Method and device for ground-ground integrated dynamic traffic flow analysis and road topology reasoning

Abstract

The present invention discloses a method and device for ground-to-space dynamic traffic flow analysis and road topology reasoning. The cloud server generates and publishes real-time traffic information through multi-level processing such as map matching, traffic flow analysis, prediction and warning, and road topology analysis and update. Specifically, the method first utilizes the high-precision positioning and communication capabilities of low-orbit satellites, combined with the data fusion of inertial measurement units and wheel speed sensors, to construct a multi-source fusion algorithm to improve positioning accuracy. Secondly, a Transformer-based interactive analysis model is used to perform dynamic analysis and prediction of traffic flow, and clustering algorithms and graph theory models are used to achieve real-time updates of road topology structures in the cloud. Finally, traffic flow and road topology information are sent to relevant vehicles and traffic management systems to achieve real-time monitoring, warning, and optimization of traffic status. The present invention can effectively improve the accuracy and real-time performance of traffic information collection, processing, and publication, and enhance the intelligent level of traffic management and decision-making.

Description

Technical Field

[0001] The present invention relates to the fields of intelligent transportation systems and autonomous driving, and in particular to a device and method for traffic flow analysis and road topology reasoning and updating based on low-orbit satellite (LEO) communication and high-precision positioning technology. Background Art

[0002] With the rapid development of intelligent transportation systems (ITS) and autonomous driving technologies, vehicle-to-vehicle collaboration and accurate, real-time perception of traffic information have become key technologies for improving traffic efficiency and safety. Traditional methods of collecting traffic information, such as fixed traffic monitoring equipment, information collection vehicles, or ground-based communication technologies, suffer from limited coverage and slow update speeds, making them inadequate for meeting the demands of future highly automated traffic scenarios. Low-orbit satellites, with their low latency, high bandwidth, and high-precision satellite-based positioning, have become a crucial technological foundation for supporting future intelligent transportation systems. However, research and application of high-precision traffic flow collection and road topology information inference strategies using collaboration between low-orbit satellites and vehicles is currently underdeveloped. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention provides a method and device for dynamic traffic flow analysis and road topology inference based on real-time low-orbit satellite communications and high-precision positioning. This method enables the accurate collection, inference, publication, and updating of traffic information. Combining intelligent vehicle information collection equipment, low-orbit satellite communications, and a cloud-based data processing and distribution system, this method forms a highly coordinated intelligent transportation system. This method enables real-time inference, publication, and updating of traffic information, enabling functions such as traffic flow monitoring, road topology analysis, and real-time early warning.

[0004] The present invention achieves the above technical objectives through the following technical means.

[0005] The ground-to-space dynamic traffic flow analysis and road topology reasoning device includes:

[0006] Ground monitoring stations to monitor data from satellites in orbit;

[0007] The intelligent vehicle onboard module obtains the vehicle's real-time heading, speed, and location based on low-orbit satellite data and ground monitoring station data;

[0008] The cloud server receives data transmitted by the onboard modules of intelligent vehicles, performs multi-level calculations and analysis, and implements two sets of task flows based on map matching, including traffic flow distribution, prediction and warning, and road topology analysis and update;

[0009] The ground information interaction module is used to receive road topology and traffic flow information processed by the cloud server and distribute it to the target vehicle or traffic management system to realize the sharing of road status and traffic information.

[0010] In the above technical solution, the intelligent vehicle on-board module includes but is not limited to an inertial measurement unit, a wheel speed sensor, a low-orbit satellite communication module and a data processing unit.

[0011] A ground-ground integrated dynamic traffic flow analysis and road topology reasoning method:

[0012] The intelligent vehicle on-board module obtains the vehicle's real-time motion status data based on low-orbit satellite data and ground monitoring station data, and uploads it to the cloud server;

[0013] The cloud server uses the motion status data of smart vehicles passing through key traffic areas to achieve preliminary matching of map data using navigation-level maps. It then conducts high-precision docking of vehicle trajectories with road maps to achieve precise matching of each vehicle's map position.

[0014] Traffic flow distribution, prediction, and warning: The cloud server collects all motion status data within a specified time window, calculates average vehicle speed, traffic density, and vehicle spacing distribution, and uses a Transformer-based driving scenario heterogeneous element interaction model to analyze the dynamic relationship between vehicles and road map elements to determine traffic flow status, congestion, and danger conditions.

[0015] The vehicle motion state data accumulated over a long period of time in the detection area is used to analyze and update the lane centerline and lane topology diagram of the road, thereby realizing the analysis and update of the road topology structure;

[0016] Distribute traffic flow information and road topology to target vehicles or traffic management systems.

[0017] Furthermore, the data structure collected by the cloud server includes the collected vehicle motion status information and prior map information : , ;in, For the scene The time series state vector of each vehicle, For the scene vectorized representation of map elements, and Respectively represent the vehicles and road map elements contained in the currently detected driving scene; and Input into the Transformer-based heterogeneous scene element interaction model, analyze the comprehensive spatiotemporal interaction relationship of scene elements, and obtain the scene dynamic interaction feature information; use the scene dynamic interaction feature information to analyze traffic status, congestion, and dangerous situations, and further analyze the traffic density and speed in each traffic area. Combined with historical data and current interaction feature information, predict and identify congestion or collision risk areas and their risk probabilities and output them.

[0018] Furthermore, combining the lane centerline and the turning and lane-changing behaviors in the vehicle trajectory, a lane topology relationship diagram is constructed using a graph theory model, and it is defined that: nodes represent intersections or key locations, edges represent lane connection relationships, edge weights are traffic constraints, and special topological structures are marked.

[0019] Furthermore, the road topology structure is updated by comparing the existing high-precision road map information with the lane topology relationship diagram, detecting newly added, closed or adjusted lanes, marking the changed positions, and incrementally updating the high-precision map; at the same time, based on the detection results, the navigation map is updated and sent to the relevant vehicles.

[0020] Furthermore, when uploading to the cloud server, traffic anomaly information adopts an instant transmission strategy and is uploaded first, while regular vehicle status information is transmitted at a predetermined frequency.

[0021] Furthermore, when uploading to the cloud server, the time slot allocation for data transmission is dynamically adjusted according to the number of vehicles and communication bandwidth.

[0022] Furthermore, the traffic flow analysis has the highest priority and performs real-time corresponding traffic flow prediction and warning; the road topology analysis serves as an auxiliary and is responsible for long-term updating and improving information sharing.

[0023] Furthermore, the distribution rule is: according to the vehicle position and driving direction, the target vehicle that needs to receive the data is determined, and the information is preferentially distributed to the target vehicle that is closer to the event point.

[0024] The beneficial effects of the present invention are:

[0025] (1) This paper designs a multi-source data fusion strategy by combining low-orbit satellite high-precision positioning and vehicle status information collection, which can significantly improve the vehicle status data collection accuracy and data transmission reliability, and ensure the stability and adaptability of the system decision-making and control strategy in complex traffic environments.

[0026] (2) The present invention constructs a real-time traffic flow analysis and prediction model, uses short-term trajectory data to monitor dynamic traffic flow status, and combines neural networks to predict future traffic flow trends, thereby achieving real-time detection and early warning of traffic congestion and dangerous situations, effectively improving the initiative and intelligence level of traffic management.

[0027] (3) Based on the long-term vehicle trajectory data, the present invention adopts clustering and graph theory methods to dynamically generate road topology structure, and combines it with the existing navigation map for real-time update. It can quickly identify road changes (such as new lanes or intersection adjustments), ensure the accuracy and real-time nature of traffic information, and improve the path planning efficiency of the vehicle's automatic driving system.

[0028] (4) The present invention combines traffic flow analysis, vehicle flow prediction and road topology update through a multi-task collaborative processing architecture, optimizes data utilization efficiency, enhances the system's global perception capability, and provides an efficient support solution for intelligent transportation systems and autonomous driving technologies.

[0029] (5) The method of the present invention can be seamlessly integrated with existing traffic management platforms and autonomous driving vehicle control systems, providing technical support for the comprehensive upgrade of intelligent transportation infrastructure and the widespread application of autonomous driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 Schematic diagram of the system architecture of the disclosed embodiment of the present invention.

[0031] Figure 2 The present invention discloses a flow chart of traffic flow analysis and road topology reasoning based on low-orbit satellite communication and high-precision positioning in an embodiment.

[0032] Figure 3 The present invention discloses a data processing method flow chart of an embodiment. DETAILED DESCRIPTION

[0033] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but the protection scope of the present invention is not limited thereto.

[0034] The first aspect is the specific implementation of the system architecture. The system architecture of the present invention consists of an intelligent vehicle onboard module, a cloud server, a ground monitoring station, and a ground information interaction module. Each module works together to achieve high integration and efficient operation. The specific implementation is as follows:

[0035] Smart vehicle onboard modules include but are not limited to an inertial measurement unit (IMU), wheel speed sensors, a low-orbit satellite communication module, and a data processing unit. For example, in a sedan, the IMU is installed at the center of the vehicle chassis and secured by a rigid bracket to ensure accurate sensing of changes in vehicle acceleration and angular velocity. The wheel speed sensor is installed near the wheel hub, concentric with the wheel's axis of rotation, to accurately measure wheel rotational speed. After the IMU and wheel speed sensor are installed, calibration is performed. Static calibration of the IMU collects IMU data in multiple attitudes to calculate and compensate for errors such as the IMU's zero bias and scale factor. Dynamic calibration of the wheel speed sensor uses a standard speed source. The actual speed is compared with the speed measured by the wheel speed sensor, and the sensor parameters are adjusted to minimize measurement errors. The low-orbit satellite communication module receives communication signals from multiple satellites and data from ground monitoring stations, corrects errors, and implements positioning and navigation data. The data collected by the inertial measurement unit, wheel speed sensor, and low-orbit satellite communication module are processed by the data processing unit through Kalman filtering or nonlinear least squares optimization to remove noise interference and realize the output of real-time vehicle heading, speed, position and other information.

[0036] Cloud servers utilize a high-performance server cluster architecture to collect intelligent vehicle positioning and navigation data, meeting the demands of concurrently processing large amounts of vehicle data. For example, when processing traffic flow analysis tasks, the cluster management system distributes computing tasks to individual nodes for parallel execution, significantly increasing data processing speed and ensuring efficient operation of the entire system. This provides a solid hardware foundation for subsequent tasks such as map matching, traffic flow analysis, and road topology generation.

[0037] The ground monitoring station is used to monitor in-orbit satellite data, conduct two-way communication with low-orbit satellites, provide data forwarding, signal reception and transmission functions, and ensure the stability and accuracy of the signal between the satellite and user equipment.

[0038] Ground information interaction modules, such as satellite communication receivers and 4G / 5G network information receivers, are used to receive road topology and traffic flow information processed by cloud servers, thereby realizing the sharing of road status and traffic information.

[0039] The second aspect is the specific implementation of the method flow, which is used to illustrate the data processing method and link based on low-orbit satellite communication and high-precision positioning, specifically including:

[0040] S1. Data Acquisition. The low-orbit satellite communication module receives multiple sets of satellite signals and calculates the vehicle's real-time position, speed, and time synchronization. An improved differential positioning algorithm based on ground monitoring stations is used to improve positioning accuracy. Specifically, the vehicle's low-orbit satellite communication module continuously receives multiple sets of low-orbit satellite data and calculates the pseudorange between the vehicle and the satellite by measuring the propagation time of the satellite signals. Precise positioning is also achieved using data from ground monitoring stations, which are distributed around the target area and have precisely known location coordinates. After receiving the satellite signals, the ground monitoring stations calculate the measurement errors of the satellite signals, including errors caused by factors such as atmospheric delay and satellite orbit errors, and transmit this error information to the vehicle via the network.

[0041] After receiving satellite signals, the low-orbit satellite communication module, based on the high-precision positioning provided by low-orbit satellites, combines onboard state estimation sensors, such as the inertial measurement unit (IMU) and wheel speed sensors, to achieve high-precision acquisition of the vehicle's motion state. The IMU measures the vehicle's acceleration and angular velocity, and integrates them to obtain the vehicle's velocity and position change. During the positioning calculation process, a Kalman filter algorithm is used to fuse satellite positioning data with inertial navigation data. The Kalman filter's state vector includes the vehicle's position, velocity, acceleration, and IMU error state. First, low-orbit satellites provide global positioning information for the vehicle, accurately calculating its position, velocity, and heading. Then, the IMU provides acceleration and angular velocity data for estimating the vehicle's dynamic changes. This inertial measurement compensates for positioning gaps, particularly in environments with unstable or blocked satellite signals. Wheel speed sensors measure tire rotation speed in real time, inferring the vehicle's speed and further calibrating the vehicle's position. By fusing these sensor data, errors from individual sensors are effectively eliminated, improving positioning accuracy. This ensures continuous and accurate acquisition of vehicle status, especially in complex terrain or where signals are blocked.

[0042] S2. Data Upload. Processed vehicle data is quickly transmitted to the cloud server via a low-orbit satellite communication module. This involves data packaging and compression. According to a predefined communication protocol, the processed vehicle data is compressed into small data packets to reduce transmission latency. The data packets contain a unique vehicle identifier (such as a license plate number or device ID) to allow the cloud server to distinguish the data source.

[0043] During the transmission process, a priority strategy is established to prioritize data. Traffic anomalies (such as sudden braking or collisions) are uploaded first, while regular status information (position, speed, and heading) is transmitted at a predetermined frequency. Traffic anomaly detection uses a method based on acceleration mutation thresholds: the vehicle's inertial measurement unit (IMU) monitors acceleration changes in real time. When the absolute value of acceleration exceeds a preset threshold (such as 8 m / s²) within a short period of time (e.g., 2 seconds), the system uses wheel speed sensor data to determine whether the vehicle is in a sudden braking or collision state.

[0044] In terms of data transmission frequency and time slot allocation, an instant transmission strategy is adopted for traffic anomaly information. Once an anomaly is detected, the data is immediately sent to the cloud server via the low-orbit satellite communication module, giving priority to communication resources to ensure the timeliness of the information. For routine status information, it is transmitted at a predetermined frequency, for example, every 10 seconds. In terms of time slot allocation, dynamic adjustments are made based on the number of vehicles and communication bandwidth: when there are fewer vehicles on the road and communication resources are more abundant, the transmission time slots for routine status information are appropriately increased to improve the real-time nature of the data; when there are more vehicles and communication is more busy, the transmission time slots for routine status information are reasonably reduced to ensure the priority transmission of emergency information, avoid network congestion, and ensure the communication stability of the entire system and the efficiency of data transmission.

[0045] Data integrity and reliability are ensured during transmission by combining the TCP / IP protocol with the link redundancy mechanisms of low-orbit satellites. A retransmission mechanism is implemented to prevent data loss. After sending a data packet, the sender waits for an acknowledgment (ACK) from the receiver. If no ACK is received within a certain period of time, the packet is automatically retransmitted, up to a maximum of 10 times, to prevent data loss due to network congestion or brief satellite signal interruptions.

[0046] Furthermore, a multi-link transmission strategy is employed, leveraging the link redundancy mechanisms of low-orbit satellites. For example, the vehicle's low-orbit satellite communication module simultaneously establishes connections with four or more low-orbit satellites, sending data to the cloud server via multiple links. On the receiving end, the cloud server compares and verifies the multiple copies of data received, selecting the most complete and accurate data for processing, effectively improving data transmission reliability.

[0047] S3. Data Processing. The cloud server performs multi-level computation and analysis on the uploaded data and implements two sets of task flows based on map matching, including traffic flow distribution inference, prediction and warning, and road topology analysis and update. The details are as follows:

[0048] S31. Map Matching: The cloud server uses a key traffic area as the smallest unit, collecting motion status data (i.e., the vehicle's real-time heading, speed, and location) transmitted by intelligent vehicles passing through it within a certain period, and performs multiple tasks. Taking the intersection of a city's bustling commercial district as an example of a key traffic area, the cloud server sets the scope for collecting intelligent vehicle trajectory status data to a circular area with a radius of 2 kilometers centered on the commercial district, and the time period is set to collect data every 10 minutes. A navigation-level map is used to achieve preliminary matching of map data, and particle filtering technology is further used to accurately match vehicle trajectories with road maps, ensuring that each vehicle's position on the map is accurately matched, removing obvious noise data while completing missing state quantities.

[0049] Particle filtering technology is used to accurately connect vehicle trajectories with road maps. Specifically: ;in, is the posterior probability of the state quantity at the time to be inferred; It is the observation model, which is obtained from the data uploaded by the vehicle; is the prior distribution, i.e., the state quantity inferred by the vehicle kinematic model and historical state input; is the normalization factor of the observation, and is the selected empirical scalar.

[0050] S32. Real-time traffic flow analysis, prediction, and early warning. This task flow analyzes the traffic flow status of the current traffic area based on short-term collected vehicle status data, promptly identifying congestion or dangerous situations. The cloud server collects all trajectory data within a specified time window and calculates macro indicators such as average vehicle speed, traffic density, and vehicle spacing distribution. It also uses a Transformer-based driving scenario heterogeneous element interaction model to analyze the dynamic relationships between vehicles and road map elements to determine traffic flow status, congestion, and dangerous situations.

[0051] The data structure collected by the cloud server includes the collected vehicle motion status information and prior map information : , ;in, For the scene The time series state vector of each vehicle, For the scene vectorized representation of map elements, and They represent the vehicles and road map elements contained in the currently detected driving scene respectively.

[0052] During the data screening and preprocessing phase, a statistical outlier detection method is employed. This involves setting reasonable thresholds based on the statistical distribution of parameters such as speed and acceleration during normal vehicle operation. For example, for speed data, the mean and standard deviation of all vehicle speeds within a region are calculated. Speed ​​values ​​outside the range of ±3 times the standard deviation of the mean are considered outliers and removed. Similar statistical methods are used for outlier processing of acceleration data to remove abnormal data caused by sensor failures or unusual driving situations (such as rapid restarts after sudden braking).

[0053] Then and Input into the Transformer-based heterogeneous scene element interaction model to analyze the comprehensive spatiotemporal interaction relationship of scene elements: Where C is the pre-processed macro-indicator characteristics such as average vehicle speed, traffic density, and vehicle spacing distribution, and H is the scene dynamic interactive feature information. Multiple decoder output heads will use this interactive feature information to analyze and warn of traffic flow status, congestion, and dangerous situations: The decoder analyzes traffic density and speed within each area, combining historical data with current spatiotemporal interaction characteristics to predict and identify areas of congestion or collision risk and their associated risk probabilities. R consists of multiple sequences, each containing the location of areas with congestion or collision risk and their corresponding risk values.

[0054] This invention achieves scenario risk recognition and prediction in cloud-based traffic flow analysis, further flagging and transmitting congestion or hazard information to other relevant vehicles to provide risk warnings. It also outputs a visual traffic flow status diagram for real-time monitoring and decision-making in the cloud or at a traffic management center. New real-time data is continuously collected to update the prediction model (i.e., a Transformer-based heterogeneous scene element interaction model) to improve long-term accuracy. The generation and display of the visual traffic flow status diagram utilizes Geographic Information System (GIS) technology to present traffic flow data in an intuitive graphical format. On the cloud platform, different colored lines represent vehicle speeds in different lanes: green indicates unobstructed lanes with speeds exceeding 40 km / h, yellow indicates slow-moving lanes with speeds between 20 and 40 km / h, and red indicates congested lanes with speeds below 20 km / h. At the same time, flashing icons are used to mark the locations where abnormal traffic events (such as accidents, sudden braking, etc.) occur, so that traffic management personnel can clearly understand the traffic conditions in the entire area, provide strong support for traffic decision-making, and also make it convenient for vehicle drivers to obtain real-time traffic flow information through on-board displays or mobile phone applications and optimize travel routes.

[0055] When training and validating the heterogeneous scene element interaction model using historical traffic data, the past year's traffic data for the city's monitored intersections was used as an example, with 80% of the data used as the training set and 20% as the validation set. During training, an adaptive learning rate adjustment strategy was employed, dynamically adjusting the learning rate based on the model's loss on the validation set. Regarding the model's online update mechanism, incremental learning updates are performed every hour using newly collected real-time traffic data to adapt to dynamic changes in traffic conditions, ensuring that the model's predictive performance remains at a consistently high level and providing continuous and reliable support for traffic management.

[0056] S33. Road topology analysis and update. The long-term accumulated vehicle motion data within the detection area can be used to analyze and update the lane centerline and topology of the road to ensure the real-time and accuracy of road information. The details are as follows:

[0057] Trajectory Data Accumulation: This system collects trajectory data for all vehicles within the detection area over a longer time window (e.g., several hours to several days), including historical locations, driving directions, and speeds. In this embodiment, the time window for long-term trajectory data accumulation is set to several days, specifically three days. In terms of data processing, a coordinate conversion algorithm based on spatial projection is used to convert the vehicle's latitude and longitude coordinates into planar coordinates based on the road, forming a set of road trajectory points.

[0058] Lane Centerline Extraction: Using a clustering algorithm (DBSCAN or MeanShift), trajectory points are clustered into lane clusters. Polynomial regression or B-spline curve fitting is then applied to the data points in each cluster to generate lane centerlines. When using the clustering algorithm, clustering parameters are optimized based on the actual road width and traffic volume. For main roads with high traffic volume and a lane width of 3.5 meters, the neighborhood radius is set to 2 meters, and the minimum number of sample points is set to 50 to ensure accurate clustering of trajectory points into lane clusters. For branch roads with low traffic volume and a lane width of 3 meters, the neighborhood radius is set to 1.5 meters, and the minimum number of sample points is set to 30. Regarding curve fitting, second-order polynomial regression is used for straight lanes, while third-order B-spline curve fitting is used for curves with a curvature radius of less than 100 meters to ensure that the centerline accurately describes the lane shape, keeping the average error of the lane centerline within 0.3 meters.

[0059] Construction of a lane topology diagram: A graph theory model is used to construct a road topology diagram based on the turning and lane-changing behaviors of lane centerlines and vehicle trajectories. The diagram is defined as follows: nodes represent intersections or key locations, edges represent lane connections, and edge weights represent traffic constraints (e.g., whether there are traffic restrictions, speed limits, flow restrictions, etc.). Special topological structures (e.g., ramps, roundabouts, or merging lanes) are also marked.

[0060] Based on actual road changes, such as a newly built ramp in the current area, vehicle trajectory data is analyzed to identify vehicle steering behavior and speed change characteristics in the area, determine the starting point, end point, and connection relationship of the ramp, and accurately add it to the road topology diagram. In terms of navigation map comparison and updates, the generated road topology diagram is compared with the lane information in the existing navigation map on a daily basis to detect newly added, closed, or adjusted lanes, mark the changed locations, and perform incremental updates to the high-precision map. Based on the detection results, the navigation map is updated and sent to relevant vehicles to ensure the accuracy of the information. For example, when a lane closure is detected on a road due to construction, the location and impact range of the closed lane are determined through statistical analysis of vehicle trajectory data, and the change information is promptly updated to the navigation map to ensure that the navigation map can accurately reflect the actual road conditions, provide a reliable basis for vehicle navigation and route planning, and improve the overall operational efficiency of the intelligent transportation system.

[0061] S34. Comprehensive Strategy and Task Collaboration. To efficiently execute these tasks, this invention designs task coordination and priority management under cloud-based control. Real-time traffic flow analysis has the highest priority, providing real-time traffic flow prediction and early warning. Road topology analysis serves as an auxiliary tool, primarily responsible for long-term updates and improvements to information sharing. The output results of each task can be shared via cloud servers to various vehicle terminals and traffic management centers, utilizing low-orbit satellite communications to ensure timely and efficient information transmission. The dynamic priority adjustment mechanism among tasks such as real-time traffic flow analysis, traffic flow prediction and warning, and road topology analysis is as follows: Under normal traffic conditions, real-time traffic flow analysis is set to the highest priority and is executed every 5 minutes; the traffic flow prediction and warning task is triggered based on the results of real-time traffic flow analysis. When there is a significant change in traffic flow (such as a sudden change in vehicle speed or an increase in traffic density), the prediction and warning task is executed immediately; road topology analysis is updated on a daily basis to assist other tasks; when abnormal traffic events occur (such as accidents, bad weather, etc.), the priority of real-time traffic flow analysis is further increased, and the execution cycle is shortened to once every 2 minutes. At the same time, the execution frequency of traffic flow prediction and warning tasks is also accelerated accordingly to respond to traffic changes more promptly and ensure traffic safety.

[0062] Information sharing utilizes a cloud-based publish / subscribe model, with data formatted in JSON (JavaScript Object Notation) to facilitate parsing and interaction between different systems and devices. For example, road topology information is published as a JSON object containing fields such as road nodes, lane connectivity, and lane attributes (such as width, direction, and speed limit); traffic flow information is published as a JSON array containing fields such as area ID, average speed, traffic density, and congestion status. This approach enables vehicles and traffic management systems to quickly and accurately obtain the required information, enabling efficient information sharing.

[0063] S4. Information distribution: Distribute the traffic flow and road topology information processed by the cloud server to relevant vehicles or traffic management systems. The specific steps include:

[0064] Distribution rules: Target vehicles are identified based on their location and direction of travel. A method combining geofencing technology and vehicle trajectory prediction is employed. First, a geographic information system (GIS) is used to construct a geofence around the target area on the cloud server, dividing it into multiple sub-areas, each with a unique identifier. When a vehicle uploads its location information, the cloud server identifies the geofence sub-area it is in and quickly selects the vehicles that may need to receive the information.

[0065] At the same time, the system combines historical vehicle trajectory data with real-time speed and direction information, applying a linear regression model from machine learning to make short-term predictions about the vehicle's trajectory. For example, by analyzing a vehicle's trajectory points over the past 10 minutes, it can determine its direction and speed trends and predict the area the vehicle is likely to reach within the next 20 minutes. The predictions are then matched against geofenced sub-regions to further pinpoint the target vehicles that need to receive specific information.

[0066] To prioritize information distribution to vehicles closer to the event point, a distance threshold of 20 kilometers is set. When a traffic event (such as congestion or an accident) is detected, the cloud server immediately calculates the number of vehicles within a 20-kilometer radius of the event point and prioritizes the delivery of relevant traffic flow and road topology information to these vehicles. Regarding the timeliness of information delivery, the delay from the event occurrence to the delivery of information to the target vehicle must not exceed 5 seconds to ensure that vehicles can obtain the latest traffic information and make decisions in a timely manner.

[0067] Data content: Road topology information: including real-time updated lane divisions, road section capacity, road closure or congestion status; traffic flow information: such as average speed, traffic density, and predicted driving paths.

[0068] Within road topology, lane division information is stored in a vector data format. Each lane consists of a series of ordered coordinate points with an accuracy of 0.2 meters, accurately describing the lane's shape and position. Road segment capacity is estimated based on design standards (such as the number of lanes and road grade) and historical traffic flow data. This data is measured in vehicles per hour and updated every half hour to reflect real-time changes in road capacity.

[0069] Transmission mechanism: Low-orbit satellite communications are used to achieve real-time data distribution; for nearby vehicles, information can be directly shared through V2V (vehicle-to-vehicle communication).

[0070] The embodiments described are preferred implementations of the present invention, but the present invention is not limited to the above implementations. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essence of the present invention are within the scope of protection of the present invention.

Claims

1. A ground-to-ground dynamic traffic flow analysis and road topology reasoning method, characterized by: The method is implemented based on the following device, which includes: Ground monitoring stations to monitor data from satellites in orbit; The intelligent vehicle onboard module obtains the vehicle's real-time heading, speed, and location based on low-orbit satellite data and ground monitoring station data; The cloud server receives data transmitted by the onboard modules of intelligent vehicles, performs multi-level calculations and analysis, and implements two sets of task flows based on map matching, including traffic flow distribution, prediction and warning, and road topology analysis and update; The ground information interaction module is used to receive road topology and traffic flow information processed by the cloud server and distribute it to the target vehicle or traffic management system to realize the sharing of road status and traffic information; The method is specifically as follows: The intelligent vehicle on-board module obtains the vehicle's real-time motion status data based on low-orbit satellite data and ground monitoring station data, and uploads it to the cloud server; The cloud server uses the motion status data of intelligent vehicles passing through key traffic areas to achieve preliminary matching of map data using navigation-level maps. It then performs high-precision alignment of vehicle trajectories with road maps to achieve precise matching of each vehicle's map position. Traffic flow distribution, prediction, and warning: The cloud server collects all motion status data within a specified time window, calculates average vehicle speed, traffic density, and vehicle spacing distribution, and uses a Transformer-based driving scenario heterogeneous element interaction model to analyze the dynamic relationship between vehicles and road map elements to determine traffic flow status, congestion, and danger conditions. The vehicle motion state data accumulated over a long period of time in the detection area is used to analyze and update the lane centerline and lane topology diagram of the road, thereby realizing the analysis and update of the road topology structure; Traffic flow information and road topology are distributed to target vehicles or traffic management systems. The data structure collected by the cloud server includes the collected vehicle motion state information X and prior map information M: Among them, x i is the temporal state vector of the i-th vehicle in the scene, m k is the kth vectorized representation map element in the scene, N and M represent the vehicle and road map elements contained in the currently detected driving scene, respectively; Input X and M into the Transformer-based heterogeneous scene element interaction model, analyze the comprehensive spatiotemporal interaction relationship of scene elements, and obtain the scene dynamic interaction feature information; The dynamic interactive feature information of the scene is used to analyze the traffic status, congestion, and dangerous situations, and further analyze the traffic density and speed in each traffic area. Combined with historical data and current interactive feature information, congestion or collision risk areas and their risk probabilities are predicted and identified and output.

2. The method for analyzing dynamic traffic flow and road topology inference based on the ground-ground integration according to claim 1, characterized in that: Combined with the lane centerline and the turning and lane changing behaviors in the vehicle trajectory, a lane topology relationship diagram is constructed using a graph theory model, and the following definitions are defined: nodes represent intersections or key locations, edges represent lane connection relationships, edge weights are traffic constraints, and special topological structures are marked. The lane centerline uses a clustering algorithm to cluster trajectory points into lane clusters, and polynomial regression or B-spline curve fitting is used to generate the data points of each cluster. When using the clustering algorithm, the clustering parameters are optimized according to the actual width of the road and traffic flow: for main roads with large traffic volume and a lane width of 3.5 meters, The neighborhood radius is set to 2 meters and the minimum number of sample points is set to 50 to ensure that trajectory points can be accurately clustered into lane clusters; for branches with light traffic and lane widths of 3 meters, the neighborhood radius is set to 1.5 meters and the minimum number of sample points is set to 30; in terms of the application of curve fitting methods, for straight lanes, second-order polynomial regression is used for fitting; for curves with a curvature radius of less than 100 meters, third-order B-spline curve fitting is used to ensure that the centerline can accurately describe the shape of the lane, so that the average error of the lane centerline is controlled within 0.3 meters.

3. The method for analyzing dynamic traffic flow and road topology inference based on the ground-ground integration according to claim 2 is characterized in that: The road topology structure is updated by comparing the existing high-precision road map information with the lane topology diagram, detecting newly added, closed, or adjusted lanes, marking the changed locations, and incrementally updating the high-precision map; at the same time, based on the detection results, the navigation map is updated and sent to the relevant vehicles.

4. The method for analyzing dynamic traffic flow and road topology inference based on the ground-ground integration according to claim 1, characterized in that: When uploading to the cloud server, traffic anomaly information adopts an immediate transmission strategy and is uploaded first, while regular vehicle status information is transmitted at a predetermined frequency.

5. The method for analyzing dynamic traffic flow and road topology inference based on the ground-ground integration according to claim 1 is characterized in that: When uploading to the cloud server, the time slot allocation for data transmission is dynamically adjusted according to the number of vehicles and communication bandwidth.

6. The method for ground-to-ground dynamic traffic flow analysis and road topology reasoning according to claim 1, characterized in that: The traffic flow analysis has the highest priority and performs real-time traffic flow prediction and warning; the road topology analysis serves as an auxiliary and is responsible for long-term updating and improving information sharing; the dynamic adjustment mechanism of priority among real-time traffic flow analysis, traffic flow prediction and warning, and road topology analysis is as follows: under normal traffic conditions, the priority of real-time traffic flow analysis is set to the highest and is executed every 5 minutes; the traffic flow prediction and warning task is triggered based on the results of real-time traffic flow analysis, and when there is a significant change in traffic flow, the prediction and warning task is executed immediately; road topology analysis is updated once a day for a long time to assist other tasks; when a traffic abnormality occurs, the priority of real-time traffic flow analysis is further improved, and the execution cycle is shortened to once every 2 minutes. At the same time, the traffic flow prediction and warning task is also accelerated accordingly to respond to traffic changes more promptly and ensure traffic safety.

7. The method for analyzing dynamic traffic flow and road topology inference based on the ground-ground integration according to claim 1, characterized in that: The distribution rule is: according to the vehicle position and driving direction, the target vehicle that needs to receive data is determined, and the information is distributed preferentially to the target vehicle that is closer to the event point.

8. The method for analyzing dynamic traffic flow and road topology inference based on the ground-ground integration according to claim 1, characterized in that: The intelligent vehicle on-board module includes but is not limited to an inertial measurement unit, a wheel speed sensor, a low-orbit satellite communication module and a data processing unit.

Images