Multi-source fusion and V2X network-connected vehicle high-precision positioning system and method

Abstract

The invention relates to the technical field of farm irrigation, in particular to a multi-source fusion and V2X networked vehicle high-precision positioning system and method, and the system comprises a vehicle-mounted terminal which integrates the following modules: a multi-source sensor group which comprises a laser radar (LiDAR), a binocular vision camera (BVC), a global positioning system (GPS), an inertial measurement unit (IMU) and a wheel speed sensor; the V2X communication module supports vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communication and is used for receiving positioning data of an adjacent vehicle and roadside equipment; the system effectively solves the problems that in the prior art, a traditional RTK technology depends on a ground base station, a positioning blind area exists in a signal shielding area, data conflicts are easily caused by independent work of multiple sensors, RTK needs to be covered and supported by dense base stations, and the positioning accuracy is poor. And high-precision positioning cannot be realized in signal shielding scenes such as remote areas, tunnels and mountainous areas, so that an automatic driving function fails.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent connected vehicles, and particularly to a high-precision positioning system and method for connected vehicles with multi-source fusion and V2X. Background Art

[0002] At present, some advanced intelligent connected vehicles using real-time kinematic (RTK) positioning technology have high positioning accuracy, which can reach about 1-10 centimeters. This high-precision positioning is mainly achieved by differential calculation between a satellite positioning system (such as Beidou, GPS, etc.) and a ground base station. However, the coverage range of RTK technology for high-precision positioning is limited, and sufficient base stations are required to ensure signal transmission and differential calculation. In some remote areas or areas with poor signal coverage, centimeter-level positioning may not be achievable. On the other hand, the cost of the RTK system is relatively high, including the construction of base stations, software and hardware. For higher-level autonomous driving functions such as automatic lane change, a large number of vision systems are required to support, and even for a pure vision solution, using cameras on low-end vehicles is obviously not accurate enough, and using lidar on high-end vehicles will also be interfered by radio waves, radar waves, etc., resulting in risks. This undoubtedly limits the wide application of autonomous driving in intelligent connected vehicles.

[0003] Traditional RTK technology relies on ground base stations, has positioning blind spots in signal occlusion areas, and the independent operation of multiple sensors easily leads to data conflicts. RTK requires dense base station coverage support and cannot achieve high-precision positioning in signal occlusion scenarios such as remote areas, tunnels, and mountains, resulting in the failure of autonomous driving functions. The reliability of a single-sensor solution is insufficient; pure vision positioning is easily affected by light and weather, lidar is easily interfered by electromagnetic fields, and inertial navigation has cumulative errors, making it difficult to meet the requirements of complex environments. The communication between vehicles and roads, and between vehicles and clouds depends on cellular networks. When the network coverage is uneven or interrupted, the collaborative decision-making and safety control of autonomous driving face risks. Summary of the Invention

[0004] In view of the above situation, to overcome the defects of the prior art, the present invention provides a high-precision positioning system and method for connected vehicles with multi-source fusion and V2X, which effectively solves the problems in the prior art.

[0005] To achieve the above object, the present invention adopts the following technical solutions: A high-precision positioning system and method for connected vehicles with multi-source fusion and V2X, including: an in-vehicle terminal: integrating the following modules: a multi-source sensor group: including a lidar (LiDAR), a binocular vision camera (BVC), a global positioning system (GPS), an inertial measurement unit (IMU), and a wheel speed sensor;

[0006] V2X communication module: It supports vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications, and is used to receive the positioning data of neighboring vehicles and roadside devices;

[0007] An edge computing unit, connected to the in-vehicle terminal, is configured to: fuse the data of the multi-source sensor group in real time, eliminate errors through the Kalman filtering algorithm, and generate the initial vehicle positioning coordinates;

[0008] Combine the collaborative positioning data received by the V2X communication module, optimize the initial positioning coordinates, and output the corrected high-precision positioning result;

[0009] Cloud server: Stores high-precision map data, receives and processes the disaster scene images and positioning information uploaded by the in-vehicle terminal, generates a three-dimensional post-disaster map and rescue path instructions, and issues them to the target vehicle through V2X communication.

[0010] Preferably, in an area without base station signals, the V2X communication module constructs a dynamic positioning network through V2V communication, including the following steps:

[0011] Vehicle A broadcasts its own positioning coordinates and signal strength through V2V;

[0012] After receiving the signal, Vehicle B calculates the relative distance from Vehicle A in combination with its own sensor data and signal transmission time;

[0013] Generate continuous positioning coordinates for the area without signals based on the collaborative data of multiple vehicles.

[0014] Preferably, the edge computing unit is further configured with an error rejection algorithm, specifically including:

[0015] Perform spatial matching on the LiDAR point cloud data and the BVC image, and eliminate occluded or noisy data;

[0016] Based on the time series analysis of GPS and IMU data, correct short-term positioning drift;

[0017] Verify through the relative distance between vehicles and screen out reliable collaborative positioning data.

[0018] Preferably, the cloud server further includes a disaster scene analysis module, configured to:

[0019] Receive the post-disaster environment images and high-precision positioning coordinates captured by the binocular vision camera uploaded by the in-vehicle terminal;

[0020] Identify obstacles (such as cracks and rubble) through deep learning algorithms and mark three-dimensional coordinates;

[0021] Generate an obstacle avoidance path in combination with the high-precision map and issue it to the rescue vehicle in real time.

[0022] Preferably, it includes the following steps:

[0023] S1 Data acquisition: Collect LiDAR point cloud, BVC images, GPS coordinates, IMU, and wheel speed data through an in-vehicle multi-source sensor group;

[0024] S2 Local positioning: Integrate the multi-source data in an edge computing unit to generate initial vehicle positioning coordinates;

[0025] S3 Cooperative positioning: Obtain the positioning information of neighboring vehicles and roadside devices through V2X communication to optimize the initial coordinates;

[0026] S4 Dynamic compensation: In a signal-free area, construct a relative positioning network between vehicles based on V2V communication to maintain positioning continuity;

[0027] S5 Cloud processing: Upload the post-disaster environment images and positioning data to a cloud server to generate rescue path instructions and transmit them back to the vehicle.

[0028] Preferably, step 4 specifically includes: Before the vehicle enters a signal-free area, cache the positioning data of surrounding vehicles;

[0029] Within the signal-free area, relay the positioning information through V2V communication, and combine IMU data to predict the vehicle's movement trajectory;

[0030] Dynamically correct the relative distance between vehicles according to the signal strength attenuation model.

[0031] Preferably, the generation of the rescue path instructions in step 5 includes:

[0032] Perform semantic segmentation on the post-disaster images to identify passable areas and obstacles;

[0033] Overlay the obstacle coordinates with a high-precision map to generate a three-dimensional obstacle avoidance path;

[0034] Adjust the path planning in real time according to the positioning data of the rescue vehicle.

[0035] Compared with the prior art, the beneficial effects of the present invention are:

[0036] 1. The present invention covers traditional blind spots: By constructing a dynamic positioning network through V2X communication, continuous centimeter-level positioning (accuracy ≤ 15 cm) is achieved in areas without base station signals (such as tunnels and underground parking lots), solving the limitations of traditional RTK technology that relies on base stations;

[0037] 2. Multi-source data fusion: By integrating data from multiple sensors such as LiDAR, binocular vision (BVC), GPS, and IMU, and through Kalman filtering and error elimination algorithms, the positioning accuracy in open areas is improved to ±2 cm (RTK mode), and the accuracy still remains at ±30 cm in complex scenarios.

[0038] 3. Anti-interference ability: The redundant design of multiple sensors reduces the risk of single-point failures. For example, in rainy and foggy weather, the LiDAR point cloud and BVC images complement each other, and the positioning success rate is ≥95%; in an electromagnetic interference environment, the IMU and wheel speed sensors provide short-term motion compensation to avoid positioning interruptions. Dynamic adaptability: The V2X communication module supports vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) cooperation to optimize positioning data in real time.

[0039] 4. Real-time environmental perception: The in-vehicle binocular vision camera (BVC) captures post-disaster images (such as cracks and rubble), and combined with high-precision positioning data (error ≤10 cm), uploads them to the cloud to generate a 3D obstacle avoidance map; Intelligent path planning: The cloud improves the A* algorithm and combines terrain complexity weights to generate the optimal rescue path. Actual measurements show that the response time for post-disaster path planning is ≤200 ms, and the time for rescue vehicles to reach the target point is shortened by 30%. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 FIG. is the overall system architecture diagram of the present invention, showing the collaborative working process of the in-vehicle terminal, edge computing unit, and cloud server;

[0041] Figure 2 FIG. is the flowchart of multi-sensor data fusion of the present invention;

[0042] Figure 3 FIG. is the construction of a dynamic positioning network based on V2V communication in a tunnel scenario of the present invention;

[0043] Figure 4 FIG. is the flowchart of post-disaster rescue path generation of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0044] As Figures 1-4 shown, a high-precision positioning system and method for networked vehicles with multi-source fusion and V2X, 1. System architecture and hardware configuration

[0045] The system of the present invention consists of an in-vehicle terminal, an edge computing unit, and a cloud server, and the specific implementation is as follows:

[0046] (1) In-vehicle terminal

[0047] Multi-source sensor group:

[0048] Light Detection and Ranging (LiDAR): Model Velodyne VLP-16, installed at the front of the vehicle roof, with a horizontal field of view of 360°, a vertical field of view of 30°, a scanning frequency of 10 Hz, generating environmental point cloud data (accuracy ±3 cm).

[0049] Binocular Vision Camera (BVC): Using ZED 2 camera, with a resolution of 3840×1080, a frame rate of 30 fps, and a baseline distance of 120 mm, used to capture RGB images and depth information.

[0050] GPS module: Equipped with a high-precision U-blox ZED-F9P receiver, supporting Beidou / GPS dual-mode positioning, with an output frequency of 5 Hz, and an original positioning accuracy of ±1.5 m (±2 cm in RTK mode).

[0051] Inertial Measurement Unit (IMU): Using Bosch BMI088, with a sampling rate of 100 Hz, measuring acceleration (±16 g) and angular velocity (±2000° / s), used for short-term motion compensation.

[0052] Wheel speed sensor: Integrated into the vehicle's ABS system, collecting the wheel speed in real time (accuracy ±0.1 km / h).

[0053] V2X communication module: Supports DSRC protocol and C-V2X mode, with a working frequency band of 5.9 GHz, a maximum communication distance of 300 m, and supports vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) data exchange 10 times per second.

[0054] (2) Edge computing unit

[0055] Hardware configuration:

[0056] Processor: NVIDIA Jetson AGX Xavier, with a built-in 512-core GPU and 64 GB of memory;

[0057] Storage: 1TB NVMe SSD, used to cache sensor data.

[0058] Software functions:

[0059] Multi-source data synchronization: Align LiDAR, BVC, GPS, and IMU data through timestamps, with a time synchronization error ≤10 ms.

[0060] Kalman filter algorithm:

[0061] State equation:

[0062] X k =F k X k-1 +B k U k +Wk

[0063] Among them, Xk is the vehicle state (position, speed), Fk is the state transition matrix, Uk is the control input (wheel speed data), and Wk is the process noise

[0064] Observation equation:

[0065] Z k = H k X k + V k

[0066] Observation data Z k from the collaborative positioning of GPS and V2X, V k is the observation noise.

[0067] Error rejection algorithm:

[0068] Spatial matching: Project the LiDAR point cloud onto the BVC image coordinate system, and reject the abnormal points (such as flying point noise) beyond the depth range;

[0069] Time series correction: Predict the short-term position drift based on the IMU data. If the deviation between the GPS data and the predicted value exceeds the threshold (±50 cm), the IMU data is preferred.

[0070] (3) Cloud server

[0071] Architecture: Deployed in a distributed cluster, using Kubernetes to manage containerized services;

[0072] High-precision map storage:

[0073] Data format: NDS (Navigation Data Standard), including lane curvature, slope, and traffic sign semantic information (accuracy ±10 cm);

[0074] Update mechanism: Incrementally update every 6 hours through crowdsourced vehicle data.

[0075] Post-disaster analysis module:

[0076] Image processing:

[0077] Input: Post-disaster images uploaded by BVC (resolution 1920×1080, JPEG format);

[0078] Algorithm: A semantic segmentation model based on U-Net to identify cracks, rubble, and tilted buildings (mIoU≥85%);

[0079] Output: Obstacle mask map and three-dimensional coordinates (error ≤20 cm).

[0080] Path Planning:

[0081] Input: Obstacle coordinates + high-precision map;

[0082] Algorithm: Improved A* algorithm, introducing terrain complexity weight into the cost function to generate an obstacle avoidance path;

[0083] Output: The path instructions are sent to the rescue vehicle in JSON format.

[0084] 2. Implementation Process of Dynamic Positioning in Signal-Free Areas

[0085] Scenario: The vehicle enters a tunnel area without base station coverage.

[0086] Step 1: Signal Pre-Caching

[0087] 500m before entering the tunnel, the vehicle receives the positioning data of surrounding vehicles (including coordinates, speed, direction) broadcast by the roadside unit (RSU) through V2I and caches it in the edge computing unit.

[0088] Step 2: V2V Cooperative Positioning

[0089] Data Broadcasting: Vehicle A (the leading vehicle) broadcasts its own coordinates (XA, YA) and signal strength (RSSI) through V2V every second;

[0090] Distance Calculation: After receiving the signal, Vehicle B calculates the relative distance according to the signal transmission time t (synchronized by IEEE 802.11p timestamp):

[0091]

[0092] where c is the speed of light, RSSI0 is the reference signal strength (at 1m), and n is the path loss exponent (2.5 in the tunnel).

[0093] Coordinate Optimization: Vehicle B combines its own IMU data (acceleration a, angular velocity ω), predicts the position through the kinematic model, and performs weighted fusion with the cooperative data:

[0094] x B =α·x IMU +(1-α)·(x A +d AB ·v A )

[0095] where α = 0.7 is the weight coefficient, V A is the velocity direction vector of Vehicle A.

[0096] Step 3: Signal Relay Transmission

[0097] Vehicle C receives the positioning data forwarded by vehicle B in the middle section of the tunnel, repeats step 2 to form a multi-hop communication link, and the maximum relay distance ≤ 150m.

[0098] 3. Post-disaster rescue path generation implementation process

[0099] Scenario: An earthquake causes road collapses, and rescue vehicles need to cross obstacle areas.

[0100] Step 1: Environmental data collection and upload

[0101] The rescue vehicle BVC takes post-disaster images (such as collapsed areas with crack widths ≥ 50 cm), synchronously records high-precision positioning coordinates (error ≤ 10 cm), and uploads them to the cloud through the 5G network.

[0102] Step 2: Cloud obstacle recognition and map reconstruction

[0103] Semantic segmentation:

[0104] Input: Post-disaster images (size 3840×1080);

[0105] Model: Pre-trained U-Net model (Backbone is ResNet-50), outputting pixel-level obstacle labels;

[0106] Post-processing: Eliminating noise through morphological closing operations and extracting the contours of connected regions.

[0107] Three-dimensional coordinate mapping:

[0108] Project the obstacle contours onto the LiDAR point cloud coordinate system and calculate the centroid coordinates (x, y, z);

[0109] Overlay with the high-precision map and label the collapsed areas as "inaccessible".

[0110] Step 3: Obstacle avoidance path planning and distribution

[0111] Path generation:

[0112] Starting point: The current position of the rescue vehicle (x0, y0);

[0113] End point: The target point (x t , y t ) in the disaster area;

[0114] Constraints: Avoid all obstacle coordinates, and the path curvature radius ≥ 5m (to adapt to the turning ability of the rescue vehicle).

[0115] Real-time adjustment:

[0116] If the rescue vehicle deviates from the path, the cloud re-plans and issues new instructions (update frequency 1Hz).

[0117] 4. Algorithm Parameters and Performance Metrics

[0118] Positioning Accuracy:

[0119] Open area: ±2 cm in RTK mode;

[0120] Inside tunnel (V2V cooperation): ±15 cm;

[0121] Disaster area after disaster: ±30 cm (affected by image segmentation error).

[0122] Latency:

[0123] Processing delay of edge computing unit ≤50 ms;

[0124] Response time of cloud path planning ≤200 ms.

[0125] Robustness:

[0126] In rainy and foggy weather, the success rate of LiDAR + BVC fusion positioning ≥95%;

[0127] Packet loss rate of V2V communication ≤5% (inside tunnel).

[0128] Industrial Practicality

[0129] This system has passed the on-vehicle test and verified its performance in the following scenarios:

[0130] Urban tunnel: The total length of the tunnel is 3.2 km, without GPS signal throughout the whole process, and the positioning error ≤20 cm;

[0131] Mountain disaster rescue: Simulating the landslide scenario of the Wenchuan earthquake, the rescue vehicle successfully avoided obstacles and reached the target point, and the path planning accuracy rate ≥90%.

[0132] Build a dynamic positioning network through V2X communication, achieve continuous centimeter-level positioning (accuracy ≤15 cm) in areas without base station signals (such as tunnels and underground parking lots), solve the limitation of traditional RTK technology relying on base stations, combine multi-sensor data such as lidar (LiDAR), binocular vision (BVC), GPS, and IMU, and through Kalman filtering and error elimination algorithms, improve the positioning accuracy in the open area to ±2 cm (RTK mode), and the accuracy in complex scenarios still remains at ±30 cm. Obtain the position and motion state of neighboring vehicles in real time through V2X communication, issue a warning 200 ms before the potential collision risk (such as lane change in blind area), the braking system intervenes in advance, and the accident rate is reduced by 25%. In areas without signals, vehicles relay and transmit positioning information through V2V communication, combined with IMU motion prediction (error ≤0.1 m / s 2) Ensure positioning continuity, avoid the failure of the autonomous driving function, replace a single high-cost device (such as a high-precision lidar) through multi-sensor collaboration, reduce the hardware cost by 35%, the V2X module supports DSRC and C-V2X dual-mode communication, improve the data bandwidth utilization rate by 20%, and reduce the dependence on the 5G network. The V2X module supports DSRC and C-V2X dual-mode communication, improve the data bandwidth utilization rate by 20%, and reduce the dependence on the 5G network.

[0133] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

[0134] In the description of this specification, the description with reference to terms such as "an embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0135] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A high-precision positioning system for a connected vehicle, characterized in that: include: Vehicle terminal: Integrates the following modules: Multi-source sensor group: including laser radar (LiDAR), binocular vision camera (BVC), global positioning system (GPS), inertial measurement unit (IMU) and wheel speed sensor; V2X communication module: supports vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications, and is used to receive positioning data from neighboring vehicles and roadside equipment; An edge computing unit is connected to the vehicle-mounted terminal and is configured to: fuse the data of the multi-source sensor group in real time, eliminate errors through a Kalman filter algorithm, and generate initial positioning coordinates of the vehicle; In combination with the collaborative positioning data received by the V2X communication module, the initial positioning coordinates are optimized, and a corrected high-precision positioning result is output; Cloud server: stores high-precision map data, receives and processes disaster scene images and positioning information uploaded by vehicle-mounted terminals, generates three-dimensional post-disaster maps and rescue path instructions, and sends them to target vehicles through V2X communication.

2. A high-precision positioning system for connected vehicles according to claim 1, characterized in that: The V2X communication module builds a dynamic positioning network through V2V communication in an area without base station signals, including the following steps: Vehicle A broadcasts its own positioning coordinates and signal strength via V2V; After receiving the signal, vehicle B calculates the relative distance to vehicle A based on its own sensor data and signal transmission time; Based on multi-vehicle collaborative data, continuous positioning coordinates of signal-free areas are generated.

3. A high-precision positioning system for connected vehicles according to claim 1, characterized in that: The edge computing unit is further configured with an error elimination algorithm, specifically including: Spatial matching of LiDAR point cloud data and BVC images to remove occluded or noisy data; Correct short-term positioning drift based on time series analysis of GPS and IMU data; Filter the trusted co-location data by verifying the relative distance between vehicles.

4. A high-precision positioning system for connected vehicles according to claim 1, characterized in that: The cloud server also includes a disaster scenario analysis module configured as follows: Receive post-disaster environment images and high-precision positioning coordinates captured by the binocular vision camera uploaded by the vehicle terminal; Identify obstacles (such as cracks and rubble) and mark their 3D coordinates through deep learning algorithms; Combined with high-precision maps, obstacle avoidance paths are generated and sent to rescue vehicles in real time.

5. A high-precision positioning method for a connected vehicle, characterized in that: The following steps are involved: S1 data acquisition: collect LiDAR point cloud, BVC image, GPS coordinates, IMU and wheel speed data through the vehicle-mounted multi-source sensor group; S2 local positioning: fusing the multi-source data in the edge computing unit to generate the initial positioning coordinates of the vehicle; S3 collaborative positioning: obtaining positioning information of neighboring vehicles and roadside equipment through V2X communication to optimize the initial coordinates; S4 dynamic compensation: In areas without signals, a relative positioning network between vehicles is built based on V2V communication to maintain positioning continuity; S5 Cloud Processing: Upload post-disaster environment images and positioning data to the cloud server, generate rescue path instructions and transmit them back to the vehicle.

6. A high-precision positioning method for a connected vehicle according to claim 5, characterized in that: Step 4 specifically includes: before the vehicle enters the no-signal area, caching the positioning data of surrounding vehicles; In areas without signals, positioning information is relayed through V2V communication, and the vehicle trajectory is predicted in combination with IMU data; The relative distance between vehicles is dynamically corrected according to the signal strength attenuation model.

7. A high-precision positioning method for a connected vehicle according to claim 5, characterized in that: The generation of the rescue path instruction in step 5 includes: Perform semantic segmentation on post-disaster images to identify traversable areas and obstacles; Overlay obstacle coordinates with high-precision maps to generate a three-dimensional obstacle avoidance path; Adjust the path planning in real time based on the rescue vehicle positioning data.

Images