, and discretize the simulation three-dimensional space into multiple semantic attribute grid units. By spatial intersection, the attribute parameters of each landscape object are mapped to the corresponding grid unit to obtain the effective physical attributes of each grid unit and configure dynamic state variables for each grid unit. The physical field generation module is used to construct an offline feature base library based on the effective physical attributes of the semantic attribute grid cells, interpolate the reference physical field from the offline feature base library according to the real-time macro meteorological input, and generate a landscape disturbance field according to the dynamic state variables of each grid cell at the current time and the effective physical attributes. The coupled solution module is used to determine the global coupled physical field based on the landscape disturbance field and the reference physical field, and to update the dynamic state variables of each grid cell in real time according to the global coupled physical field, so as to form a physical feedback coupling between the landscape and the weather.

[0017] According to another aspect of this application, a computer-readable medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the steps of any of the above methods.

[0018] Compared with existing technologies, this application obtains a set of landscape objects from a target city model and configures attribute parameters for each landscape object; it discretizes the simulation 3D space into multiple semantic attribute grid cells, maps the attribute parameters of each landscape object to the corresponding grid cell through spatial intersection, obtains the effective physical attributes of each grid cell, and configures dynamic state variables for each grid cell; it constructs an offline feature base library based on the effective physical attributes of the semantic attribute grid cells, interpolates from the offline feature base library to obtain a reference physical field based on real-time macro-meteorological input, and generates a landscape perturbation field based on the dynamic state variables and effective physical attributes of each grid cell at the current time; it determines a global coupled physical field based on the landscape perturbation field and the reference physical field, and updates the dynamic state variables of each grid cell based on the global coupled physical field to form a physical feedback coupling between landscape and meteorology. This achieves integrated simulation of multiple physics fields and multiple modes, improves physical fidelity, adapts to real-time operation requirements, and has good versatility. Attached Figure Description

[0019] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the specific embodiments of this application will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 A flowchart illustrating a physical feedback-based modeling method according to one aspect of this application is shown. Figure 2 This diagram illustrates the full-process closed-loop calculation of landscape and meteorological physical feedback in one embodiment of this application; Figure 3 A schematic diagram of the framework of a physical feedback-based modeling system provided according to another aspect of this application is shown. Figure 4 This diagram illustrates the overall architecture of a physics-based aircraft simulation system according to one embodiment of this application.

[0020] The same or similar reference numerals in the accompanying drawings represent the same or similar parts. Detailed Implementation

[0021] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the specific embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0022] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein, and therefore this application is not limited to the specific embodiments disclosed below.

[0023] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0024] Existing flight simulation technologies mainly include: (1) Decoupled "static terrain + statistical meteorology" model. This model is the mainstream commercial simulator solution and is the common technology used by mainstream full-motion flight simulators (FFS) such as FlightSafety and CAE, as well as consumer software such as Microsoft Flight Simulator and X-Plane. In terms of terrain modeling, a static geometric model is used. A three-dimensional mesh is constructed using a digital elevation model (DEM) and building outlines (GIS). Buildings are only used as visual obstacles (collision boxes) and radar reflectors, and do not have physical properties (such as roughness and heat capacity). In terms of meteorological modeling, a global / regional statistical model is used, and random noise is generated using the Dryden / Von Kármán turbulence model, or a uniform wind field is generated based on METAR data interpolation. The meteorological system operates independently of the terrain.

[0025] (2) Offline CFD Pre-calculated Wind Field Database (Scientific Research and Route Planning Scheme). This is currently a method used in UAM (Urban Air Traffic) route planning and high-end scientific research simulation to address the complexity of urban wind fields. The core technology is the use of computational fluid dynamics (CFD) pre-calculation. For a specific urban area, fixed boundary conditions are set (such as wind direction intervals of 22.5°, wind speeds of 5m / s, 10m / s, etc.), and RANS or LES simulations are run. The results (velocity field, pressure field, turbulent kinetic energy field) are stored in a large database. Using look-up table interpolation, the current wind vector is linearly interpolated from the database based on the aircraft's current position and the current macroscopic wind direction during simulation.

[0026] (3) Simple dynamic coupling based on aerodynamic meshes (such as the Blade Element Theory variant of X-Plane). The technology used by the X-Plane simulator couples the terrain with aerodynamics at a low order, which is currently the most "coupled" solution in consumer software. Surface roughness mapping is used to map terrain textures (such as grass, concrete, and water) to different surface roughness coefficients (Z0). When the aircraft approaches the ground, different induced turbulence is generated according to the roughness. The wind direction changes with the terrain; in mountainous areas, the wind field will attempt to deflect along the valley direction (by processing the terrain normal vector).

[0027] (4) The microclimate simulation tool is a solution for the building industry, solving the coupled Navier-Stokes equations, energy equations, and radiation equations. It can simulate the urban heat island effect, cold pools caused by building shadows, temperature differences between different surface materials, and the resulting local circulation. It is applied to urban design and environmental assessment. The output is a static, averaged climate analysis map, rather than millisecond-updated flight data.

[0028] However, the existing technology still has the following problems: (1) The contradiction between fidelity and real-time performance: high-fidelity schemes (CFD, microclimate model) are not real-time; real-time schemes (statistical model, roughness mapping) are not high-fidelity.

[0029] (2) The separation of vision and physics: The existing "weather + landscape" integration mainly solves the superposition of the display layer, rather than the interaction of the physical layer.

[0030] (3) Lack of specificity of eVTOL scale: Existing civil aviation simulators focus on "runway headwind shear" or "high-altitude jet stream", and their meteorological grid resolution (kilometer level) is much larger than the "building gap" (meter level) where eVTOL is located, resulting in the complete absence of micro meteorological phenomena in urban canyons.

[0031] To address the aforementioned issues, this application proposes a dynamic coupling modeling method for urban landscape and meteorology based on a physical feedback mechanism. This method upgrades urban architectural landscapes from static visual models and collision bodies into "active environmental participants" with physical properties such as mass, heat capacity, permeability, and radiation characteristics. It performs real-time multi-physics bidirectional interactive calculations with macro-meteorological inputs and aircraft disturbances, and synchronously feeds the interaction results back to the visual rendering, flight mechanics, and acoustic simulation modules. This forms a closed-loop, self-consistent, and dynamically evolving high-fidelity simulation environment, filling the gaps in existing technologies.

[0032] Figure 1 The diagram shows a flowchart of a physical feedback-based modeling method according to one aspect of this application, the modeling method including steps S11 to S14.

[0033] Step S11: Obtain the set of landscape objects of the target city model and configure attribute parameters for each landscape object.

[0034] A landscape object collection refers to various three-dimensional entities in a city, such as buildings, roads, green spaces (vegetation), and water bodies (water bodies). Each object is assigned an attribute parameter, elevating it from a visual object to an interactive physical entity. This attribute parameter can include physical and thermodynamic properties, such as aerodynamic, thermodynamic, hydrological, and acoustic properties. Specifically, buildings can be assigned surface roughness and thermal conductivity; vegetation can be assigned porosity and leaf area index; and roads can be assigned albedo and heat capacity.

[0035] In practical applications, core physical attributes are defined and associated parametrically, and a physical attribute parameter vector is defined for each landscape object in the landscape object set. :

[0036] Among them, aerodynamic property vector , The length is the surface roughness. Permeability (for porous media). To meet the angle of attack of the incoming flow Related shape drag coefficient model; thermodynamic property vector , It refers to surface albedo (short-wavelength reflectance). For surface thermal emissivity, For volumetric heat capacity, Thermal conductivity, For the intensity of the internal heat source of the object; hydrological attribute vector , Surface water absorption rate, To the maximum water holding capacity of the surface, This is the runoff coefficient. Surface evaporation efficiency coefficient; acoustic property vector , For frequency The relevant acoustic reflection coefficient, For frequency The relevant sound absorption coefficient.

[0037] Step S12: Discretize the simulated 3D space into multiple semantic attribute grid cells. Map the attribute parameters of each landscape object to the corresponding grid cell through spatial intersection to obtain the effective physical attributes of each grid cell and configure dynamic state variables for each grid cell.

[0038] To achieve efficient spatial querying and physics calculations, a three-dimensional regular mesh is established, dividing the entire target city space into tiny cubes (i.e., mesh cells) with clearly defined spatial coordinates and indices. The semantic attribute mesh is not only a physical division of geometric space but also carries semantic information about the scene. For example, the entire three-dimensional space can be discretized into a semantic attribute mesh. . It is a three-dimensional regular mesh, each mesh cell By its index ( i , j , k ) and spatial location ( x , y , z )definition.

[0039] After spatial discretization, it is necessary to establish the association between the mesh and the landscape objects and their attribute parameters generated in step S11. Through spatial geometric calculations (such as collision detection or volume / area intersection tests), it is determined which landscape objects or their parts occupy specific mesh cells. Since a mesh cell may contain multiple landscape objects simultaneously, such as building facades and some vegetation, the physical attribute parameter vectors configured in the previous stage (such as aerodynamic, thermodynamic, and hydrological attributes) are weighted and fused according to the geometric proportion occupied by each object within the mesh, thereby calculating the comprehensive effective physical attributes of the mesh cell.

[0040] Specifically, for each unit By using spatial intersection tests, the set of objects occupying the main volume is determined. And calculate its effective physical properties. For example, the effective value of surface roughness can be obtained by area-weighted averaging:

[0041] in, It is an object In unit Internal surface area This is its surface roughness. Similarly, other properties such as thermal conductivity and albedo can also be mapped to the effective physical parameters of the mesh through a similar mechanism.

[0042] Effective physical properties represent the static material characteristics of mesh cells. To support real-time physics simulation and interaction, each mesh cell also needs to be assigned a dynamic physical state that changes over time. These dynamic state variables serve as initial values ​​and iteration vectors for solving physical equations (such as fluid dynamics and thermodynamics equations), enabling the mesh to record and update the physical state of its microenvironment in real time.

[0043] Specifically, a set of dynamic state variables is maintained for each grid cell. .in, For surface temperature, Surface moisture content This represents net radiation flux. During the simulation, these variables are dynamically updated based on physical equations and real-time meteorological inputs.

[0044] Through the complete process of spatial discretization, attribute mapping, and state configuration described above, following the principles of geometric semanticization, parameterization assignment, and hierarchical organization, the traditional visual 3D city model is transformed into a digital twin that can participate in multi-physics interaction, providing a data foundation for coupled simulation.

[0045] In one embodiment of this application, the parametric model can also be calibrated. Specifically, this is done using measured data. or high-fidelity simulation benchmark solution Calibrate the parameterized model. Define the loss function. L To minimize the difference between the model predictions and the reference data:

[0046] in, The set of physical property parameters to be calibrated. For twin-based and parameters Θ A simplified physical model. Adjusted through optimization algorithms. Θ ,make L ( Θ Minimize the physical behavior of the digital twin in typical scenarios to ensure its realism.

[0047] The final output, a structured and parameterized digital twin of the city's multiple physical attributes, can be mathematically represented as a triple:

[0048] in, A collection of semantic objects To create a semantic mesh that carries valid physical attributes, To update dynamic states based on physical laws This is a set of rules. This twin serves as the precise input and interactive foundation for subsequent real-time multiphysics coupling solutions.

[0049] Step S13: Construct an offline feature base library based on the effective physical attributes of the semantic attribute grid cells, interpolate the reference physical field from the offline feature base library according to the real-time macro meteorological input, and generate a landscape disturbance field according to the dynamic state variables and effective physical attributes of each grid cell at the current time.

[0050] To overcome the bottleneck of traditional high-fidelity physics simulations being time-consuming and unable to meet real-time interaction requirements, the computationally intensive macroscopic background field solution is performed offline, while the lightweight microscopic dynamic correction is performed online, establishing a hierarchical decoupled multiphysics coupling solution framework.

[0051] A hybrid solution framework balancing computational accuracy and real-time performance is established, decoupling computational tasks in layers: computationally intensive, high-fidelity multiphysics coupled simulations are pre-calculated and characterized offline, while in the online phase, a lightweight real-time solver rapidly responds to dynamic changes, thus ensuring both physical accuracy and stringent real-time requirements. Mathematically, this framework can be represented as:

[0052] in, A module for building an offline feature base library is constructed to perform high-fidelity, non-real-time simulations and generate features of urban microclimates. As an online real-time response solver, it performs millisecond-level physics reconstruction and interactive calculations while the simulator is running. This indicates the collaboration and data flow coupling between the two modules.

[0053] Specifically, before the simulation system runs, an offline high-performance computing platform is used to treat the previously generated semantic attribute mesh and its effective physical attributes as complex surface boundary conditions. For massive discrete meteorological scenarios covering typical climate conditions of the target city, high-fidelity multiphysics coupled control equations (such as fluid dynamics and energy conservation) are solved simultaneously. The calculated global three-dimensional physical field data is then feature-extracted and stored as structured tensors to construct an offline feature base library.

[0054] Next, during the real-time operation of the simulator, the current real-time macro-meteorological input parameters are received. Since the real-time weather conditions are continuously changing and usually do not completely overlap with the discrete scenes in the offline library, multiple reference scene points adjacent to the current input are selected in the multi-dimensional meteorological parameter space. A lightweight multi-dimensional linear interpolation algorithm is then used to quickly synthesize the basic background physical field under the current weather conditions.

[0055] Since offline base libraries cannot cover the dynamic thermal and hydrological changes of urban underlying surfaces caused by real-time sunlight, precipitation, evaporation, etc., the online solver needs further local correction. It reads the current dynamic state variables (such as real-time surface temperature, instantaneous water content, and net radiation flux) and static effective physical properties of each grid cell in real time. Using a preset lightweight multi-physical perturbation generator, it quickly calculates the perturbation components of the local wind field, pressure field, temperature field, and humidity field, generating a landscape perturbation field.

[0056] For example, based on the real-time dynamic state variables of each semantic grid cell at the current moment. (i.e., surface temperature, surface water content, and net radiative flux), combined with semantic attribute mesh. and the vector of physical parameters of ground features The lightweight ground feature multi-physics perturbation generation operator is generated through a preset method. Quickly calculate the local corrected disturbance field .

[0057] The local corrected disturbance field Used to superimpose onto the macroscopic base meteorological field to characterize local micro-meteorological disturbances caused by building geometry, topographic relief, surface thermals, and surface cover. Its mathematical expression is:

[0058] In the formula, Three-dimensional disturbance wind speed; This is a static pressure disturbance; To add turbulent kinetic energy; For temperature disturbance; Humidity disturbance; Operators for generating multiple physical disturbances of ground features.

[0059] This allows for the superposition and fusion of the macroscopic baseline physical field obtained through offline interpolation with the microscopic landscape disturbance field calculated online in real time, ultimately generating a high-fidelity, real-time interactive global coupled physical field, thus realizing the physical feedback coupling between landscape and meteorology in the digital twin city.

[0060] Step S14: Determine the global coupled physical field based on the landscape disturbance field and the reference physical field, and update the dynamic state variables of each grid cell according to the global coupled physical field to form a physical feedback coupling between the landscape and the weather.

[0061] The macroscopic baseline physical field constructed offline is superimposed and fused with the microscopic landscape disturbance field calculated online in real time to generate a high-fidelity global coupled physical field with real-time interactive capabilities. The dynamic state variables of each grid cell are dynamically updated through the global coupled physical field, thereby realizing bidirectional physical feedback between landscape and meteorology.

[0062] Specifically, in the real-time coupled total physical field synthesis, in order to achieve comprehensive coupling between the landscape and the weather, the physical fields are linearly superimposed to generate the current dynamic coupled physical field of the entire domain:

[0063] in, The macroscopic reference physical field (including background wind speed, pressure, temperature, etc.) is obtained through offline interpolation. The landscape disturbance field is calculated online in real time (caused by the real-time state of landscape objects such as buildings and vegetation). This is an optional aircraft disturbance field that can be superimposed as an additional disturbance field.

[0064] It contains complete three-dimensional physical fields such as wind speed, pressure, turbulent kinetic energy, temperature, and humidity, providing input for subsequent modules (such as the multimodal sensing module and the risk field calculation module).

[0065] The online solver employs an explicit first-order Euler scheme for time discretization, ensuring real-time performance and reducing computational latency. The dynamic state variables of the semantic grid are updated through single-step iterations.

[0066] in, This represents the dynamic state variable of the grid cell at the current moment; The simulation step size is set to 10ms (usually to match the simulator refresh rate). Current total physics field With semantic grid properties Combine these factors to calculate the rate of change of the state variables. The merging results of offline and online operators are achieved by combining the baseline physical field (such as long-term meteorological statistics and static terrain effects) provided by offline operators and the dynamic disturbance field (such as real-time meteorological changes and landscape object interactions) calculated in real time by online operators, thus realizing accurate updates of state variables.

[0067] Through iterative updates, the grid cells respond to changes in the meteorological field in real time, and their state changes (such as surface heating and water evaporation) in turn affect the physical field calculation at the next moment, forming a closed-loop feedback.

[0068] Integrating the entire online calculation process, the total operator expression for real-time solution is:

[0069] in, The current real-time macro-meteorological input (such as wind direction, wind speed, and radiation intensity); Real-time spatial position of the aircraft (if it exists); Reference field data provided for offline feature base library.

[0070] The complete landscape-meteorological dynamic coupling update general equation, combining offline and online operators, is as follows:

[0071] It describes the complete closed-loop process from offline reference field construction, online real-time field overlay to grid state update, ensuring dynamic feedback between landscape and meteorology. That is, the state changes of landscape objects (such as buildings and vegetation) affect the meteorological field through the perturbation field, and the changes of the meteorological field (such as wind speed and temperature) in turn affect the physical state of the landscape (such as surface temperature and humidity).

[0072] In one embodiment of this application, in step S11, multi-source geospatial data of the target city is acquired and fused to obtain fused data; semantic segmentation is performed on the fused data, a set of structured landscape objects is output, and a multi-level-of-detail model is generated for each structured landscape object, wherein each structured landscape object is associated with triangular mesh geometry and semantic labels.

[0073] Single-type geospatial data often suffers from limited perspective, inconsistent accuracy, or limited information dimensions. Therefore, it is necessary to collect heterogeneous data encompassing multiple dimensions, including geometric morphology, surface texture, spatial topology, and physical attributes. Subsequently, through spatial registration, coordinate system unification, data cleaning, and feature-level fusion, these heterogeneous data are aligned and complemented within a unified three-dimensional spatial coordinate system, thereby forming a comprehensive and high-confidence dataset. In practical applications, this involves acquiring and integrating multi-source geospatial data for the target urban area:

[0074] in, It can provide high-precision three-dimensional spatial coordinates for lidar point clouds; For oblique photographic images, realistic surface textures can be provided; Vector data for geographic information systems can provide macroscopic spatial topology and attribute information; As a building information model, it can provide detailed internal structure and physical properties of buildings; It is a high-resolution satellite image.

[0075] For the merged data By employing a semantic segmentation algorithm that combines deep learning with rule-based processing, urban scenes can be identified and divided into a set of specific structured landscape objects such as buildings, roads, green spaces, and water bodies. This transforms unstructured geographic data into a set of structured objects that are easily processed by computers. For each segmented entity, not only is a semantic identity (semantic label) assigned, but a three-dimensional geometric shell (triangular mesh) is also constructed for it. Furthermore, Level of Detail (LOD) technology is used to generate multiple geometric versions of the same object, ranging from coarse to fine.

[0076] In practical applications, the fused data Perform semantic segmentation combining deep learning and rules, and output a structured set of scene objects:

[0077] Each object Associated with its fine triangular mesh geometry With semantic tags . It is a mesh-pattern model, using a multi-level-of-details model; Enumeration is used to represent Type enumeration values. Next, 3D model reconstruction and lightweighting are performed for each object. Generate multi-level detail models:

[0078] In this model, LOD0 is the bounding box, LOD1 is a simple geometry, and LOD2 is a high-fidelity mesh, to accommodate different computational granularities. This allows for dynamic scheduling of models with varying precision in subsequent calculations, based on viewing distance, computational granularity, or hardware resource requirements, achieving an optimal balance between geometric representation accuracy and computational performance.

[0079] In one embodiment of this application, in step S13, a scene parameter space covering a set range of meteorological parameters is constructed, and discrete reference scene points are taken within the scene parameter space; for each reference scene point, the effective physical attributes of the semantic attribute grid unit are used as the surface boundary conditions, and the multiphysics coupling control equations are solved simultaneously to obtain the global base physical field corresponding to the reference scene point; based on the global base physical fields corresponding to all reference scene points, an offline feature base library is constructed.

[0080] To encompass various typical weather and climate conditions that may occur in the target city, a multi-dimensional meteorological parameter space needs to be defined. Continuous meteorological parameters are discretized into a finite number of states according to certain physical meanings and resolutions. By combining and sampling these discrete states, a series of representative benchmark scene points are generated. These scene points constitute the input condition set for subsequent offline high-fidelity calculations, ensuring that the calculation results cover the vast majority of actual meteorological conditions. This constructs a high-fidelity background physical field feature database covering typical meteorological scenarios for the entire urban area, providing physically reliable benchmark solutions for real-time solutions.

[0081] For example, define a multidimensional parameter space that covers typical climatic conditions of the target city:

[0082] in, The direction of the incoming wind is indicated by intervals of 45°. (m / s) is the reference height wind speed, discretized into 11 levels; The solar altitude angle is related to the time of day and the season, and is discretized at 15° intervals. There are six levels for atmospheric stability. K represents the ambient background temperature, covering -20℃ to 40℃, and is discretized in 3℃ increments. Six levels of relative humidity were set for the ambient background. In summary, all scene parameters were combined and sampled to form a finite number of discrete reference scene points.

[0083] For each discrete reference scene point High-fidelity physical field simulations were performed in an offline high-performance computing environment. In this process, the semantic attribute grid cells constructed earlier and their effective physical properties (such as roughness, albedo, and heat capacity) were used as inputs for complex surface boundary conditions. Subsequently, multiple physical governing equations describing fluid motion, heat transfer, radiation exchange, and water evaporation were solved simultaneously to accurately simulate the physical evolution of the urban microenvironment under these meteorological conditions. Finally, the physical state distribution across the entire three-dimensional spatial grid, i.e., the global baseline physical field, was calculated.

[0084] Specifically, for each benchmark scene point The four sets of coupled equations for fluid dynamics, energy conservation, radiation transfer, and surface hydrology were solved simultaneously on an offline high-performance computing platform.

[0085] ① Navier-Stokes governing equations for incompressible fluids (momentum + continuity)

[0086] In the formula: air density; It is a three-dimensional velocity vector; For hydrostatic pressure; It is the viscous stress tensor; It is a volume force (gravity); This is the thermal buoyancy source term, generated by the temperature field gradient.

[0087] ② Atmospheric energy conservation equation (temperature field evolution)

[0088] In the formula: The specific heat capacity of air at constant pressure; The thermal conductivity of air; This is the net radiation flux source term; The heat transfer source term between ground features and air convection is determined by the dynamic state of semantic grid cells. Coupled input.

[0089] ③ Surface radiation transfer balance equation (net radiation of grid cell)

[0090] semantic grid unit :

[0091] In the formula: This refers to solar shortwave irradiance; The unit equivalent albedo; The unit's equivalent thermal emissivity; It is the Stefan-Boltzmann constant; The equivalent radiation temperature of the sky; The instantaneous surface temperature of the unit.

[0092] ④ Surface hydrological evolution equation (dynamic update of water content)

[0093] In the formula: Precipitation intensity; This is the runoff coefficient; Evaporation efficiency; For surface runoff; constraints .

[0094] After solving the physics for all benchmark scene points, the massive computational results need to be structured, packaged, and persistently stored. The global 3D mesh physics data calculated for each scene point is encapsulated into a multidimensional data tensor. The set of tensors corresponding to all scenes together constitutes an offline feature basis library. This basis library will serve as a priori knowledge base in subsequent online real-time simulations, quickly providing high-fidelity background physics through interpolation and other algorithms, thereby significantly reducing the resource consumption of online computation and achieving heavy offline computation and lightweight online access.

[0095] For example, for each scenario The solved global 3D mesh physics field is packaged into a basis field tensor:

[0096] In the formula: For three-dimensional wind speed, For static pressure, For turbulent kinetic energy, For surface temperature, Surface moisture content The net radiation flux is calculated; tensors from all scenes are aggregated and stored in an offline feature basis library. .

[0097] In one embodiment of this application, in step S13, the macro-meteorological input parameters at the current moment are obtained. The macro-meteorological input parameters include at least wind direction, wind speed, solar altitude angle, atmospheric stability, ambient temperature, and ambient relative humidity. Multiple reference scene points adjacent to the macro-meteorological input parameters are selected from the offline feature base library. Multidimensional linear interpolation is performed on the global base physical field corresponding to the multiple adjacent reference scene points to obtain the reference physical field.

[0098] Online solutions can be categorized into overlay calculations for baseline field interpolation and dynamic ground feature correction, and can also overlay aircraft disturbances; all calculations are lightweight, meeting millisecond-level real-time updates. In the macro-meteorological baseline field interpolation, current real-world or pre-defined large-scale meteorological conditions need to be introduced as macro-boundary inputs driving urban micro-meteorological evolution. Macro-meteorological input parameters include at least wind direction, wind speed, solar altitude angle, atmospheric stability, ambient temperature, and relative humidity. These parameters define the overall physical state of the atmospheric environment above the target city and are the driving force for subsequent matching, retrieval, and interpolation calculations in the offline database. For example, obtaining the macro-meteorological parameter vector input from the simulator in real time. ,in, These are the incoming wind direction, reference altitude wind speed, solar altitude angle, atmospheric stability category, ambient background temperature, and ambient background relative humidity.

[0099] Because offline high-fidelity computation is limited by computing power and cannot exhaustively cover all continuous weather states, the real-time input weather parameters usually do not completely overlap with the discrete scene points in the offline baseline library. To obtain the physical field under the current state, it is necessary to find several discrete baseline scene points in a pre-constructed multidimensional weather parameter space that are closest to the current real-time input parameters in each dimension. These neighboring points constitute the multidimensional spatial vertices surrounding the current real-time state point, providing known high-fidelity data support for interpolation calculations.

[0100] By utilizing multiple selected adjacent benchmark scene points and their corresponding high-fidelity global baseline physical field data, multi-dimensional linear weighted calculations are performed based on the relative spatial positions (i.e., interpolation weights) of real-time input parameters among these adjacent points. Through interpolation operators, a physically reliable and continuous basic background physical field under current real-time meteorological conditions is rapidly synthesized without the need for costly online hydrodynamic partial differential equation solving. This achieves a smooth and efficient transition from discrete offline data to continuous real-time conditions.

[0101] For example, for real-time input parameters In a pre-built set of discrete scenes In this process, multiple reference scene points are selected based on their nearest neighbors in various dimensions such as wind direction, wind speed, and temperature, through multi-dimensional spatial distance or grid index calculation. , A multidimensional linear interpolation operator is employed. Adjacent scene points , The corresponding global basis physics tensor and As known input, combined with real-time input parameters Calculate the interpolation weights for each dimension, and finally interpolate to obtain the underlying background field at the current moment:

[0102] This allows for the rapid provision of the three-dimensional spatial distribution of fundamental physical quantities such as wind speed, static pressure, temperature, and turbulent kinetic energy under the current environment.

[0103] In one embodiment of this application, in step S13, the attribute parameters include aerodynamic attribute parameters, and the landscape disturbance field includes a landscape-induced horizontal disturbance wind speed component, a landscape-induced static pressure disturbance component, and a landscape-induced additional turbulent kinetic energy component; the dynamic state variables and effective physical attributes of each grid cell at the current moment are obtained; the landscape-induced horizontal disturbance wind speed component is determined according to the preset mapping relationship between the geometric shading effect of the landscape object at each grid cell and the aerodynamic attribute parameters; the landscape-induced static pressure disturbance component is determined based on air density, base flow wind speed, and post-disturbance composite wind speed; the landscape-induced additional turbulent kinetic energy component is determined according to the landscape-induced horizontal disturbance wind speed component and the building turbulence generation coefficient associated with the urban morphological feature zone to which the grid cell belongs, wherein the value of the building turbulence generation coefficient is determined by the urban morphological feature zone.

[0104] The landscape disturbance field refers to the local disturbance of the landscape to the weather, which can be generated based on the dynamic state variables and effective physical properties of each grid cell at the current moment. First, it is necessary to clarify the core parameters involved in the aerodynamic calculation and the specific physical components of the disturbance field. The dynamic physical state (such as surface temperature, water content, etc.) of each semantic attribute grid cell at the current moment is read in real time, along with the effective physical properties obtained through spatial intersection mapping in the previous stage. Based on the real-time state and static properties, subsequent calculations of local disturbances in the wind field, pressure field, and turbulence field are then performed. For example, the attribute parameters of the landscape object include aerodynamic properties (such as surface roughness length, permeability, shape drag coefficient, etc.). First, the current dynamic state variables of the semantic grid cells are obtained in real time. And effective physical properties.

[0105] Urban landscape objects such as buildings and terrain exert aerodynamic effects on the macroscopic base airflow, including obstruction, bypassing, wake formation, and narrowing of streets and alleys, thereby altering local horizontal wind speeds. By analyzing the geometric characteristics of landscape objects at grid cells (such as windward and leeward sides) and combining their aerodynamic properties, a simplified engineering mapping model between geometric obstruction effects and wind speed changes is established. Using this model, acceleration or attenuation corrections are applied to the base incoming wind speed to calculate the horizontal wind speed disturbance caused by the landscape.

[0106] The horizontal flow disturbance wind speed is calculated using a simplified model based on the building shading coefficient:

[0107] In the formula, The building shading coefficient is a positive value in the windward area of ​​the building, indicating airflow acceleration, and a negative value in the leeward wake area, indicating airflow deceleration. , The base horizontal inflow velocity is obtained through previous interpolation.

[0108] Based on the principle of energy conservation in fluid mechanics (such as the simplified Bernoulli equation), when airflow is slowed down by terrain obstruction or accelerated at a narrow passage, its dynamic and static pressures will interconvert. Based on the ambient air density, the undisturbed base inflow wind speed, and the combined wind speed after adding horizontal disturbances, the static pressure change caused by local wind speed variations is calculated. This allows the simulation to accurately reflect the positive pressure zone on the windward side of the building and the negative pressure zone on the leeward side / narrow passage.

[0109] For example, the static pressure disturbance component induced by the building is calculated as follows:

[0110] In the formula, air density, The base flow velocity. To calculate the horizontal disturbance wind speed; to accurately quantify the physical phenomenon of increased pressure on the windward side and decreased negative pressure on the leeward side.

[0111] The complex underlying surfaces of cities (such as building angles, narrow streets, and buildings of varying heights) tear apart airflow, generating strong mechanical turbulence. To efficiently and physically consistent with calculations of this additional turbulent kinetic energy, a building turbulence generation coefficient is introduced. This coefficient is not a globally fixed value but is dynamically assigned based on the urban morphological feature zone to which the grid cell belongs. The urban morphological feature zone refers to dividing the simulation space into several regions with different aerodynamic roughness characteristics based on morphological parameters such as building density, building height distribution, street width, building geometric sharpness (e.g., right angles, curved facades), and the proportion of open space. These regions include, but are not limited to, open, unobstructed areas, low-rise, uniformly built areas, densely built-up high-rise areas, and narrow street areas. Combined with the calculated modulus of the horizontal disturbance wind speed, the additional turbulent kinetic energy induced by the building geometry of the grid cell is calculated.

[0112] For example, the formula for calculating additional turbulent kinetic energy is as follows:

[0113] In the formula, the building turbulence generation coefficient is... The value is determined by the urban morphological characteristics of the zoning: in open areas without building obstruction (such as squares and large green spaces). ; In areas with flat building facades and low, uniform buildings (such as those without obvious corners or narrow alleys). In areas where strong turbulence is generated, such as right-angled corners of buildings, gaps between buildings, narrow streets and alleys, T-junctions, and leeward areas of high-rise buildings, .

[0114] Furthermore, the landscape disturbance field also includes landscape-induced vertical thermal buoyancy disturbance wind speed component, landscape-induced temperature disturbance component, and landscape-induced humidity disturbance component; the landscape-induced vertical thermal buoyancy disturbance wind speed component is determined based on the local temperature difference between the surface temperature of each grid unit and the ambient reference temperature, combined with gravitational acceleration, thermal buoyancy correction coefficient, and near-ground reference height, and the vertical airflow direction is determined based on the sign of the local temperature difference; an exponential decay factor is constructed based on the ground height of the grid unit and the thermal diffusion attenuation height, and the landscape-induced temperature disturbance component is determined based on the local temperature difference and the exponential decay factor; the landscape-induced humidity disturbance component is determined based on the instantaneous surface water content and building shadow shading coefficient of each grid unit, respectively, by performing weighted difference calculation with corresponding first and second empirical correction coefficients.

[0115] The landscape disturbance field also includes landscape-induced vertical thermal buoyancy disturbance wind speed components, landscape-induced temperature disturbance components, and landscape-induced humidity disturbance components. Urban underlying surfaces (such as building facades, asphalt roads, green spaces, and water bodies) exhibit surface temperatures that differ from ambient air temperatures due to variations in material heat capacity and albedo, thus triggering thermal convection (hot air rises, cold air sinks). The intensity of this thermal drive can be quantified by calculating the local temperature difference between the grid cell surface temperature and the ambient reference temperature. Subsequently, combining gravitational acceleration, empirical thermal buoyancy correction coefficients, and near-ground reference height, a simplified thermal buoyancy physical model is used to calculate the vertical disturbance wind speed. Simultaneously, the mathematical sign of the local temperature difference is directly used to determine the physical direction of the vertical airflow.

[0116] For example, first calculate the local temperature difference disturbance term:

[0117] In the formula, The grid corresponds to the surface temperature of the earth / building. The ambient base temperature (i.e., the ambient reference temperature) is used as the environmental baseline temperature. The vertical thermal buoyancy disturbance wind speed is calculated based on the local temperature difference disturbance term.

[0118] In the formula, g It is the acceleration due to gravity; This is the thermal buoyancy correction factor; This is the near-Earth reference altitude; when At that time, it was determined to be a thermal updraft ( );when At that time, it was determined to be a descending airflow in the cold pool ( ).

[0119] As thermal disturbances from the Earth's surface are transmitted to higher altitudes, they are diluted by atmospheric turbulence and background airflow, causing the disturbance intensity to gradually weaken with increasing altitude. To accurately reflect this physical process, a characteristic parameter of thermal diffusion attenuation height is introduced. Combined with the current ground elevation of the grid cells, an exponential attenuation factor is constructed. Multiplying this attenuation factor by the local temperature difference at the Earth's surface allows for the calculation of the temperature disturbance component at any altitude in three-dimensional space, thus achieving a reasonable attenuation distribution of temperature disturbances across the vertical profile.

[0120] For example, the grid temperature perturbation component is calculated as follows:

[0121] In the formula, z This refers to the height of the grid cell above the ground. The thermal diffusion attenuation height is determined by the exponential decay factor. This ensures that the higher the altitude, the weaker the temperature disturbance, which conforms to the physical laws of thermal diffusion in the urban boundary layer.

[0122] Local humidity variations in urban microenvironments are primarily influenced by the combined effects of surface water evaporation (humidification) and intense evapotranspiration caused by solar radiation (potentially accompanied by localized water vapor diffusion and dehumidification). To address this, the current instantaneous surface water content (reflecting the source of water evaporation) and building shadowing coefficient (reflecting differences in solar radiation intensity) of each grid cell are extracted and assigned first and second empirical correction coefficients, respectively. Through weighted difference calculations, the humidification effect in waterlogged / humid areas and the dehumidification / diffusion effect in areas of intense sunlight are quantified, ultimately determining the local relative humidity disturbance component.

[0123] For example, based on differences in surface water holding capacity and shading evaporation, the formula for calculating the humidity disturbance component is:

[0124] In the formula, This refers to the instantaneous surface water content. The building's shadow shading coefficient; , These are the first and second empirical correction coefficients, respectively. This difference calculation makes the humidity disturbance in waterlogged or damp areas positive, while the humidity disturbance in areas outside the building's shade due to solar evaporation is negative.

[0125] In one embodiment of this application, real-time parameters of the electric vertical take-off and landing aircraft are obtained, and the real-time parameters are input into a preset disturbance calculation model. The disturbance calculation model outputs flow field disturbance parameters, and an aircraft disturbance field is constructed based on the flow field disturbance parameters. The aircraft disturbance field is superimposed on the local position of the aircraft in the global coupled physical field as the input for the calculation of the aircraft's aerodynamic load.

[0126] During aircraft dynamics simulation, the aircraft's rotor rotation and fuselage motion generate strong aerodynamic disturbances to the surrounding air (such as downwash and wake vortices). To assess these disturbances in real time, it is necessary to continuously collect the aircraft's current motion and aerodynamic state parameters and input them into a pre-defined lightweight disturbance calculation model. This avoids time-consuming computational fluid dynamics (CFD) iterations and uses engineering-simplified algorithms to quickly solve for the flow field changes caused by the aircraft itself. For example, the real-time flight state of the eVTOL aircraft can be acquired at millisecond-level frequencies and input into a pre-built fast calculation operator for rotor downwash and wake disturbances to meet the stringent real-time requirements of the simulator.

[0127] The disturbance calculation model, based on input parameters, calculates the multi-dimensional physical quantity changes (such as wind speed changes, pressure changes, and turbulence enhancement) induced by the aircraft in three-dimensional space. By distributing and synthesizing these physical quantities on a spatial grid around the aircraft, a disturbance field characterizing the aerodynamic disturbance characteristics of the aircraft itself can be constructed.

[0128] When an aircraft flies at low altitudes over a city, the airflow it actually experiences is a superposition of the urban micro-meteorological environment and the aircraft's own aerodynamic disturbances. The constructed aircraft disturbance field is precisely superimposed onto the local spatial location of the aircraft within the previously calculated global coupled physical field of the city. This high-fidelity flow field fusion provides extremely realistic and complex aerodynamic load inputs for the aircraft's six-degree-of-freedom dynamic model, thereby accurately assessing the aircraft's flight attitude, control response, and safety in complex urban wind fields.

[0129] Further, the real-time parameters of the electric vertical takeoff and landing aircraft are obtained, including real-time spatial position, rotor-induced velocity field, and rotor speed vector. The real-time parameters are input into a preset rotor downwash disturbance calculation operator, which outputs the three-dimensional disturbance wind speed component induced by the aircraft, the static pressure disturbance component induced by the aircraft, and the turbulent kinetic energy component induced by the aircraft, which together constitute the aircraft disturbance field. The rotor downwash disturbance calculation operator is a fast calculation operator pre-established based on sheet theory and a finite-length vortex wake model.

[0130] To accurately and in real-time assess the aerodynamic disturbances generated by the aircraft on the surrounding air, the system needs to continuously collect the core motion and aerodynamic state variables that determine the spatial distribution and intensity of the disturbances. These parameters are then input into a pre-defined lightweight disturbance calculation operator. This operator abandons time-consuming computational fluid dynamics (CFD) iterations and employs a simplified engineering algorithm to adapt to the simulator's millisecond-level online solution requirements.

[0131] For example, the real-time parameters acquired for an electric vertical takeoff and landing (eVTOL) aircraft specifically include: the aircraft's real-time spatial position. Rotor-induced velocity field and rotor speed vector These parameters are input into a preset fast calculation operator for rotor downwash and wake disturbance. In this context, the operator does not involve CFD iteration and is used for real-time, fast computation.

[0132] The operator employs two classical aerodynamic engineering simplification theories to solve for the three-dimensional disturbance wind speed components, including vertical and horizontal disturbance wind speed components. First, it uses a simplified model based on strip theory to calculate the average induced downwash velocity of the rotor disk and establishes a spatial attenuation model to characterize the vertical downwash airflow. Second, it uses a finite-length, equal-intensity straight vortex model to simulate the tip vortex, calculates the horizontal tangential wind speed induced by the tip vortex using the Biot-Savart law, and introduces a dissipation coefficient to simulate the attenuation of the vortex with distance. Finally, the disturbances from all rotors are linearly superimposed.

[0133] In practical applications, the single-rotor foundation-induced downwash model (a simplification of the strip theory used to calculate the vertical disturbance wind speed component) is based on rotor thrust. T Rotor radius R Solve for the average shuffling velocity of the paddle disc:

[0134] In the formula, air density; The linear velocity of the propeller tip. This represents the rotor speed.

[0135] Next, the attenuation distribution of the downward wash in the space below the rotor disk is calculated along the vertical downward direction of the rotor. z To establish an attenuation model and perform horizontal radial diffusion correction, calculate the vertical downwash disturbance wind speed of the aircraft (negative values ​​for downward movement):

[0136] in, The horizontal distance from a grid point to the rotor center; This refers to the rotor height; The horizontal diffusion radius of the wash; The vertical attenuation height is the height of the wash.

[0137] The rapid calculation model for propeller tip vortex disturbance (based on a finite straight vortex line model for calculating the horizontal disturbance wind speed component) uses a finite-length, uniformly strong straight vortex line as the wake model. Each rotor generates two symmetrical tip vortices extending rearward along the flight direction. The simplified Biot-Savart formula for the induced velocity of a single vortex line is as follows:

[0138] Among them, tip vortex circulation Calculated from rotor thrust:

[0139] h The perpendicular distance from the grid point to the vortex line; , The angle between the grid point and the two ends of the vortex line; Here, represents the eddy-induced tangential unit vector. To simulate the dissipation of eddy intensity with distance and time, an eddy dissipation coefficient is introduced to reduce far-end disturbances: The wake vortex dissipation coefficient is taken as 0.08~0.15 for urban low-altitude areas; The distance from the grid point to the vortex initiation point is given. Finally, the downwash and wake vortex-induced wind speeds of all rotors are linearly superimposed to obtain the total horizontal disturbance wind speed of the aircraft. , .

[0140] In addition to wind speed changes, the operator also needs to calculate the local static pressure changes caused by airflow acceleration (based on the principle of energy conservation in fluid dynamics), as well as the additional mechanical turbulence caused by tip vortex breaking and wake shear layer (based on a linear correlation empirical model). Finally, the calculated three-dimensional wind speed disturbances (including horizontal and vertical), static pressure disturbances, and turbulent kinetic energy disturbances are combined to form a complete aircraft disturbance field.

[0141] For example, changes in the flow pressure of the underflow air beneath the rotor generate localized static pressure disturbances: the airflow accelerates beneath the rotor, reducing static pressure and creating a negative pressure zone. The formula for calculating static pressure disturbances is:

[0142] Rotor-induced turbulent kinetic energy disturbance Additional turbulence is generated by tip vortex breaking and wake shear layer, which is calculated using a simplified linear correlation model:

[0143] in, The turbulence generation coefficient for rotors is 0.12 to 0.25 for conventional multi-rotor eVTOLs; the upper limit is taken near the rotor disk and the core region of the wake vortex, and the lower limit is taken far away from the wake region.

[0144] In summary, the overall disturbance field of the aircraft, by integrating the above components, is represented as follows:

[0145] The disturbance field comprehensively encompasses the three-dimensional disturbance wind speed component induced by the aircraft, the static pressure disturbance component induced by the aircraft, and the turbulent kinetic energy component induced by the aircraft.

[0146] In one embodiment of this application, the attribute parameters include acoustic attribute parameters, and the flight simulator includes a visual rendering channel, a flight dynamics calculation channel, and an acoustic simulation channel; the same global coupled physical field is synchronously output to the visual rendering channel, the flight dynamics calculation channel, and the acoustic simulation channel; the visual rendering channel updates the rendering parameters of the visual image according to the global coupled physical field; the flight dynamics calculation channel extracts the global coupled physical field at the location of the electric vertical take-off and landing aircraft and uses it as an aerodynamic load input term into the six-degree-of-freedom dynamic equations of the aircraft; the acoustic simulation channel determines the spatial environmental acoustic parameters of the listening point location based on the global coupled physical field and the acoustic attribute parameters.

[0147] The multiphysics tensor output in real time by the online solver is used as a unified input to synchronously drive the visual rendering channel, flight dynamics solution channel, and acoustic simulation channel of the flight simulator, achieving closed-loop interaction between environmental perception and flight control. For example, the online solver outputs a global real-time coupled multiphysics tensor as follows:

[0148] in, = For any grid coordinate in space; This represents a three-dimensional wind speed vector field. For fluid pressure field; For turbulent kinetic energy fields; For the surface / air temperature field; For the surface water content field; This is the net radiation flux field.

[0149] The parallel channels of multimodal perception simulation include a visual rendering channel, a flight dynamics channel, and a 3D acoustic channel, each channel being configured with... To unify the input, modal mapping operators are constructed separately to achieve quantitative conversion of physical fields into simulated sensing quantities.

[0150] The visual rendering pipeline renders dynamic environments using a unified rendering engine, updating rendering parameters for the visual scene. For example, it inputs coupled physical fields. Urban Semantic Grid Physical attributes of land features Through visual rendering operators Convert the physical field into a pixel-level rendering visual texture parameter field. To achieve dynamic environment visualization:

[0151] This channel is broken down into the following four types of visual sub-effect models, all of which are driven by coupled physical fields: ① Rendering of building hot spots and surface thermal radiation (temperature field) drive) The pixel equivalent radiance model is as follows:

[0152] in, This is the solar elevation angle irradiance attenuation function; , Equivalent emissivity and albedo of the grid cells; Overlaying onto the base texture generates the visual effect of high-temperature hot spots on glass curtain walls and roofs.

[0153] ② Fog and cold pool visibility field (driven by temperature and humidity)

[0154] Local saturated vapor pressure and visibility attenuation coefficient:

[0155] in, The critical relative humidity for fogging; The atmospheric extinction coefficient is used. Integral extinction is performed along the view path to achieve cold fog rendering of streets and cold pools of building shadows:

[0156] ③ Surface water accumulation and specular reflection of rain streaks (water content field) drive)

[0157] Equivalent specular reflectance of the Earth's surface:

[0158] in, For dry surface reflectivity, Water reflectance; when precipitation input is present At that time, the longitudinal rain mark intensity of the building facade was:

[0159] in, The wind speed is horizontal; the higher the wind speed, the more pronounced the stretching of rain streaks.

[0160] ④ Spatial motion trajectory of rain and snow particles (three-dimensional wind field) drive)

[0161] Differential equations of motion for rain and snow particles:

[0162] in, The natural falling velocity of rain and snow particles is used; the real-time position of the particles is solved by numerical integration to achieve a dynamic effect of oblique rain and snow that deflects with the local wind field.

[0163] Combining the above four types of visual sub-effects, the final screen output pixel color is:

[0164] in, Based on the base landscape texture colors The ambient light level of the sky is adjusted. An immersive visual scene is achieved by integrating sub-effects such as basic textures, thermal radiation, specular reflection, and rain streaks.

[0165] The flight dynamics solution channel extracts the global coupled physical field at the location of the eVTOL and uses it as an aerodynamic load input term to input the six-degree-of-freedom dynamic equations of the aircraft.

[0166] At the center of mass of the spacecraft place, from The local wind field, turbulent kinetic energy, and pressure gradient are extracted using the following formulas:

[0167] in, It is a three-dimensional mesh trilinear interpolation operator.

[0168] Based on turbulent kinetic energy Dryden turbulence perturbation superposition is performed. The turbulence velocity perturbation components are:

[0169] in, The normalized Gaussian random turbulence vector generated for the Dryden model; the total effective inflow velocity is:

[0170] The pressure gradient produces additional acceleration; the volumetric acceleration caused by the pressure gradient is:

[0171] Substitute the above physical field quantities into the eVTOL six-degree-of-freedom dynamic equations. Assume the aircraft attitude... Body speed Rotor aerodynamic load The governing equations for the dynamics are:

[0172] in, For the mass of the aircraft, For rotational inertia, For turbulent aerodynamic forces and moments, The rotor force / torque is determined by the coupled wind field. Real-time correction enables dynamic disturbance of flight attitude and control response by urban canyon winds.

[0173] The acoustic simulation channel determines the spatial environmental acoustic parameters of the listening point location based on the globally coupled physical field and acoustic property parameters.

[0174] In acoustic simulation of three-dimensional spatialized environments, acoustic synthesis operators With coupled wind field acoustic properties of ground features Input, output spatial location The ambient sound pressure level field at the (pilot's cockpit) location. This includes the following four components: ① Free-flow wind noise basic sound source Local wind speed corresponds to wind noise source intensity:

[0175] in, The reference wind speed is used as the reference sound level.

[0176] ② Correction item for building canyon eddy noise

[0177] Based on turbulent kinetic energy Increase eddy current howling noise:

[0178] Total sound source sound level:

[0179] ③Architectural acoustic reflection / absorption propagation attenuation

[0180] Sound from the source Distributed to listening points The path passes through the building grid cells, frequency f Total attenuation:

[0181] in, Geometric distance attenuation; The sound absorption coefficient of the grid unit; is the sound reflection coefficient.

[0182] ④ Real-time sound pressure level at the cockpit listening point

[0183] For the full frequency f Integral synthesis creates three-dimensional spatial sound effects, enabling the dynamic changes of canyon wind sounds, building windbreak noise reduction, and corner vortex howling with the real-time wind field.

[0184] The three channels described above are solved in parallel, all sharing the same coupled physical field input. :

[0185] The three-channel output is simultaneously sent to the simulator terminal, forming an integrated immersive simulation that combines vision, motion control, and hearing; at the same time, each mode provides real-time feedback of the aircraft's position and attitude to the online solver, completing a complete closed-loop feedback between the landscape, weather, aircraft, and multimodal perception.

[0186] In one embodiment of this application, the global coupled physical field is used as the only physical field input. Based on the corresponding physical quantities of the global coupled physical field, risk subfields are calculated and determined according to a preset flight safety threshold parameter set. The risk subfields include vertical airflow risk subfields, turbulent eddy current risk subfields, thermal airflow disturbance risk subfields, and sensor performance degradation risk subfields. The risk subfields are weighted and summed to obtain a comprehensive risk value. The comprehensive risk value is mapped to different risk levels according to a preset grading threshold.

[0187] The global coupled physics tensor output by the online solver is used as the sole physics input for risk calculation:

[0188] in, For any grid coordinate in space; This represents a three-dimensional wind speed vector field. It is static pressure; For turbulent kinetic energy fields; For vorticity field; For the surface / air temperature field; For the surface water content field; This represents the relative humidity field.

[0189] It should be noted that the risk calculation process also involves non-physical field parameters, such as urban semantic grids and flight safety threshold parameter sets. These are configuration parameters rather than physical field data, and therefore do not constitute a negation of the unique physical field input.

[0190] Specifically, a general generation operator for risk fields is defined. Input Coupled Physical Field Urban Semantic Grid eVTOL safety threshold parameter set Output a standardized risk scalar field across the entire domain. :

[0191] The preset flight safety threshold parameter set is as follows: These correspond to the critical thresholds for vertical wind speed, turbulent kinetic energy, vorticity, visibility, slippery surface, and electromagnetic interference, respectively. Furthermore, the risk grading stage also involves grading thresholds. , , The aforementioned threshold parameter set Γ constitutes the criterion for determining whether each physical quantity has reached the risk level, and can be dynamically adjusted according to eVTOL aircraft type, flight altitude, and other conditions.

[0192] The risk subfield represents the quantitative scalar distribution of various flight risks on a spatial grid. It consists of four independent risk subfields superimposed, then normalized, graded, and visualized. The overall process is divided into four sections: sub-risk field solution, comprehensive risk fusion, risk grading mapping, and visualization layer rendering. The first three sections will be illustrated with examples below.

[0193] First, we will develop quantitative solution formulas for four types of independent flight risk subfields.

[0194] (1) Vertical airflow risk subfield

[0195] This subfield is used to identify the risks posed by micro-downflow / upflow to the vertical control of eVTOL.

[0196] from The corresponding physical quantity extracted is three-dimensional wind speed. of z Vertical wind speed component Define the vertical airflow risk factor:

[0197] In the formula: The vertical airflow velocity of the grid cell is positive for upward and negative for downward. The threshold for mild vertical airflow alarm; The threshold for severe danger of vertical airflow; The higher the value, the greater the risk of downward / upward airflow.

[0198] (2) Turbulent eddy risk subfield

[0199] This subfield is used to identify threats to the attitude stability of aircraft, such as building leeward wakes, street corner separation vortices, and strong turbulence zones between buildings.

[0200] from The corresponding physical quantity extracted is turbulent kinetic energy. With vorticity modulus Joint weighted calculation of turbulent eddy current risk:

[0201] Among them, the constraints are: ; , For turbulent kinetic energy and vorticity weighting coefficients, ; The threshold for light / heavy turbulent kinetic energy; The threshold for light / heavy eddy current is used to identify building leeward wakes, street corner vortices, and strong turbulent regions between buildings.

[0202] (3) Risk field of thermal airflow disturbance

[0203] This subfield is used to identify disturbances to eVTOL control caused by thermal buoyancy vertical airflow above high-temperature buildings and asphalt roads due to surface temperature differences.

[0204] from The corresponding physical quantity extracted is the temperature field. First, calculate the vertical temperature gradient:

[0205] in, For vertical temperature gradient, It is a vertical unit vector; The hot gas flow risk scalar is calculated using a piecewise linear function:

[0206] Strong upward hot air currents form above high-temperature buildings and asphalt roads, increasing the disturbance to eVTOL control.

[0207] (4) Sensor performance degradation risk field

[0208] This subfield is used to quantify the degrading impact of environmental factors on the performance of eVTOL airborne sensors (vision, lidar, radio communication, etc.).

[0209] from The extracted physical quantities include relative humidity, surface water content, and turbulent kinetic energy, which correspond to three sub-factors: visibility in fog and rain, reflectivity on slippery surfaces, and radio interference from turbulence, respectively. ① Fog and rain visibility attenuation factor:

[0210] ②Reflective interference factors of slippery ground surface:

[0211] ③ Turbulent radio interference factor:

[0212] Then, the three sub-factors are weighted and fused:

[0213] constraint ; , , These represent the weights of each interference item.

[0214] The weighted summation of the aforementioned risk subfields yields the comprehensive risk value. The weighted superposition of the four types of individual risk fields yields the total risk value for the grid cell. : For various risk weighting coefficients, satisfying Weighted summation followed by unified normalization constraint:

[0215] in, 0 represents no risk, and 1 represents an extremely high-risk area.

[0216] The comprehensive risk value is mapped to different risk levels according to a preset grading threshold. Continuous risk scalars are mapped to four levels of security alarms using a discrete piecewise function:

[0217] in, , , The preset risk classification threshold can be dynamically adjusted based on the eVTOL aircraft type and flight altitude. This classification result can directly drive the rendering of risk visualization layers in the flight simulator. For example, risk heatmap layers can be overlaid on the visual screen using different color codes (green / yellow / orange / red), allowing pilots to perceive the safety situation of the airspace ahead in real time during immersive simulation.

[0218] Furthermore, the comprehensive risk value is integrated and attenuated based on the simulated viewpoint and line of sight to obtain the pixel risk superposition intensity; based on the pixel risk superposition intensity, the warning color and layer transparency coefficient matching the risk level are superimposed to generate a risk visualization scene; the risk level corresponding to the center of mass position of the electric vertical take-off and landing aircraft is extracted, and when the risk level is equal to or higher than the preset alarm level, an alarm trigger signal is sent.

[0219] In the visualization layer rendering of the risk field, the comprehensive risk value is first integrated and attenuated based on the simulation viewpoint and the direction of the line of sight to obtain the pixel risk superposition intensity.

[0220] Simulated viewpoint and line of sight are based on the pilot's viewpoint. Direction of sight Based on this, the starting point and direction of the three-dimensional spatial ray corresponding to each pixel in the simulator's view are determined. For any pixel in the view... Its corresponding line of sight ray comes from Departure, along The direction extends to the far-view cutting surface. .

[0221] Line-of-sight integration refers to spatially integrating the comprehensive risk value of each grid cell along the aforementioned line-of-sight ray, thereby projecting the three-dimensional risk scalar field onto the two-dimensional viewpoint to generate the risk superposition intensity corresponding to each pixel. The specific formula is as follows: The field of view is the clipping plane; the exponential term is the fog visibility attenuation correction. The exponential attenuation term in the formula realizes the fog visibility attenuation correction, that is, the farther away from the pilot and the stronger the fog and rain extinction in the path, the smaller the contribution of the risk value to the pixel superposition intensity. Thus, visually, distant risks are blurred by fog, while nearby risks are clear and conspicuous, which is consistent with the perception of limited visibility in real flight scenarios.

[0222] Based on the pixel risk overlay intensity, the warning color matching the risk level, and the layer transparency coefficient, a risk visualization visual image is generated.

[0223] RGBA overlay layer blending operation, combining the risk warning color with the original view pixel color. (i.e., the base scene pixel colors output by the visual rendering channel) are blended and overlaid:

[0224] In the formula, Warning colors are assigned to different risk levels (e.g., transparent for safe, yellow for mild, orange for moderate, and red for severe). This is the transparency factor of the risk layer; the higher the hazard level, the lower the transparency and the more prominent the warning color.

[0225] The final output pixel color is composed of a weighted blend of two parts: one part is the original scene color, which is attenuated by the transparency coefficient and retained as the background; the other part is the risk warning color, whose overlay intensity is determined by the warning color itself, the transparency coefficient, and the pixel risk overlay intensity. When the risk level is low, When the risk level is low, the original view dominates, and the risk layer is almost invisible; as the risk level increases, The increased intensity significantly enhances the product of warning color and risk overlay, allowing the risk heat map to cover the original view with a striking color in the corresponding airspace area, enabling pilots to have an immediate visual perception of high-risk areas.

[0226] Extract the risk level corresponding to the center of mass position of the electric vertical take-off and landing aircraft. When the risk level is equal to or higher than the preset alarm level, send an alarm trigger signal.

[0227] The risk level corresponding to the center of mass position of an electric vertical takeoff and landing (eVTOL) aircraft refers to the risk level associated with extracting the eVTOL center of mass position. The comprehensive risk value at a given location is mapped using the aforementioned risk grading operator to obtain the corresponding risk level for that location. The logic for triggering the audible and visual alarm is as follows:

[0228] Level 2 (moderate alarm) is used as the critical level for alarm triggering. When the risk level of the location of the aircraft's center of gravity reaches the level of moderate alarm or severe danger, the alarm condition is met; when the aircraft is in a safe area or only at the level of mild warning, the alarm condition is not met.

[0229] When the alarm conditions are met (i.e., Level ≥ 2), an alarm signal is simultaneously pushed to the simulator control panel, and the alarm feedback mechanism is activated. For example, visual highlighting (enhancing the visual marking of high-risk areas where the aircraft is located in the visual scene) and voice prompts (broadcasting the risk type and level to the pilot through the acoustic simulation channel or an independent voice channel) allow the pilot to receive risk warning information through both visual and auditory channels in an immersive simulation environment.

[0230] In summary, the complete process of connecting physical field input, solving for multiple types of risks, integrating comprehensive risks, and hierarchical visualization is as follows: Solving operators for sub-risks; For risk classification operators; To visualize and overlay rendering operators.

[0231] exist Figure 2The illustrated embodiment demonstrates a closed-loop computational process for the interaction between landscape and meteorological physics. Steps 1-4 correspond to the complete construction process of a multi-physical attribute digital twin for the city, including multi-source data fusion, semantic geometric modeling, assignment of multi-dimensional physical parameters of land features, discretization of the 3D semantic attribute mesh, weighted attribute calculation, digital twin calibration, and output of a triplet. Step 5 corresponds to offline basis library simulation, including offline CFD multiphysics simulation to construct a feature basis library. Step 6 corresponds to online real-time physics field solving, with the online real-time solver outputting the coupled physics field. Step 7 corresponds to multimodal perception simulation, performing parallel computation of visual, flight dynamics, and acoustic channels. Step 8 corresponds to dynamic risk field quantification and visualization, including dynamic risk field calculation, hierarchical visualization, and alarms. Step 9 is the simulation state closed-loop feedback iteration step, which feeds back the state data in a closed loop to update the dynamic state of the mesh. This achieves bidirectional dynamic updates between landscape, meteorology, and aircraft, forming a closed loop in the overall process, reflecting the core dynamic coupling mechanism of this scheme.

[0232] Figure 3 The diagram illustrates a framework of a physics-based modeling system according to another aspect of this application, applied to an electric vertical takeoff and landing (EVTOL) flight simulator. The modeling system includes a digital twin construction module 10, a physics field generation module 20, and a coupled solution module 30. The digital twin construction module 10 acquires a set of landscape objects from a target city model, configures attribute parameters for each landscape object, and discretizes the simulation 3D space into multiple semantic attribute grid cells. Through spatial intersection, the attribute parameters of each landscape object are mapped to the corresponding grid cell, obtaining the effective physical attributes of each grid cell and providing each grid cell with its own physical attributes. The system configures dynamic state variables; the physical field generation module 20 is used to construct an offline feature base library based on the effective physical attributes of the semantic attribute grid cells, interpolate from the offline feature base library according to real-time macro-meteorological input to obtain a reference physical field, and generate a landscape disturbance field according to the dynamic state variables and effective physical attributes of each grid cell at the current time; the coupling solution module 30 is used to determine a global coupled physical field based on the landscape disturbance field and the reference physical field, and update the dynamic state variables of each grid cell in real time according to the global coupled physical field to form a physical feedback coupling between the landscape and the meteorology.

[0233] The physical feedback-based modeling system is a modular implementation of the aforementioned modeling method, logically constructed along the lines of digital landscape modeling, multi-physics layered solution, and bidirectional feedback closed loop. The digital twin construction module 10 corresponds to the landscape modeling steps in the method, used to complete the semantic segmentation, attribute configuration, and spatial discretization of the urban landscape. It transforms the traditional visual landscape into a semantic attribute mesh with effective physical attributes and capable of supporting dynamic states, providing the basic data carrier for subsequent full-process physical field calculations.

[0234] The physical field generation module 20 corresponds to the hierarchical solution steps in the method, and adopts an architecture of offline pre-computation and online lightweight solution. In the offline stage, a high-fidelity meteorological feature base library is built based on the semantic attribute grid. In the online stage, the benchmark physical field is obtained by interpolation combined with real-time macro meteorological input, and the landscape disturbance field is calculated based on the current dynamic state of the grid and the physical attributes. It is the core unit of physical field calculation.

[0235] The coupling feedback step in the corresponding method of the coupling solution module 30 is used to synthesize the global coupled physical field and update the dynamic state variables of each grid cell in reverse with the coupling result, forming a physical feedback closed loop of bidirectional interaction between landscape and weather. The output global coupled physical field can be synchronously supplied to downstream simulation modules such as visual rendering and flight dynamics calculation.

[0236] The proposed landscape-meteorology two-way physical feedback mechanism breaks the traditional decoupled simulation mode, transforming urban landscapes from passive visual carriers into actively interactive physical entities. It realistically recreates the natural interaction between terrain, buildings, and meteorology in the urban low-altitude environment, significantly improving simulation realism. Employing a layered decoupled architecture that balances accuracy and real-time performance, it performs high-computational simulations offline and lightweight online computations, perfectly adapting to the high real-time requirements of flight simulators without compromising physical fidelity, making it highly practical for engineering applications. Specifically adapted to eVTOL urban low-altitude simulation scenarios, it uses meter-level grid resolution to specifically simulate urban canyon micro-meteorology, filling the technical gap where traditional civil aviation simulators cannot adapt to eVTOL low-altitude flight simulation. Multi-physics, multi-modal integrated simulation integrates fluid, thermal, water, and acoustic multiple physics fields, linking visual, flight dynamics, and auditory perception channels to achieve full-dimensional immersive simulation. It possesses intelligent risk identification and visualization capabilities, capable of real-time analysis of flight risk areas and visual warnings, suitable not only for pilot simulation training but also supporting extended applications such as eVTOL route planning and flight safety assessment.

[0237] exist Figure 4 The illustrated embodiment presents the overall architecture of a physics-based aircraft simulation system, where 1 represents the multi-source geospatial data input module; 2 represents the urban multi-physical attribute digital twin construction module; 3 represents the offline high-fidelity feature base library construction module; and 4 represents the online hierarchical decoupling real-time coupling solver, outputting a global coupled physical field. Label 5 represents the multimodal perception simulation parallel channel, which includes three sub-channels: visual rendering, flight dynamics, and 3D acoustics; Label 6 represents the dynamic environment risk field generation and visualization module; Label 7 represents the external real-time input interface; and Label 8 represents the simulator's integrated output terminal. Solid arrows between modules indicate forward data transmission paths, while dashed lines indicate closed-loop feedback paths for aircraft status.

[0238] For the external real-time data access interface unit labeled 7, it is divided into two independent data stream channels: a macro-meteorological real-time input sub-interface and an eVTOL whole-aircraft flight status real-time input sub-interface. All parameters are pushed to the online real-time physical response solver at high frequency in a standardized time-series data stream for solving the base meteorological field. Aircraft disturbance fields such as rotor thrust / speed / position This enables dynamic coupling between the terrain, weather, and aircraft. The data stream format is millisecond-level time-series structured data with a unified timestamp. t Synchronously align with the three-dimensional mesh space coordinate reference.

[0239] The macro-meteorological real-time input sub-interface includes three types of parameters. The first type is basic meteorological observation parameters (real-time push from meteorological stations), the second type is short-term gust / turbulence additional parameters, and the third type is solar thermal boundary parameters (driving surface thermal buoyancy disturbances).

[0240] The eVTOL real-time flight status input interface includes three types of parameters. The first type is the airframe spatial motion state parameters, the second type is the independent dynamic parameters of each rotor (multi-rotor output channel by channel), and the third type is the global aerodynamic auxiliary parameters of the aircraft.

[0241] Interface data transmission and timing constraints follow these rules: Weather data packets and aircraft status data packets carry the same global timestamp. t Timing alignment is completed before being fed into the solver; refresh rate: 5Hz for meteorological input; 100Hz high-frequency push for eVTOL flight status, matching the flight dynamics step size; fault tolerance: when a single frame parameter is missing, the cached value of the previous moment is automatically used to avoid jumps in the calculation of the disturbance field; data output flow: all parameters are input into the online real-time physical response solver, and participate in the coupled solution of the three major modules of base wind field calculation, landscape disturbance operator, and rotor wake downwash disturbance operator.

[0242] The interface output coupling relationship can be summarized into the following parts: macroscopic meteorological interface parameters are used for solving... (Base wind, temperature, humidity, pressure, background turbulence); Building BIM static texture data (pre-module) and meteorological and solar radiation parameters are used together to solve the landscape disturbance field. eVTOL flight status and rotor dynamic parameters are used to solve the rotor disturbance field of the aircraft. Three games overlapping It serves as the input source for eVTOL six-degree-of-freedom aerodynamic solution and risk field quantization.

[0243] For terminal 8, which is the simulator's integrated output terminal, it receives output data from the preceding modules, including: the global coupled micro-meteorological field. The eVTOL six-degree-of-freedom dynamics solution results and dynamic risk quantification and classification data are output to the 3D visual rendering unit, flight control hardware feedback unit, simulation audio effect unit, and low-altitude risk audio-visual alarm unit via four standardized interfaces. All outputs carry a synchronized global timestamp. t The output refresh rate matches the simulation master step size of 100Hz.

[0244] The four output data streams are independent and synchronized in time, achieving integrated output of visuals, control forces, sound effects, and safety alarms. The first stream is a 3D city scene rendering output interface, including global flow field rendering parameters (for wind field cloud maps and airflow streamline visualization), risk layer overlay rendering parameters, and eVTOL aircraft attitude and motion output parameters. The second stream is a flight control hardware control force feedback output interface, which can connect to simulator joysticks, throttle levers, and other hardware, outputting servo drive quantities at a refresh rate of 100Hz. The third stream is a simulation environment sound effect output interface, with three categories of audio control parameters: environmental wind noise, rotor noise, and alarm sound effects, outputting to a sound card / audio amplifier. The fourth stream is a low-altitude flight graded audio-visual risk alarm output interface, corresponding to alarm logic, outputting to screen indicator lights, voice broadcast module, and interface text pop-ups. Parameters include graded control flag quantities and visual alarm auxiliary parameters.

[0245] The interface timing synchronization and output constraints follow these rules: A global synchronization mechanism requires all four outputs to share the same simulation timestamp 't', ensuring complete synchronization of the visual display, control feedback, audio effects, and alarms. A tiered refresh rate rule requires flight control force feedback and aircraft attitude output to use a 100Hz frequency, corresponding to the main dynamic step size; visual rendering and audio parameters to use a 60Hz frequency; and risk alarm status updates to use a 30Hz frequency. A fault-tolerant output rule requires that when the solver calculates an anomaly, it retains the output value of the previous frame to avoid sudden changes in the display and control impacts. The data flow to the front-end link executes sequentially from coupled meteorological field to aircraft aerodynamics calculation, then to risk quantification and grading, and finally enters the integrated output terminal to complete the synchronized distribution of the four interfaces.

[0246] The overall coupling logic of the interface can be summarized into the following four parts: The visual interface can output the global weather field, risk layer, and aircraft attitude to complete the low-altitude visualization simulation of digital cities; the flight control force feedback interface can output the additional control loads brought by turbulence and downwash to achieve a realistic flight control feel; the sound effect interface can dynamically adjust the environment and rotor noise based on wind speed, rotor speed, and turbulence intensity; and the risk alarm interface can output the risk level, risk type, and danger distance to provide multi-dimensional safety warnings with sound and light, supporting pilot low-altitude simulation training and flight route safety assessment.

[0247] The proposed solution adopts a hierarchical architecture of offline high-fidelity simulation + online lightweight solution. Complex multiphysics field calculations are performed offline, while online calculations are only performed on lightweight operations such as interpolation and local perturbation. This approach retains the high physical accuracy of CFD and microclimate models while meeting the millisecond-level real-time operation requirements of flight simulators, thus resolving the contradiction between fidelity and real-time performance.

[0248] The traditional static visual landscape is upgraded into an interactive entity with full-dimensional physical attributes, realizing two-way physical feedback between meteorology and landscape, and between landscape and meteorology. Meteorology changes the building temperature and surface water content; the building form and materials, in turn, change the local wind field, thermal circulation and turbulence, achieving deep coupling of the physical layer rather than simple visual layer superposition, thus solving the problem of the separation between visual and physical scenes.

[0249] Using a meter-level three-dimensional semantic attribute mesh, a refined physical field solution is performed for gaps between urban buildings and street canyons, accurately reproducing the micro-meteorological conditions faced by eVTOL low-altitude flight, such as local eddies, building wind shear, and thermal airflow. This adapts to the eVTOL operational scale and solves the problem of simulation scale mismatch with eVTOL.

[0250] By integrating multiple parameters from aerodynamics, thermodynamics, hydrology, and acoustics, it achieves coordinated simulation of wind, temperature, humidity, precipitation, and sound environment, comprehensively recreating the complex urban low-altitude environment and filling the gaps in multi-physics field analysis.

[0251] The system adds dynamic flight risk perception capabilities, which automatically identify risks such as turbulence, wind shear, thermal disturbances, and sensor failure zones based on real-time physical fields, and provides safety warning capabilities for eVTOL simulation training and flight simulation through visualization.

[0252] The proposed solution possesses excellent versatility, flexibly adaptable to eVTOL simulations in different urban areas and for different aircraft models. Its physical parameter library supports rapid expansion and iteration, allowing for continuous optimization of simulation accuracy based on measured data, and providing room for long-term technological iteration. It has a wide range of applications, including eVTOL pilot simulation training, urban air traffic route simulation, aircraft aerodynamic performance testing, and urban low-altitude environment assessment. Leveraging a highly realistic simulation environment, it reduces the frequency of on-site test flights, thereby lowering the safety risks and economic costs of eVTOL R&D, test flights, and personnel training. It also provides a novel modeling framework for the field of urban air traffic simulation.

[0253] This application also provides a computer-readable medium having computer instructions stored thereon, which can be executed by a processor to implement a physical feedback-based modeling method as described above.

[0254] When a physical feedback-based modeling method is implemented as a computer program, it can also be stored as an article of art in a computer-readable storage medium. For example, computer-readable storage media can include, but are not limited to, magnetic storage devices (e.g., hard disks, floppy disks, magnetic stripes), optical discs (e.g., compact discs (CDs), digital multifunction discs (DVDs)), smart cards, and flash memory devices (e.g., electrically erasable programmable read-only memory (EPROM), cards, sticks, key drives). Furthermore, the various storage media described herein can represent one or more devices and / or other machine-readable media used for storing information. The term "machine-readable medium" can include, but is not limited to, wireless channels and various other media (and / or storage media) capable of storing, containing, and / or carrying code and / or instructions and / or data.

[0255] It should be understood that the embodiments described above are merely illustrative. The embodiments described herein may be implemented in hardware, software, firmware, middleware, microcode, or any combination thereof. For hardware implementation, the processor may be implemented within one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, and / or other electronic units designed to perform the functions described herein, or combinations thereof.

[0256] Some aspects of this application can be executed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The aforementioned hardware or software may be referred to as a "data block," "module," "engine," "unit," "component," or "system." The processor may be one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DAPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, or combinations thereof. Furthermore, aspects of this application may manifest as computer products residing in one or more computer-readable media, including computer-readable program code. For example, computer-readable media may include, but are not limited to, magnetic storage devices (e.g., hard disks, floppy disks, magnetic tapes, etc.), optical discs (e.g., compressed CDs, digital multifunction DVDs, etc.), smart cards, and flash memory devices (e.g., cards, sticks, key drives, etc.).

[0257] A computer-readable medium may contain a propagated data signal containing computer program code, for example, on baseband or as part of a carrier wave. This propagated signal may take various forms, including electromagnetic, optical, and so on, or suitable combinations thereof. A computer-readable medium can be any computer-readable medium other than a computer-readable storage medium, which can be connected to an instruction execution system, apparatus, or device to enable communication, propagation, or transmission of a program for use. The program code located on the computer-readable medium can be propagated through any suitable medium, including radio, cable, fiber optic cable, radio frequency signals, or similar media, or any combination of the above media.

[0258] The basic concepts have been described above. Obviously, for those skilled in the art, the above disclosure is merely illustrative and does not constitute a limitation of this application. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements, and corrections to this application. Such modifications, improvements, and corrections are suggested in this application, and therefore remain within the spirit and scope of the exemplary embodiments of this application.

[0259] Furthermore, this application uses specific terms to describe embodiments of the application. For example, "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic related to at least one embodiment of the application. Therefore, it should be emphasized and noted that "an embodiment," "one embodiment," or "an alternative embodiment" mentioned twice or more in different locations in this specification do not necessarily refer to the same embodiment. In addition, certain features, structures, or characteristics in one or more embodiments of the application can be appropriately combined.

[0260] In some embodiments, numbers describing the quantity of components and attributes are used. It should be understood that such numbers used in the description of embodiments are modified in some examples with the terms "approximately," "approximately," or "generally." Unless otherwise stated, "approximately," "approximately," or "generally" indicates that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may be changed depending on the characteristics required by individual embodiments. In some embodiments, numerical parameters should take into account specified significant digits and employ a general method of digit reservation. Although the numerical ranges and parameters used to confirm their breadth of scope in some embodiments of this application are approximate values, in specific embodiments, such values ​​are set as precisely as feasible.

Claims

1. A physical feedback-based modeling method applied to a flight simulator for electric vertical takeoff and landing aircraft, characterized in that, The modeling method includes: Obtain the collection of landscape objects of the target city model and configure attribute parameters for each landscape object; The simulated 3D space is discretized into multiple semantic attribute grid cells. By spatial intersection, the attribute parameters of each landscape object are mapped to the corresponding grid cell, the effective physical attributes of each grid cell are obtained, and dynamic state variables are configured for each grid cell. An offline feature base library is constructed based on the effective physical properties of the semantic attribute grid cells. A reference physical field is obtained by interpolation from the offline feature base library according to the real-time macro meteorological input. A landscape disturbance field is generated according to the dynamic state variables of each grid cell at the current time and the effective physical properties. The global coupled physical field is determined based on the landscape disturbance field and the reference physical field, and the dynamic state variables of each grid cell are updated according to the global coupled physical field to form a physical feedback coupling between the landscape and the weather.

2. The modeling method according to claim 1, characterized in that, The acquisition of the landscape object set of the target city model includes: Acquire and fuse multi-source geospatial data of the target city to obtain fused data; The fused data is semantically segmented to output a set of structured landscape objects and generate a multi-level-of-detail model for each structured landscape object, wherein each structured landscape object is associated with triangular mesh geometry and semantic labels.

3. The modeling method according to claim 1, characterized in that, The modeling method also includes: The real-time parameters of the electric vertical take-off and landing aircraft are obtained, and the real-time parameters are input into a preset disturbance calculation model. The disturbance calculation model outputs flow field disturbance parameters, and the aircraft disturbance field is constructed based on the flow field disturbance parameters. The perturbation field of the aircraft is superimposed on the local position of the aircraft in the global coupled physical field, and used as the input for the calculation of the aerodynamic load of the aircraft.

4. The modeling method according to claim 1, characterized in that, The process of interpolating from the offline feature base library based on real-time macro-meteorological input to obtain the reference physical field includes: Obtain the current macro-meteorological input parameters, which include at least wind direction, wind speed, solar altitude angle, atmospheric stability, ambient temperature, and ambient relative humidity; Select multiple reference scene points adjacent to the macro-meteorological input parameters from the offline feature base library; Multidimensional linear interpolation is performed on the global basis physics field corresponding to the multiple adjacent reference scene points to obtain the reference physics field.

5. The modeling method according to claim 1, characterized in that, The construction of the offline feature base library based on the effective physical attributes of the semantic attribute grid cells includes: Construct a scene parameter space that covers a set range of meteorological parameters, and take discrete reference scene points within the scene parameter space; For each benchmark scene point, the effective physical properties of the semantic attribute grid unit are used as the surface boundary conditions. The multiphysics coupling control equations are solved simultaneously to obtain the global base physical field corresponding to the benchmark scene point. An offline feature base library is constructed based on the global base physical field corresponding to all benchmark scene points.

6. The modeling method according to claim 1, characterized in that, The attribute parameters include aerodynamic attribute parameters, and the landscape disturbance field includes landscape-induced horizontal disturbance wind speed components, landscape-induced static pressure disturbance components, and landscape-induced additional turbulent kinetic energy components. Generating the landscape disturbance field based on the dynamic state variables of each grid cell at the current moment and the effective physical attributes includes: Obtain the dynamic state variables and effective physical properties of each grid cell at the current moment; The landscape-induced horizontal disturbance wind speed component is determined based on the preset mapping relationship between the geometric occlusion effect of landscape objects at each grid unit and the aerodynamic property parameters. The static pressure disturbance component induced by the landscape is determined based on air density, base flow wind speed, and synthesized wind speed after disturbance. The additional turbulent kinetic energy component induced by the landscape is determined based on the horizontal disturbance wind speed component induced by the landscape and the building turbulence generation coefficient associated with the urban morphological feature zone to which the grid cell belongs. The value of the building turbulence generation coefficient is determined by the urban morphological feature zone.

7. The modeling method according to claim 6, characterized in that, The landscape disturbance field also includes a landscape-induced vertical thermal buoyancy disturbance wind speed component, a landscape-induced temperature disturbance component, and a landscape-induced humidity disturbance component. The generation of the landscape disturbance field further includes: Based on the local temperature difference between the surface temperature of each grid cell and the ambient reference temperature, the vertical thermal buoyancy disturbance wind speed component induced by the landscape is determined by combining the gravitational acceleration, thermal buoyancy correction coefficient and near-ground reference height, and the vertical airflow direction is determined based on the sign of the local temperature difference. An exponential decay factor is constructed based on the ground height of the grid unit and the thermal diffusion attenuation height, and the landscape-induced temperature disturbance component is determined based on the local temperature difference and the exponential decay factor. Based on the instantaneous surface water content and building shadow shading coefficient of each grid unit, a weighted difference calculation is performed using the corresponding first empirical correction coefficient and second empirical correction coefficient to determine the landscape-induced humidity disturbance component.

8. The modeling method according to claim 3, characterized in that, The process of acquiring real-time parameters of the electric vertical takeoff and landing (EVTOL) aircraft, inputting these real-time parameters into a preset disturbance calculation model, having the disturbance calculation model output flow field disturbance parameters, and constructing the aircraft disturbance field based on these flow field disturbance parameters includes: The real-time parameters of the electric vertical takeoff and landing aircraft are obtained, wherein the real-time parameters include real-time spatial position, rotor induced velocity field and rotor speed vector; The real-time parameters are input into a preset rotor downwash disturbance calculation operator. The rotor downwash disturbance calculation operator outputs the three-dimensional disturbance wind speed component induced by the aircraft, the static pressure disturbance component induced by the aircraft, and the turbulent kinetic energy component induced by the aircraft, which together constitute the aircraft disturbance field. The rotor downwash disturbance calculation operator is a fast calculation operator pre-established based on the sheet theory and the finite-length vortex wake model.

9. The modeling method according to claim 1, characterized in that, The attribute parameters include acoustic attribute parameters; the flight simulator includes a visual rendering channel, a flight dynamics calculation channel, and an acoustic simulation channel; the modeling method further includes: The same global coupled physical field is synchronously output to the visual rendering channel, flight dynamics calculation channel and acoustic simulation channel; The rendering parameters of the visual scene are updated by the visual rendering channel based on the global coupled physical field; The flight dynamics solution channel extracts the global coupled physical field at the location of the electric vertical takeoff and landing aircraft and uses it as an aerodynamic load input term into the six-degree-of-freedom dynamic equations of the aircraft. The acoustic simulation channel determines the spatial environmental acoustic parameters of the listening point location based on the global coupled physical field and the acoustic attribute parameters.

10. The modeling method according to claim 1, characterized in that, The modeling method also includes: Using the global coupled physical field as the only physical field input, and based on the corresponding physical quantities of the global coupled physical field, risk subfields are calculated and determined according to a preset flight safety threshold parameter set. The risk subfields include vertical airflow risk subfields, turbulent eddy current risk subfields, thermal airflow disturbance risk subfields, and sensor performance degradation risk subfields. The risk subfields are weighted and summed to obtain a comprehensive risk value, which is then mapped to different risk levels according to a preset grading threshold.

11. The modeling method according to claim 10, characterized in that, The modeling method also includes: Based on the simulated viewpoint and line of sight, the comprehensive risk value is integrated by line of sight and attenuated to obtain the pixel risk superposition intensity. Based on the pixel risk overlay intensity, the warning color matching the risk level, and the layer transparency coefficient, a risk visualization visual image is generated by overlaying. Extract the risk level corresponding to the center of mass position of the electric vertical take-off and landing aircraft. When the risk level is equal to or higher than the preset alarm level, send an alarm trigger signal.

12. A physical feedback-based modeling system applied to an electric vertical takeoff and landing (EVTOL) aircraft flight simulator, characterized in that, The modeling system includes a digital twin construction module, a physical field generation module, and a coupled solution module; The digital twin construction module is used to obtain a set of landscape objects of the target city model, configure attribute parameters for each landscape object, and discretize the simulation three-dimensional space into multiple semantic attribute grid units. By spatial intersection, the attribute parameters of each landscape object are mapped to the corresponding grid unit to obtain the effective physical attributes of each grid unit and configure dynamic state variables for each grid unit. The physical field generation module is used to construct an offline feature base library based on the effective physical attributes of the semantic attribute grid cells, interpolate the reference physical field from the offline feature base library according to the real-time macro meteorological input, and generate a landscape disturbance field according to the dynamic state variables of each grid cell at the current time and the effective physical attributes. The coupled solution module is used to determine the global coupled physical field based on the landscape disturbance field and the reference physical field, and to update the dynamic state variables of each grid cell in real time according to the global coupled physical field, so as to form a physical feedback coupling between the landscape and the weather.

13. A computer-readable medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 11.

Images