Road and air integrated three-dimensional traffic safety control twin system and method

Abstract

The application discloses a road-air integrated three-dimensional traffic safety control twin system and method, belonging to the field of traffic safety control and digital twin technology, comprising a digital twin modeling layer for constructing a three-dimensional integrated twin model, realizing real-time mapping and trajectory prediction of road-air traffic state; a road-air collaborative perception layer for generating state data sets in a unified format; an intelligent anti-collision control layer for risk prediction and collision avoidance strategy generation; a full-link anti-interference layer for ensuring safe and reliable transmission of the collision avoidance strategy; an execution feedback layer for executing the collision avoidance strategy and collecting feedback data to optimize the twin model and anti-interference strategy; the layers are sequentially connected in communication to form a closed-loop management and control architecture. The application realizes precise management and control, safe anti-collision and reliable communication of road-air traffic by innovative digital twin modeling architecture, dynamic anti-collision mechanism and full-link anti-interference technology, and integrates self-destruction protection and black box intelligent learning function.

Description

Technical Field

[0001] This invention relates to the field of traffic safety control and digital twin technology, and in particular to an integrated road-air three-dimensional traffic safety control twin system and method. Background Technology

[0002] With the rapid development of the low-altitude economy and intelligent transportation, the operating scenarios of low-altitude aircraft and ground vehicles highly overlap, posing multiple challenges to the safety management of three-dimensional transportation.

[0003] In existing technologies, the application of digital twin technology in the transportation sector is mostly limited to single ground or low-altitude scenarios, lacking integrated modeling covering "ground road network - low-altitude airspace - infrastructure," and thus unable to achieve real-time mapping and collaborative analysis of road-air traffic conditions. Furthermore, road-air traffic conflict prevention relies on single-height hierarchical or static interval rules, which are insufficient to cope with dynamic conflicts in high-density scenarios, and lack intelligent collision avoidance mechanisms incorporating multi-dimensional parameters. In terms of communication, low-altitude communication frequency bands are susceptible to interference, existing encryption technologies lack adaptability, and there is a lack of end-to-end anti-interference design from signal transmission and propagation to reception, as well as emergency protection and adaptive optimization mechanisms to address encryption cracking, leading to delays, failures, or intrusion and tampering in traffic control command transmission. Simultaneously, the identification, status awareness, and collaborative scheduling of heterogeneous road-air traffic lack a unified system, failing to support the safe and efficient operation of large-scale three-dimensional transportation.

[0004] In summary, existing technologies are insufficient to simultaneously address the three core challenges of integrated digital twin modeling, dynamic collision avoidance, and end-to-end anti-interference, necessitating the development of an innovative integrated road-air safety control system. Summary of the Invention

[0005] To address the aforementioned issues, this invention proposes a road-air integrated three-dimensional traffic safety control twin system and method. By innovating the digital twin modeling architecture, dynamic anti-collision mechanism, and full-link anti-interference technology, and integrating self-destruct protection and black box intelligent learning functions, it achieves precise control, safe collision avoidance, and reliable communication of road and air traffic.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A first aspect of the present invention provides a road-air integrated three-dimensional traffic safety control twin system, comprising:

[0008] The digital twin modeling layer is used to construct a three-dimensional integrated twin model of the ground, low-altitude, and facilities to achieve real-time mapping and trajectory prediction of road and air traffic conditions.

[0009] The road-air collaborative perception layer is used to collect and fuse multi-source traffic and environmental data to generate a unified format state dataset.

[0010] The intelligent collision avoidance control layer, based on the twin model and state dataset, predicts risks through a spatiotemporal probability conflict model, and generates collision avoidance strategies by combining a three-dimensional collision avoidance mechanism and a multi-agent self-coordination method.

[0011] The end-to-end anti-interference layer is used to ensure the safe and reliable transmission of collision avoidance strategies through dedicated frequency band planning, encrypted transmission, self-destruction procedures and black box intelligent learning algorithms.

[0012] The execution feedback layer is used to execute collision avoidance strategies and collect feedback data to optimize the twin model and anti-interference strategies.

[0013] The layers are connected in sequence to form a closed-loop management and control architecture that includes modeling, perception, decision-making, execution, and feedback.

[0014] As a further implementation, the digital twin modeling layer includes:

[0015] The geometric modeling module uses BIM, GIS and three-dimensional orientation and domain grid integration technology to construct an integrated three-dimensional model;

[0016] The dynamic mapping module is used to achieve real-time data synchronization between the physical space and the twin model;

[0017] The predictive modeling module performs trajectory prediction based on bidirectional long short-term memory networks, particle swarm optimization extreme learning machines, and heterogeneous traffic subject modeling methods.

[0018] The identity coding module assigns a unique three-dimensional identity code to each participant in road and air traffic.

[0019] As a further implementation method, the intelligent collision avoidance control layer includes:

[0020] The conflict prediction module is used to construct a four-dimensional conflict assessment matrix and calculate the collision probability.

[0021] The dynamic control module outputs differentiated collision avoidance strategies based on risk levels.

[0022] A self-coordination mechanism is used for autonomous negotiation of collision avoidance paths among multiple agents;

[0023] The global optimization module optimizes trajectories based on multiple objectives, including time cost, collision probability, and energy consumption.

[0024] As a further implementation method, the end-to-end anti-interference layer includes:

[0025] The frequency band planning module is used to allocate and authorize dedicated air and road communication frequency bands;

[0026] The encrypted transmission module uses the SM9-Plus algorithm and blockchain consensus mechanism for command encryption and authentication.

[0027] The core protection module embeds a self-destruct program and a black-box intelligent learning algorithm for threat detection and security data capture.

[0028] The anti-interference control module implements anti-interference measures throughout the entire signal transmission, propagation, and reception chain.

[0029] The dynamic no-fly zone module automatically delineates temporary no-fly zones based on environmental data.

[0030] As a further implementation method, the self-destruct program is triggered by conditions including the outer encryption algorithm being cracked, malicious intrusion code being detected, or the interference strength exceeding a preset threshold; the black box intelligent learning algorithm is used to capture interference features and intrusion code, and feed them back to the execution feedback layer for strategy optimization.

[0031] A second aspect of the present invention provides a road-air integrated three-dimensional traffic safety control method, based on the road-air integrated three-dimensional traffic safety control twin system described in the first aspect of the present invention, comprising the following steps:

[0032] S1: Construct an integrated digital twin model of the ground, low-altitude, and facilities, and assign unique identification codes to traffic participants;

[0033] S2: Collect road and air traffic and environmental data through multi-source sensors, and perform fusion processing to generate a standardized dataset;

[0034] S3: Based on the twin model and dataset, it predicts collision risk through a spatiotemporal probability conflict model, and generates collision avoidance strategies by combining a three-dimensional collision avoidance mechanism and self-coordinating decision-making.

[0035] S4: Employs encryption algorithms, end-to-end anti-interference technology, and self-destruct protection mechanisms to transmit collision avoidance strategies;

[0036] S5: Execute collision avoidance strategies, collect execution results and safety event data, and provide feedback to optimize the twin model and anti-interference strategies.

[0037] As a further implementation, the spatiotemporal probabilistic conflict model specifically involves: constructing a four-dimensional conflict evaluation matrix (x, y, z, t), based on the collision probability function. Calculate future time periods The probability of conflict within, where, and Vehicles and aircraft at different times The predicted location probability density, This is the conflict intensity function.

[0038] As a further implementation, step S4 further includes:

[0039] Plan and authorize dedicated communication frequency bands;

[0040] Interference prediction and spectral hole detection were performed using a Gaussian process regression model.

[0041] Adaptive power adjustment, edge relay nodes, and beamforming technology are employed to enhance signal transmission.

[0042] A self-destruct mechanism is triggered when a threat is detected, and security event data is captured using a black-box algorithm.

[0043] As a further implementation method, the trajectory prediction parameters and anti-interference strategies are optimized through artificial intelligence algorithms based on feedback data, so as to achieve adaptive iterative improvement of the system.

[0044] As a further implementation method, it also includes dynamically delineating no-fly zones based on real-time meteorological and electromagnetic environment data, and pushing the data to all traffic participants through a twin model.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0046] The integrated road-air traffic safety control twin system and method of the present invention constructs a full-domain twin model covering "ground-low altitude-facilities" through the fusion technology of BIM+GIS+three-dimensional orientation domain division grid, realizing integrated digital twin modeling. Combined with advanced prediction algorithms, it achieves millisecond-level real-time mapping and high-precision trajectory prediction, effectively identifying and avoiding traffic conflicts in advance, and improving the overall coordination capability of three-dimensional traffic.

[0047] The road-air integrated three-dimensional traffic safety control twin system and method of the present invention constructs a three-dimensional intelligent collision avoidance system, proposes a "layer-grid-temporal" three-dimensional collision avoidance mechanism, and combines a spatiotemporal probability conflict model with a multi-agent self-coordination method to achieve autonomous conflict resolution in high-density dynamic scenarios, support the safe operation of large-scale road-air traffic, and ensure the efficient passage of key tasks through priority scheduling.

[0048] The integrated road-air traffic safety control twin system and method of the present invention strengthens the anti-interference and security protection of the whole link, designs a transmission scheme that combines SM9-Plus encryption algorithm and blockchain consensus mechanism, and pioneers a dual-layer core protection of "self-destruct program + black box intelligent learning", which realizes anti-interference and anti-intrusion of the whole link from signal transmission, propagation to reception, and greatly improves communication reliability, security and adaptive optimization capabilities.

[0049] The road-air integrated three-dimensional traffic safety control twin system and method of the present invention unifies the identity recognition and collaborative scheduling system. It realizes unified identification and precise management of heterogeneous road-air traffic participants through a three-dimensional identity coding system. Combined with dynamic safety interval settings and collaborative scheduling rules, it significantly improves the accuracy of identity recognition and scheduling efficiency, and provides basic support for the efficient collaboration of three-dimensional traffic. Attached Figure Description

[0050] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0051] Figure 1 This is an architectural diagram of the integrated road-air traffic safety control twin system of the present invention;

[0052] Figure 2 This is a flowchart of the intelligent anti-collision control method of the present invention;

[0053] Figure 3 This is a flowchart of the end-to-end anti-interference control method of the present invention. Detailed Implementation

[0054] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0055] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0056] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0057] Example 1

[0058] like Figure 1 As shown, this embodiment provides a road-air integrated three-dimensional traffic safety control twin system, including:

[0059] The digital twin modeling layer is used to construct a three-dimensional integrated twin model of the ground, low-altitude, and facilities to achieve real-time mapping and trajectory prediction of road and air traffic conditions.

[0060] The road-air collaborative perception layer is used to collect and fuse multi-source traffic and environmental data to generate a unified format state dataset.

[0061] The intelligent collision avoidance control layer, based on the twin model and state dataset, predicts risks through a spatiotemporal probability conflict model, and generates collision avoidance strategies by combining a three-dimensional collision avoidance mechanism and a multi-agent self-coordination method.

[0062] The end-to-end anti-interference layer is used to ensure the safe and reliable transmission of collision avoidance strategies through dedicated frequency band planning, encrypted transmission, self-destruction procedures and black box intelligent learning algorithms.

[0063] The execution feedback layer is used to execute collision avoidance strategies and collect feedback data to optimize the twin model and anti-interference strategies.

[0064] The layers are connected in sequence to form a closed-loop management and control architecture that includes modeling, perception, decision-making, execution, and feedback.

[0065] The integrated road-air traffic safety control twin system of the present invention is specifically as follows:

[0066] The system comprises a digital twin modeling layer, a road-air cooperative perception layer, an intelligent collision avoidance control layer, a full-link anti-interference layer, and an execution feedback layer. These layers are sequentially connected to form a closed-loop control system of "modeling-perception-decision-execution-feedback."

[0067] The digital twin modeling layer is used to construct a three-dimensional integrated twin model of "ground-low altitude-infrastructure" to achieve real-time mapping and prediction of road and air traffic conditions.

[0068] Geometric modeling module: Based on the fusion technology of BIM+GIS+3D orientation domain grid, it constructs a fine model of ground road network (1m accuracy), low-altitude airspace 3D grid (10m×10m×5m) and infrastructure (charging piles, take-off and landing points, etc.).

[0069] Among them, the three-dimensional orientation domain grid is a three-dimensional spatial division system specifically designed for urban integrated road-air transportation. Using the standard latitude and longitude grid as the absolute coordinate reference, it first determines the city's unique code and city center, then integrates the north, south, east, west, northeast, southeast, northwest, and southwest orientation logic to divide the horizontal space into eight fan-shaped main areas. Each area is further refined into a horizontal grid through "concentric rings + radial sectors." Vertically, the functional height layers are matched with the first letter of their English names: Logistics Layer is L, Manned Layer is M, and Emergency Layer is E, ultimately forming a three-dimensional grid unit of "horizontal positioning + vertical layering." For example, 110000-Northwest 03A7-M05, decoded as the fifth layer of the manned layer in the radial sector of the third ring A7 in the northwest direction of Beijing (110000). The core logic of integrating with the standard latitude and longitude grid is "the standard latitude and longitude grid as the underlying support, and the three-dimensional orientation domain grid as the application layer mapping."

[0070] The three-dimensional orientation domain grid adopts the standard latitude and longitude grid coordinate system. The division of the city origin, sector boundaries, and ring / sector divisions are all based on latitude and longitude coordinates, without additional coordinate system conversion costs. Each orientation domain horizontal grid corresponds to a unique standard latitude and longitude grid range. The height range of the vertical grid is consistent with the elevation datum, and can be directly connected to the elevation data layer of the GIS system. Compared with directly using the standard latitude and longitude grid, the three-dimensional orientation domain grid converts location from numerical values ​​into information-rich symbols with inherent initial semantics, greatly simplifying the initial judgment of spatial relationships, making management rules more intuitive and interpretable, significantly improving the readability, understandability, and operational efficiency of the system, reducing the cognitive load of human decision-makers, and facilitating human-machine collaboration.

[0071] Dynamic mapping module: Through edge computing and 5G-A communication, it achieves millisecond-level synchronization (update frequency ≤50ms) between physical space data and twin models, and integrates real-time position, speed, attitude and other status data of vehicles and aircraft;

[0072] Predictive Modeling Module: This module integrates bidirectional long short-term memory networks, particle swarm optimization, extreme learning machines, and heterogeneous traffic subject modeling methods to construct a road-air traffic trajectory prediction model. It introduces intrinsic orthogonal decomposition (POD) to reduce the dimensionality and extract features from traffic state data, establishing a traffic state snapshot matrix. For manually driven vehicles, an extended intelligent driver model (IDM) is used; for intelligent connected vehicles, a cooperative adaptive cruise model is employed; and for low-altitude aircraft, a height-dimensional motion equation is introduced. ,in This is the height adjustment coefficient. For the target height, To address random wind disturbance, differentiated trajectory prediction is achieved for different types of subjects, outputting high-precision trajectories 5-10 seconds in advance; among which, the height adjustment coefficient... The range of values ​​and the method for determining them are as follows:

[0073] Physical meaning: coefficient It is the proportional gain of the height control loop, and its physical unit is s. -1 It determines the aircraft's altitude from its current position. Approaching target altitude Its response speed and stability.

[0074] Values: , For the aircraft from its current altitude Approaching target altitude The time required, in normal mode This is the ratio of vertical height to current speed. In emergency obstacle avoidance mode, The requirements are short and derived from collision avoidance strategies. α is the state adjustment coefficient, with an empirical range of 0.0 to 5.0. Lower α values ​​(0.0-1) are used for passenger comfort modes or deceleration flight during smooth climb / descent; higher α values ​​(1.0-5.0) are used for emergency maneuver modes for agile obstacle avoidance.

[0075] random wind disturbance term The calculation model is as follows:

[0076] Mathematical model: The model is created as random wind speed disturbances in the vertical direction, acquired in real time from the environmental sensing unit. This is injected as an external input into the aircraft's altitude dynamics equations, allowing the trajectory to fully account for uncertainties caused by wind disturbances.

[0077] Identity coding module: Assigns a unique "three-dimensional identity code" to all road and air traffic participants, which includes information such as type identifier, performance parameters, and operating permissions, to achieve accurate identification of heterogeneous traffic.

[0078] The road-air collaborative perception layer is used to collect road-air traffic and environmental data from multiple dimensions to support management and decision-making.

[0079] Ground sensing unit: Deploys millimeter-wave radar, vehicle-to-everything (V2X) sensors, and geomagnetic detectors to collect data such as vehicle position, speed, and load;

[0080] Low-altitude sensing unit: integrates lidar, ADS-B remote positioning, and Remote ID close-range identification equipment to achieve all-area perception of the aircraft, while forcing the aircraft to be equipped with a high-precision positioning system to adapt to different distance monitoring needs;

[0081] The method by which the low-altitude sensing unit achieves global sensing is as follows:

[0082] First, use BeiDou to assign a unified timestamp to all sensing data and unify the coordinates of each device to the GIS geographic coordinate system to align the data in time and space.

[0083] Secondly, a joint probabilistic data association algorithm based on multi-sensor multi-target tracking is adopted, using the aircraft's three-dimensional identity code as an auxiliary association feature to achieve data fusion.

[0084] Finally, a state estimator based on capacitive Kalman filtering is established to generate a unified high-confidence motion state estimate for the aircraft, thereby achieving position tracking.

[0085] Environmental sensing unit: integrates meteorological satellite data, ground meteorological station data and electromagnetic environment monitoring equipment to collect parameters such as wind speed, precipitation, and electromagnetic interference intensity;

[0086] The method for the environmental sensing unit to fuse data from multiple devices is as follows:

[0087] Meteorological data fusion: Spatial interpolation and fusion of satellite data and ground station data are performed to generate a gridded meteorological field covering the entire area, and output parameters such as wind speed and visibility.

[0088] Electromagnetic data fusion: Construct a spectrum map from spectrum scanning data from multiple monitoring stations to generate a real-time electromagnetic interference heat map.

[0089] Joint early warning: When the fused data determines that "the wind speed is ≥15m / s and there is broadband interference of ≥-70dBm in the target airspace", a joint early warning of communication risks in severe environment is generated to provide a decision basis for the dynamic no-fly module.

[0090] Data fusion module: Employs heterogeneous sensor fusion technology to eliminate data redundancy and errors, generating a unified format of road and air traffic status dataset.

[0091] The intelligent collision avoidance control layer, based on twin models and perception data, enables accurate prediction and dynamic control of road-air conflicts.

[0092] Conflict prediction module: Introduces a "spatiotemporal probabilistic conflict model," combining the vehicle's and aircraft's positions (x, y, z) and velocities (v). x ,v y ,v z ), acceleration (a x ,a y ,a z Based on the heading angle (θ) and airspace grid occupancy status, a four-dimensional conflict assessment matrix is ​​constructed; based on the heterogeneous trajectory prediction results, a collision probability function is defined. (in , ρ represents the predicted position probability density for the vehicle and the aircraft, respectively, and ρ is the conflict intensity function. T To predict the length of time, t (At the current assessment time), calculate the conflict probability; where x, y, z, and t in the four-dimensional conflict assessment matrix represent three-dimensional geospatial coordinates and the time dimension.

[0093] Each element M(x,y,z,t) of this matrix stores the probability that a spatial unit (x,y,z) will be occupied by one or more traffic participants at a future time t, or aggregated information about its state vector. By traversing and computing this matrix in parallel, the risk of overlap in the same spatiotemporal units can be detected efficiently.

[0094] The specific steps and methods for constructing the four-dimensional conflict assessment matrix are as follows:

[0095] Step 1: Input data from multiple sources.

[0096] Real-time status data: derived from a unified dataset fused by the road-air cooperative sensing layer, including the real-time position (x, y, z) and velocity (v) of all traffic participants. x ,v y ,v z ), acceleration and heading angle.

[0097] Predicted trajectory data: This data originates from the heterogeneous trajectory prediction results output by the predictive modeling module. For vehicles, the predicted trajectory data is the future trajectory output by the extended IDM model; for aircraft, the predicted trajectory data combines the model output of the height-dimensional motion equations.

[0098] Identity and performance data: derived from the three-dimensional identity code assigned by the identity coding module, used to uniquely identify participants and associate them with information such as their physical size and performance boundaries.

[0099] Step 2: Spatiotemporal state vector generation.

[0100] For each traffic participant i, its state is discretized into a spatiotemporal state matrix. At each future discrete time point t... k Based on its predicted trajectory, the center location is determined, and combined with its physical envelope size (obtained from the identity code), the set G of all spatial grid cells covered by the participant's occupied volume is calculated. i (t k Generate the spatiotemporal state vector S of the participant at that moment. i (t k The vector content must include at least position, velocity, acceleration, participant ID, and type.

[0101] Step 3: Conflict assessment matrix filling. Place participant i in t... k The spatial grid set G covered at any given time i (t k All grid coordinates (x) in ) g ,y g ,z g ) as a spatial index, t k As a time index, initialize a four-dimensional zero matrix [M], in matrix M(x g ,y g ,z g ,t k In this process, participant identification information is written or accumulated. A typical implementation involves recording a list of participant IDs within the matrix cell. If the list length of a grid cell at a given moment is greater than 1, it indicates a potential risk of physical overlap within that spatiotemporal cell.

[0102] Step 4: Conflict detection and probability calculation.

[0103] For each time slice t k Scan the entire spatial grid to find grid cells with a list length of at least 2 for all participant IDs. Each such cell identifies a potential point of conflict.

[0104] For each conflict cell, the collision intensity function ρ is called to calculate the collision probability based on the state vectors of the two or more conflicting parties within that cell. The conflict intensity is obtained by integrating along the time dimension, which is a weighted sum of the conflict intensity ρ of all conflicting units at each time point.

[0105] The conflict intensity function ρ is used to quantify the instantaneous risk intensity of a collision between a vehicle and an aircraft at predicted positions Pv(τ) and Pa(τ) at a specific future time τ. The design is based on the core principle that "risk is inversely proportional to distance and directly proportional to radial approach speed".

[0106] Function expression: ;

[0107] in, The relative velocity between the two agents is expressed in m / s. The distance between the two agents is in meters (m). The minimum distance constant is set to 0.1m to prevent division by zero. C is the comprehensive risk coefficient, measured in seconds (s). Its baseline value C0 is determined by the system's preset standard near-field critical encounter scenario, calculated using the formula C0 = R. high d min / v max Among them, R high The high-risk baseline intensity is 1.0, d min With v max The value of C is dynamically calculated based on the physical dimensions and performance parameters in the three-dimensional identity codes of both conflicting parties. The final value of C is based on C0 and can be fine-tuned according to the weights of the participant types.

[0108] Dynamic control module: Innovative "layer-grid-time" three-dimensional collision avoidance mechanism. Based on vertical layering (0-120m logistics drone layer, 120-300m manned eVTOL layer, and above 300m emergency layer), collision avoidance is achieved through time window allocation, dynamic safety interval (set according to performance differences) and path reconstruction.

[0109] Vertical stratification is based on the relevant management regulations of the Civil Aviation Administration of China, which stipulate that the combined flight operation altitude of medium, light, and small unmanned aerial vehicles shall not exceed 300 meters above sea level.

[0110] 0-120 meters: This height level is close to ground infrastructure, suitable for high-frequency, short-distance, point-to-point logistics distribution, and physically isolated from high-level airspace, simplifying control complexity.

[0111] 120-300 meters: This altitude can effectively avoid most ground obstacles and low-altitude turbulence, providing sufficient emergency response time and space margin for manned flight.

[0112] Above 300 meters: This level can be designated as an emergency level to avoid interfering with the regular commercial traffic flow below, and its height advantage can be used to achieve wide-area monitoring.

[0113] Dynamic safety separation model: The safety separation includes the maximum braking distance, the total system response distance, and the safety buffer margin (dynamically adjusted according to airspace congestion and weather conditions).

[0114] Self-coordination mechanism: Construct a multi-agent collaborative decision-making model to support road and air traffic participants in autonomously negotiating obstacle avoidance strategies, with the priority rule being: emergency rescue > medical transportation > logistics transportation > ordinary traffic;

[0115] The multi-agent collaborative decision-making model is a three-layer distributed decision-making system, specifically as follows:

[0116] Local agent layer: Each agent (aircraft and vehicle, etc.) uses its own perception and twin model data and a built-in policy generator to quickly generate local collision avoidance strategy alternatives, such as accelerating through, slowing down to give way, and climbing / descending.

[0117] Collaborative Decision-Making Group Layer: When a potential conflict is detected, the relevant agents automatically form a temporary collaborative decision-making group via V2X communication. Within the group, they negotiate by exchanging alternative solutions, with the goal of reaching a strategy acceptable to all members.

[0118] Twin verification and filing layer: The agreed collaborative strategy will be sent to the digital twin system for rapid simulation verification. After confirming that there are no secondary conflicts, the strategy will be recorded in the blockchain evidence storage module and distributed for execution.

[0119] The specific implementation method of the multi-agent cooperative decision-making model is as follows:

[0120] 1. Communication and networking mechanisms:

[0121] When a conflict risk is detected, the intelligent agent uses the V2X communication capability provided by the road-air cooperative perception layer to automatically initiate a request to other intelligent agents involved in the conflict to form a temporary cooperative decision-making group based on the three-dimensional identity code and location information.

[0122] 2. Negotiation-based decision-making algorithm:

[0123] Based on a distributed negotiation algorithm using priority rules and benefit-based voting, each agent broadcasts its locally generated collision avoidance strategy candidates to the decision-making group. After receiving other options, each agent calculates the cost impact on itself if that option is implemented. Each agent votes on all options. The voting weight is positively correlated with its priority in the current conflict. The system prioritizes the option with the highest total weighted votes and where no single agent's cost exceeds its tolerance threshold.

[0124] The specific implementation steps are as follows:

[0125] 1. Networking: The parties involved in the conflict, A and B, automatically form a network via V2X.

[0126] 2. Generation and Broadcasting: A generates a strategy based on its IDM model: "Accelerate to 25m / s to pass ahead" (cost: energy consumption +5%); B generates a strategy based on its motion model: "Climb to 125m altitude and wait" (cost: time delay +8 seconds). Broadcasting strategies and costs for both parties.

[0127] 3. Evaluation and Voting: Option A evaluates Option B, with a cost of 0; Option B evaluates Option A, with a cost of time delay + 12 seconds. According to the priority rule, passenger transport > logistics, so Option B has a higher voting weight. Calculations show that Option B receives a higher weighted number of votes, while Option A's cost is within tolerable limits.

[0128] 4. Implementation and Verification: The decision-making team adopts B's "climb-and-wait" plan. After the plan is verified by the digital twin model, it is then implemented.

[0129] Global optimization module: With time cost, collision probability and energy consumption as multiple objectives, it optimizes the trajectory through intelligent adaptive algorithm to achieve a balance between conflict resolution and operational efficiency.

[0130] The end-to-end anti-interference layer is used to ensure the secure and reliable transmission of traffic control commands and status data, and incorporates self-destruct protection and intelligent learning optimization capabilities.

[0131] Frequency band planning module: Allocate dedicated communication frequency bands for air and road use, and use different frequency bands for the logistics drone layer and the manned eVTOL layer to avoid interference between the same layers;

[0132] Encrypted transmission module: It adopts an improved national cryptographic algorithm (SM9-Plus) and combines it with the blockchain consensus mechanism to realize the encryption authentication and immutability of command transmission, while applying differential privacy technology to protect user privacy data;

[0133] Core protection module: Self-destruct programs and black-box intelligent learning algorithms are embedded in the innermost algorithms of the signal transmitter, receiver, and transmission relay nodes. The self-destruct program is preset to be triggered when the outer encryption algorithm is cracked, malicious intrusion code injection is detected, or the interference intensity exceeds the anti-interference threshold. Once triggered, the signal receiving or transmission process is immediately terminated to ensure that the data is not stolen or tampered with. The black-box algorithm monitors the communication link status in real time, automatically captures and stores key data such as interference signal characteristics and intrusion code fragments to form a security event dataset.

[0134] Anti-interference control module: A protection mechanism is designed across the entire signal link. The transmitting end employs adaptive power adjustment technology (dynamically adjusting the transmission power from 0.1W to 5W based on the communication distance); edge computing relay nodes are deployed during propagation (one every 3km, supporting signal forwarding and enhancement); an interference prediction model based on spectrum maps and historical interference data is constructed, using Gaussian process regression (GPR) for spectrum hole prediction; the receiving end is equipped with an adaptive nulling anti-interference antenna and wavelet transform signal filtering algorithm, combined with beamforming technology to achieve spatial isolation of the communication link; simultaneously, a blockchain consortium chain (nodes include traffic management departments, communication operators, and aircraft companies) is introduced to achieve tamper-proof storage of communication session authentication and frequency band usage records, providing multi-stage shielding against external electromagnetic interference.

[0135] Dynamic no-fly zone module: Based on meteorological data and electromagnetic interference intensity, it automatically delineates temporary no-fly zones to avoid communication failures and security risks in harsh environments.

[0136] The execution feedback layer is used to execute control commands and provide feedback on the execution results, thereby optimizing the twin model.

[0137] Ground execution units: traffic lights, vehicle guidance screens, vehicle-road cooperative terminals;

[0138] Low-altitude execution units: aircraft flight control system, takeoff and landing point scheduling equipment, emergency countermeasures equipment;

[0139] Feedback optimization module: On the one hand, it collects data after execution (such as obstacle avoidance trajectory and command response time) and feeds it back to the digital twin modeling layer to correct the prediction model parameters; on the other hand, it receives the security event dataset uploaded by the black box of the end-to-end anti-interference layer, analyzes the intrusion code and interference features based on artificial intelligence algorithms (integrating deep learning and reinforcement learning), automatically generates new encryption algorithm parameters or anti-interference strategies that are adapted, and updates them to the encryption transmission module and the anti-interference control module to achieve adaptive iteration of "threat capture-model learning-strategy optimization".

[0140] The following example will be used to explain the deployment method and modeling implementation of the integrated road-air traffic safety control twin system of the present invention.

[0141] Taking a megacity (urban area of ​​1500 km², daily average ground traffic of 8 million vehicles, and planned low-altitude aircraft fleet of 100,000 vehicles) as an example, the specific deployment of the system's five-layer architecture is as follows:

[0142] Digital twin modeling layer deployment: Based on Revit 2025, a BIM model of urban infrastructure (LOD400 accuracy, including elements such as bridges, roads, buildings, and take-off and landing points) is built. Low-altitude airspace is divided using ArcGIS Pro 3.0 and a self-developed 3D orientation and meshing plugin (100m×100m×5m 3D mesh). Intrinsic orthogonal decomposition (POD) is used to reduce the dimensionality of historical traffic data (50 million vehicle and aircraft operation data collected over 3 months), extracting the top 200 dominant modes to construct a traffic status snapshot matrix. Real-time rendering and interactive visualization are achieved through Unity3D 2024, supporting simultaneous online display of 100,000 road and air traffic participants, with a model update frequency ≤50ms. The identity coding module assigns a 3D identity code to each type of traffic participant (e.g., "LV-001-60-30" represents a ground vehicle - number 001 - maximum speed 60km / h - load 30 tons), and the coding information is stored in the Neo4j graph database.

[0143] Deployment of the road-air cooperative sensing layer: The ground sensing unit deploys millimeter-wave radar (detection range 500m, positioning accuracy ±0.5m) every 500m along the main urban roads, and simultaneously deploys vehicle-road cooperative sensors and geomagnetic detectors; the low-altitude sensing unit sets up lidar (360° detection, maximum distance 5km) and ADS-B receiving equipment every 1km at high points such as bridges and tall buildings, and the aircraft is required to be equipped with a Remote ID module and a centimeter-level Beidou positioning terminal; the environmental sensing unit integrates meteorological satellite data, ground meteorological station data, and electromagnetic environment monitoring station data (monitoring frequency band 2GHz-6GHz, sampling rate 100MHz); the data fusion module adopts heterogeneous sensor fusion technology, eliminates data errors through Kalman filtering, and generates a standardized dataset (update frequency 100ms).

[0144] Intelligent collision avoidance control layer deployment: A distributed computing cluster is deployed on the city-level traffic cloud platform, including 10 NVIDIA H20 GPU servers and 50 Intel Xeon Gold 6430 CPU servers; a spatiotemporal probabilistic conflict model inference engine is deployed, which reads the state data of the twin model every 100ms, and inputs position (x, y, z) and speed (v). x ,v y ,v z The system calculates the conflict probability using parameters such as [parameters]; the dynamic control module is integrated with the multi-agent self-coordination algorithm, and receives sensing data and outputs control instructions through the Kafka message queue (throughput ≥ 100,000 messages / second).

[0145] Full-link anti-interference layer deployment: Dedicated frequency bands for air and road use (2.4GHz-2.45GHz for logistics drones, 5.725GHz-5.85GHz for manned eVTOL, and 5.9GHz-5.925GHz for emergency communication); deployment of edge computing relay nodes (one every 3km, supporting signal forwarding and enhancement, with a built-in core protection module); encryption transmission module integrating SM9-Plus algorithm and blockchain consortium chain (nodes include traffic management departments, communication operators, etc.), using PBFT consensus mechanism (consensus time ≤300ms); anti-interference control module configured with adaptive zero-adjustment antenna and wavelet transform filtering algorithm, real-time monitoring of communication signal strength (sampling rate 10Hz); in the core protection module, the self-destruct program trigger threshold is set to interference intensity > -60dBm or encryption cracking feature matching degree ≥95%, the black box algorithm uses a 128GB encrypted storage unit, supporting real-time capture and classified storage of interference features and intrusion codes.

[0146] The self-destruct procedure incorporates a safety mechanism, specifically: First, interference is categorized and a progressive response strategy is designed. For minor interference, enhanced filtering is activated and frequency bands are switched; for severe interference, backup communication links are switched, such as from 5G-V2X to a satellite link; for fatal interference, a safety hold mode is entered, the aircraft hovers or returns along a preset safe route, vehicles slowly move to a safe area and stop, and a clear alarm is sent to the control center requesting manual takeover. Simultaneously, design commands and data commands are separated. Channels used for transmitting the highest priority commands such as flight control and emergency braking are isolated from channels used for transmitting large amounts of data such as traffic and video. When interference occurs, the data channel is shut down, while the command channel remains operational, avoiding widespread traffic risks.

[0147] Execution feedback layer deployment: Ground execution units include traffic lights, vehicle guidance screens, and vehicle-road cooperative OBU equipment; low-altitude execution units are connected to the aircraft flight control system, take-off and landing point scheduling equipment, and emergency countermeasure equipment; the feedback optimization module collects command execution data (such as obstacle avoidance trajectory and response time) every 500ms, corrects the BiLSTM-PSO-ELM prediction model parameters through particle swarm optimization algorithm, and updates the twin model state synchronously.

[0148] Example 2

[0149] This embodiment provides a road-air integrated three-dimensional traffic safety control method, based on a road-air integrated three-dimensional traffic safety control twin system of Embodiment 1, including the following steps:

[0150] S1: Construct an integrated digital twin model of the ground, low-altitude, and facilities, and assign unique identification codes to traffic participants;

[0151] S2: Collect road and air traffic and environmental data through multi-source sensors, and perform fusion processing to generate a standardized dataset;

[0152] S3: Based on the twin model and dataset, it predicts collision risk through a spatiotemporal probability conflict model, and generates collision avoidance strategies by combining a three-dimensional collision avoidance mechanism and self-coordinating decision-making.

[0153] S4: Employs encryption algorithms, end-to-end anti-interference technology, and self-destruct protection mechanisms to transmit collision avoidance strategies;

[0154] S5: Execute collision avoidance strategies, collect execution results and safety event data, and provide feedback to optimize the twin model and anti-interference strategies.

[0155] The integrated road-air traffic safety control method of the present invention includes an intelligent collision avoidance control method and a full-link anti-interference control method, wherein, as... Figure 2 As shown, the intelligent collision avoidance control method is as follows:

[0156] Based on the intelligent collision avoidance control layer, collision avoidance for road and air traffic is achieved through a four-step process: "trajectory prediction - conflict assessment - strategy generation - execution verification". The specific steps are as follows:

[0157] Step 1: Accurate prediction of heterogeneous trajectories. The predictive modeling module is invoked, and differentiated predictive models are loaded based on the type of traffic participant: For manually driven vehicles, an extended IDM model is used, with input parameters including desired speed, minimum safe distance, and driver reaction delay; for intelligent connected vehicles, a cooperative adaptive cruise control model is used, incorporating V2X communication data; for low-altitude aircraft, altitude-dimensional motion equations are introduced. , where k z This is the height adjustment coefficient. For the target height, The model outputs the predicted trajectory for the next 5-10 seconds every 100ms, keeping the error ≤1.5m, and synchronizes it to the digital twin modeling layer.

[0158] Step 2: Spatiotemporal conflict probability assessment. The conflict prediction module receives real-time state data from the twin model and heterogeneous predicted trajectories, constructs a four-dimensional conflict assessment matrix (x, y, z, t), and then uses the collision probability function... The system calculates the probability of conflict within the next 5 seconds. The value of the prediction window T is primarily based on the timeliness of the warning and the reliability of the prediction. The total system response time (sensing and communication ≤ 0.5 seconds + decision execution 1-2 seconds + safety buffer 1-2 seconds) is approximately 2.5-4.5 seconds. T must be greater than this value to reserve an effective response window. The error of the prediction model used (BiLSTM-PSO-ELM, etc.) is still controllable within 5 seconds. Beyond 5 seconds, the error increases non-linearly, and the reliability of the assessment decreases. T=5 seconds provides an effective buffer time of approximately 0.5-2.5 seconds while ensuring prediction accuracy. This value can be dynamically adjusted, shortening it during high-risk periods to increase urgency and extending it during low-risk periods to optimize efficiency.

[0159] Step 3: Dynamic collision avoidance strategy generation, outputting differentiated control strategies based on risk level: Low-risk scenarios, maintain the original trajectory, and the dynamic control module adjusts the safety interval according to performance differences (micro UAV ≥ 50m, eVTOL ≥ 200m); Medium-risk scenarios, activate the multi-agent self-coordination mechanism, negotiate obstacle avoidance paths according to the priority of "emergency rescue > medical transportation > logistics transportation > ordinary traffic", such as raising / lowering the aircraft to the buffer layer, and adjusting the vehicle speed or lane; High-risk scenarios, trigger the mandatory control command, the aircraft activates the emergency hovering mode, the vehicle performs emergency braking, and the emergency corridor is activated at the same time.

[0160] Step 4: Strategy verification and execution. The generated collision avoidance strategy is synchronized to the digital twin modeling layer for pre-performance verification. After confirming that there are no secondary conflicts, it is transmitted to the execution unit through the end-to-end anti-interference layer. The execution feedback layer collects data such as obstacle avoidance trajectory and command response time in real time and feeds it back to the prediction modeling module every 500ms to correct the model parameters and improve the subsequent prediction accuracy.

[0161] like Figure 3 As shown, the end-to-end anti-interference control method of the present invention is specifically as follows:

[0162] A six-step process—"frequency band planning, interference prediction, anti-interference control, encrypted transmission, security feedback, and strategy optimization"—is used to construct a full-link anti-interference system to ensure the secure and reliable transmission of control commands. The specific steps are as follows:

[0163] Step 1: Planning and confirmation of dedicated road and air frequency bands. The frequency band planning module, in conjunction with the radio management department, completes the allocation of dedicated road and air frequency bands: 2.4GHz-2.45GHz (bandwidth 50MHz) for logistics drones, 5.725GHz-5.85GHz (bandwidth 125MHz) for manned eVTOL, and 5.9GHz-5.925GHz (bandwidth 25MHz) for emergency communication. A frequency band usage filing mechanism is established, requiring traffic participants to submit frequency band usage applications 24 hours in advance. After approval, they will obtain dedicated communication frequency band authorization to avoid frequency band conflicts within and across the same layer.

[0164] Step 2: Interference signal prediction and detection. The anti-interference control module trains a Gaussian process regression (GPR) model based on historical interference data, receives data from the electromagnetic environment monitoring station in real time, and predicts the intensity and distribution of interference signals in the next 30 seconds. When the interference signal intensity of the target frequency band is detected to be >-80dBm, the spectrum hole prediction is started, and the location of the alternative frequency band with an error ≤200m is output to prepare for frequency band switching in advance.

[0165] Step 3: Execution of end-to-end anti-interference control. The transmitting end adopts adaptive power adjustment technology to dynamically adjust the transmission power according to the communication distance calculated by BeiDou positioning. During propagation, when the signal attenuation is >15dB, the surrounding edge computing relay nodes (one every 3km) are automatically activated to enhance signal directivity through beamforming technology. The receiving end is equipped with an adaptive nulling anti-interference antenna. When interference signals are detected, a 5-layer wavelet transform filtering algorithm is activated to suppress noise interference and ensure that the signal-to-noise ratio is ≥25dB.

[0166] Step 4: Encrypted transmission and core protection are activated. Control commands are encrypted using the SM9-Plus improved national cryptographic algorithm. The key is dynamically generated based on the three-dimensional identity code and is automatically updated every 8 minutes. The encrypted commands are uploaded to the blockchain consortium chain and pushed to the execution unit after being verified by the PBFT consensus mechanism. At the same time, the core protection module is in an active state throughout the process. The self-destruct program and the black box intelligent learning algorithm monitor the outer encryption status and communication link security in real time.

[0167] Step 5: Security Incident Response and Data Capture. When the outer encryption is detected to be cracked, malicious intrusion is detected, or strong interference is detected, the self-destruct program is immediately triggered to terminate signal transmission / reception. The black box algorithm simultaneously captures interference characteristics, intrusion codes, and other data, encrypts and stores them, and marks the event type (interference / intrusion).

[0168] Step 6: Feedback Optimization and Strategy Iteration. The black box uploads the security event dataset to the feedback optimization module of the execution feedback layer. The AI ​​algorithm performs deep learning based on the data features to generate optimized encryption algorithm parameters or anti-interference strategies, which are then updated to the end-to-end anti-interference layer to achieve adaptive improvement in anti-interference capabilities. When meteorological data shows wind speed ≥15m / s or visibility <500m, the dynamic no-fly zone module automatically delineates a temporary no-fly zone and pushes it to all traffic participants through a twin model to avoid communication failures in severe environments.

[0169] This invention also provides an implementation of a collision avoidance mechanism and an application of heterogeneous trajectory prediction, based on a road-air integrated three-dimensional traffic safety control twin system. The specific implementation steps are as follows:

[0170] Step 1: Heterogeneous trajectory prediction is initiated. The prediction modeling module loads the corresponding prediction model according to the type of traffic participant: manually driven vehicles call the extended IDM model, inputting the desired vehicle speed, minimum safe distance, and driver reaction delay parameters; intelligent connected vehicles call the cooperative adaptive cruise control model, integrating V2X communication data; low-altitude aircraft call the motion equations containing the altitude dimension. The prediction model, with the height adjustment coefficient as input. Target height and real-time wind speed disturbance The model outputs predicted trajectory data for the next 5-10 seconds every 100ms.

[0171] Step 2: Calculate the conflict probability. The prediction module receives real-time state data from the twin model and heterogeneous predicted trajectories, constructs a four-dimensional conflict evaluation matrix (x,y,z,t), and then calculates the conflict probability based on the collision probability function. Calculate the probability of a conflict within the next 5 seconds.

[0172] Step 3: Dynamic control strategy generation. Low-risk scenario: Maintain the original trajectory, and the dynamic control module adjusts the safety interval (micro UAV ≥ 50m, eVTOL ≥ 200m); Medium-risk scenario: Activate the self-coordination mechanism, and multiple agents negotiate obstacle avoidance strategies according to the priority of "emergency rescue > medical transportation > logistics transportation > ordinary traffic", such as the aircraft raising / lowering its altitude to the buffer layer, and the vehicle adjusting its speed or lane; High-risk scenario: Trigger the mandatory control command, the aircraft activates the emergency hovering mode, the vehicle performs emergency braking, and the emergency corridor is activated at the same time.

[0173] Step 4: Control command execution and feedback. The generated control commands are transmitted to the execution unit through the end-to-end anti-interference layer. Ground vehicles receive the commands through the vehicle-road cooperative OBU, and the aircraft responds through the flight control system. The execution feedback layer collects data such as obstacle avoidance trajectory and command response time in real time, and feeds it back to the digital twin modeling layer every 500ms to correct the prediction model parameters.

[0174] This invention also provides an anti-signal interference implementation and spectrum management application, based on a road-air integrated three-dimensional traffic safety control twin system, combining spectrum sensing, blockchain authentication, self-destruct protection, and intelligent learning technologies. The specific implementation steps are as follows:

[0175] Step 1: Dedicated frequency band planning and rights confirmation. The frequency band planning module, in conjunction with the radio management department, completes the allocation of dedicated road and air frequency bands. The allocation for logistics drones is 2.4GHz-2.45GHz (bandwidth 50MHz), the allocation for manned eVTOL is 5.725GHz-5.85GHz (bandwidth 125MHz), and the allocation for emergency communication is 5.9GHz-5.925GHz (bandwidth 25MHz). A frequency band usage registration mechanism is established. Traffic participants must submit a frequency band usage application to the management platform 24 hours in advance. After approval, they will obtain a dedicated communication frequency band authorization.

[0176] Step 2: Interference prediction and spectrum hole detection. The anti-interference control module trains a Gaussian process regression (GPR) model based on historical interference data (15,000 electromagnetic interference events collected within one year), receives data from the electromagnetic environment monitoring station in real time (sampling rate 10Hz), and predicts the intensity and distribution of interference signals in the next 30 seconds. When the interference signal intensity of the target frequency band is detected to be >-80dBm, spectrum hole prediction is started, and the location of the alternative frequency band with an error ≤200m is output.

[0177] Step 3: Execution of end-to-end anti-interference control. The transmitting end adopts adaptive power adjustment technology to dynamically adjust the transmission power (0.1W-5W) according to the communication distance (calculated through BeiDou positioning). During propagation, when the signal attenuation is >15dB, the surrounding edge computing relay nodes (one every 3km) are automatically activated, and beamforming technology is used to enhance signal directivity. The receiving end is equipped with an adaptive nulling anti-interference antenna. When interference signals are detected, a 5-layer wavelet transform filtering algorithm is activated to suppress noise interference.

[0178] Step 4: Encrypted transmission and core protection are activated. Control commands are encrypted using the SM9-Plus algorithm, and the key is dynamically generated based on the three-dimensional identity code and automatically updated every 8 minutes. The encrypted commands are uploaded to the blockchain consortium chain, verified by the PBFT consensus mechanism, and then pushed to the execution unit. The core protection module is activated throughout the process, and the self-destruct program and black box algorithm monitor communication security in real time.

[0179] Step 5: Security incident response. When the outer encryption is detected to be cracked (feature matching degree ≥ 95%), malicious intrusion code is injected, or the interference intensity is > -60dBm, the self-destruct program is immediately triggered to terminate signal transmission / reception. The black box algorithm simultaneously captures interference features, intrusion codes, and other data, encrypts and stores them, and marks the event type.

[0180] Step 6: Feedback Optimization and Strategy Iteration. The black box uploads the security event dataset to the feedback optimization module. The AI ​​deep learning algorithm performs feature extraction and pattern recognition on the data to generate optimized encryption algorithm parameters and anti-interference strategies. For example, the key update cycle is shortened to 5 minutes, a new shielding algorithm for specific frequency band interference is added, and it is updated to the end-to-end anti-interference layer. When meteorological data shows that the wind speed is ≥15m / s or the visibility is <500m, the dynamic no-fly module automatically delineates a temporary no-fly zone and pushes it to all traffic participants through a twin model.

[0181] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0182] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A road-air integrated three-dimensional traffic safety control twin system, characterized in that: include: The digital twin modeling layer is used to construct a three-dimensional integrated twin model of the ground, low-altitude, and facilities to achieve real-time mapping and trajectory prediction of road and air traffic conditions. The digital twin modeling layer includes: a predictive modeling module, which performs trajectory prediction based on bidirectional long short-term memory networks, particle swarm optimization extreme learning machines, and heterogeneous traffic subject modeling methods; it introduces intrinsic orthogonal decomposition to reduce the dimensionality and extract features from traffic state data, establishes a traffic state snapshot matrix, and adopts an extended intelligent driver model for manually driven vehicles, a cooperative adaptive cruise model for intelligent connected vehicles, and introduces altitude-dimensional motion equations for low-altitude aircraft. ,in This is the height adjustment coefficient. For the target height, For random wind disturbance, differentiated trajectory prediction is achieved for different types of subjects; height adjustment coefficient. It is the proportional gain of the height control loop, and its physical unit is s. -1 ; The road-air collaborative perception layer is used to collect and fuse multi-source traffic and environmental data to generate a unified format state dataset. The intelligent collision avoidance control layer, based on the twin model and state dataset, predicts risks through a spatiotemporal probability conflict model, and generates collision avoidance strategies by combining a three-dimensional collision avoidance mechanism and a multi-agent self-coordination method. The intelligent collision avoidance control layer includes: a conflict prediction module for constructing a four-dimensional conflict assessment matrix and calculating the collision probability; introducing a spatiotemporal probability conflict model, combining the vehicle's and aircraft's positions (x, y, z) and velocities (v)... x ,v y ,v z ), acceleration (a x ,a y ,a z Based on the heading angle (θ) and airspace grid occupancy status, a four-dimensional conflict assessment matrix is ​​constructed; based on the heterogeneous trajectory prediction results, a collision probability function is defined. ,in , ρ represents the predicted position probability density for the vehicle and the aircraft, respectively, and ρ is the conflict intensity function. T To predict the length of time, t For the current assessment moment, calculate the collision probability; the collision intensity function ρ is used to quantify the instantaneous risk intensity of a collision between the vehicle and the aircraft at predicted positions Pv(τ) and Pa(τ) at a specific future moment τ. The function expression is: ;in, The relative velocity between the two agents, in m / s; The relative distance between two agents, in meters (m). C is a minimum distance constant used to prevent division by zero; C is the comprehensive risk coefficient, in seconds (s); its baseline value C0 is determined by the system's preset standard near-field critical encounter scenario. The end-to-end anti-interference layer is used to ensure the safe and reliable transmission of collision avoidance strategies through dedicated frequency band planning, encrypted transmission, self-destruction procedures, and black-box intelligent learning algorithms. The triggering conditions of the self-destruction procedure include the outer encryption algorithm being cracked, malicious intrusion code being detected, or the interference intensity exceeding a preset threshold. The black-box intelligent learning algorithm is used to capture interference features and intrusion codes and feed them back to the execution feedback layer for strategy optimization. The execution feedback layer is used to execute collision avoidance strategies and collect feedback data to optimize the twin model and anti-interference strategies. The layers are sequentially connected to form a closed-loop management and control architecture encompassing modeling, perception, decision-making, execution, and feedback.

2. The integrated road-air traffic safety control twin system as described in claim 1, characterized in that, The digital twin modeling layer also includes: The geometric modeling module uses BIM, GIS and three-dimensional orientation and domain grid integration technology to construct an integrated three-dimensional model; The dynamic mapping module is used to achieve real-time data synchronization between the physical space and the twin model; The identity coding module assigns a unique three-dimensional identity code to each participant in road and air traffic.

3. The integrated road-air traffic safety control twin system as described in claim 1, characterized in that, The intelligent anti-collision control layer also includes; The dynamic control module outputs differentiated collision avoidance strategies based on risk levels. A self-coordination mechanism is used for autonomous negotiation of collision avoidance paths among multiple agents; The global optimization module optimizes trajectories based on multiple objectives, including time cost, collision probability, and energy consumption.

4. The integrated road-air traffic safety control twin system as described in claim 1, characterized in that, The end-to-link anti-interference layer includes: The frequency band planning module is used to allocate and authorize dedicated air and road communication frequency bands; The encrypted transmission module uses the SM9-Plus algorithm and blockchain consensus mechanism for command encryption and authentication. The core protection module embeds a self-destruct program and a black-box intelligent learning algorithm for threat detection and security data capture. The anti-interference control module implements anti-interference measures throughout the entire signal transmission, propagation, and reception chain. The dynamic no-fly zone module automatically delineates temporary no-fly zones based on environmental data.

5. A road-air integrated three-dimensional traffic safety control method, characterized in that, Based on the integrated road-air traffic safety control twin system as described in any one of claims 1-4, the system includes the following steps: S1: Construct an integrated digital twin model of the ground, low-altitude, and facilities, and assign unique identification codes to traffic participants; S2: Collect road and air traffic and environmental data through multi-source sensors, and perform fusion processing to generate a standardized dataset; S3: Based on the twin model and dataset, it predicts collision risk through a spatiotemporal probability conflict model, and generates collision avoidance strategies by combining a three-dimensional collision avoidance mechanism and self-coordinating decision-making. S4: Employs encryption algorithms, end-to-end anti-interference technology, and self-destruct protection mechanisms to transmit collision avoidance strategies; S5: Execute collision avoidance strategies, collect execution results and safety event data, and provide feedback to optimize the twin model and anti-interference strategies.

6. The integrated road-air traffic safety control method as described in claim 5, characterized in that, The spatiotemporal probabilistic conflict model specifically involves constructing a four-dimensional conflict evaluation matrix (x, y, z, t) based on the collision probability function. Calculate future time periods The probability of conflict within, where, and The time for vehicles and aircraft respectively The predicted location probability density, This is the conflict intensity function.

7. The integrated road-air traffic safety control method as described in claim 5, characterized in that, Step S4 further includes: Plan and authorize dedicated communication frequency bands; Interference prediction and spectral hole detection were performed using a Gaussian process regression model. Adaptive power adjustment, edge relay nodes, and beamforming technology are employed to enhance signal transmission. A self-destruct mechanism is triggered when a threat is detected, and security event data is captured using a black-box algorithm.

8. The integrated road-air traffic safety control method as described in claim 5, characterized in that, Based on feedback data, artificial intelligence algorithms are used to optimize trajectory prediction parameters and anti-interference strategies, thereby achieving adaptive iterative improvement of the system.

9. The integrated road-air traffic safety control method as described in claim 5, characterized in that, It also includes dynamically delineating no-fly zones based on real-time meteorological and electromagnetic environment data, and pushing the information to all traffic participants through a twin model.

Images