A vehicle exterior environment detection device based on V2X

Abstract

The present application belongs to the field of intelligent transportation and Internet of Vehicles, especially a vehicle external environment detection device based on V2X, aiming at the problem that the existing scheme of the background technology focuses on data sharing and lacks deep fusion and collaborative modeling capability of multi-source heterogeneous data. Including a V2X transceiver device supporting DSRC and C-V2X dual-mode communication protocol. The present application significantly reduces the environmental dependence of a single sensor by dynamically adjusting the sensor weight, realizes all-weather high-precision detection, solves the mis-matching problem of multi-sensor data caused by clock deviation or coordinate system difference, can trigger lane keeping assistance or adjust the following distance in advance, effectively avoids chain collision, can predict potential cutting-in risk in advance, significantly improves the response speed of the active safety system, the path planning module generates a detour trajectory accordingly, makes up for the blind spot defect of the traditional sensor, and the device triggers a three-level warning mechanism based on the fusion result of D-S evidence theory.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent transportation and Internet of Vehicles, and particularly relates to a vehicle external environment detection device based on V2X. BACKGROUND

[0002] An autonomous vehicle, also known as self-driving, unmanned driving or unmanned autonomous driving, is a vehicle that uses a computer system to control the flight without a driver. Before the development of autonomous driving technology, various components of the vehicle are operated by the driver to complete different tasks, such as red light detection, vehicle control system coordination, braking, turning, etc. The principle of the autonomous vehicle is that the computer system first inputs the current environmental information from various sensors, such as map information, road distance, vehicle in front of the vehicle, pedestrian, obstacle, etc., and then determines the next action through complex operation. The next action can be directly fed back to the sensor, such as changing the running direction or advancing of the vehicle. The autonomous driving technology mainly uses computers to synthesize complex driving environments to realize real-time control of the vehicle, combined with sensors, GPS and map information, and some specific algorithms, and sends instructions to each part of the vehicle in turn to achieve the purpose of coordinated control.

[0003] The existing autonomous vehicle mainly relies on vehicle-mounted sensors (such as cameras, radars, and laser radars) for environmental perception, which has the following defects:

[0004] 1. The sensor is easily disturbed by weather, shielding and other factors, and the detection distance and accuracy are limited;

[0005] 2. It is impossible to obtain dynamic information in a non-line-of-sight (NLOS) area;

[0006] 3. The coverage range of a single vehicle sensor is limited, and it is difficult to cope with complex traffic scenes.

[0007] Therefore, although the V2X technology can realize information interaction, the existing scheme mainly focuses on data sharing, and lacks the ability of deep fusion and collaborative modeling of multi-source heterogeneous data. SUMMARY

[0008] In view of the defects of the prior art, the present application provides a vehicle external environment detection device based on V2X, which overcomes the defects of the prior art and effectively solves the problem that the existing scheme mainly focuses on data sharing and lacks the ability of deep fusion and collaborative modeling of multi-source heterogeneous data.

[0009] In order to achieve the above purpose, the present application adopts the following technical scheme:

[0010] A vehicle external environment detection device based on V2X, comprising: supporting DSRC and

[0011] A V2X transceiver device of a C-V2X dual-mode communication protocol;

[0012] A heterogeneous data interface connected with vehicle-mounted sensors, the vehicle-mounted sensors including a camera, a millimeter wave radar, and a laser radar;

[0013] A cooperative calibration module for spatio-temporal alignment of multi-source data, a multi-source data fusion processor based on Bayesian inference;

[0014] A control unit for generating a dynamic environment model and outputting a decision instruction, and the environment model is output by the control unit;

[0015] An edge computing unit for real-time processing of camera image data and a NLOS compensation model integrated with a device.

[0016] Preferably, the cooperative calibration module includes a dual-redundancy global clock synchronization mechanism based on NTP and PTP protocols, and a dynamic coordinate system conversion algorithm based on a road side unit (RSU), wherein the dynamic coordinate system conversion algorithm supports real-time mapping of a WGS84 coordinate system and a local coordinate system.

[0017] Preferably, the multi-source data fusion processor includes a channel quality evaluation sub-module and an adaptive filter, wherein the channel quality evaluation sub-module is used to analyze the signal-to-noise ratio, packet loss rate and transmission delay of V2X data packets, and the adaptive filter dynamically adjusts the fusion weight according to data freshness, signal strength and channel quality.

[0018] Preferably, the adaptive filter adopts a hybrid model of Kalman filtering and particle filtering, and introduces a confidence threshold to filter low-reliability data.

[0019] Preferably, the environment model includes a real-time updated three-dimensional occupancy grid map, trajectory prediction data of traffic participants, and emergency event semantic information based on V2X broadcast, wherein the three-dimensional occupancy grid map is used to label obstacle positions and confidence levels, the trajectory prediction data of traffic participants includes speed vectors and motion intention analysis results, and the emergency event semantic information includes traffic accidents, road construction and sudden weather.

[0020] Preferably, the edge computing unit deploys a lightweight convolutional neural network (CNN) model and performs feature-level fusion with V2X data.

[0021] Preferably, the NLOS compensation model obtains reflection signals of the occluded area through RSU relaying, and reconstructs a virtual perception field based on geometric topological relationship.

[0022] Preferably, the NLOS compensation model adopts a multipath signal analysis algorithm and dynamically optimizes the compensation accuracy in combination with the RSU deployment density.

[0023] Preferably, the vehicle external environment detection method of the device comprises the following steps:

[0024] Step S1: receiving and parsing BSM (basic safety message), SPAT (signal phase and timing) and MAP (map data) in the V2X message;

[0025] Step S2: performing space-time calibration on multi-source data to eliminate clock deviation and coordinate system difference;

[0026] Step S3: performing multi-modal data fusion based on D-S evidence theory to generate environment perception results;

[0027] Step S4: triggering hierarchical warning according to the fusion results, including HUD prompt, sound alarm and automatic emergency braking.

[0028] Preferably, the space-time calibration step comprises inter-device synchronization using a high-precision clock signal broadcast by the RSU and real-time coordinate system alignment of the local environment through the SLAM (simultaneous localization and mapping) technology.

[0029] The beneficial effects of the present application are:

[0030] 1. The vehicle external environment detection device based on V2X of the present application has full-scene coverage and high-precision perception. By supporting DSRC and C-V2X dual-mode communication protocols, combining heterogeneous fusion of vehicle-mounted sensors (cameras, millimeter wave radars, laser radars) and V2X data, the device can simultaneously obtain dynamic information in the line-of-sight (LOS) and non-line-of-sight (NLOS) areas. For example, in the intersection scene, the reflected signal of the RSU relay can reconstruct the position of the blocked pedestrian, with a confidence of 90%; in bad weather (such as heavy fog, heavy rain), by dynamically adjusting the sensor weight (such as increasing the V2X data weight to 0.8), the environmental dependence of a single sensor is significantly reduced, realizing all-weather high-precision detection.

[0031] 2. The vehicle external environment detection device based on V2X of the present application has multi-source data deep fusion and space-time consistency guarantee. The device uses a multi-source data fusion processor based on Bayesian inference, combined with dual-redundancy global clock synchronization (NTP / PTP protocol) and dynamic coordinate system conversion algorithm (WGS84 and real-time mapping of local coordinate system), solving the problem of false matching of multi-sensor data caused by clock deviation or coordinate system difference. For example, in the urban intersection scene, after space-time alignment of laser radar point cloud and V2X data, the error is controlled within centimeters, providing a reliable data basis for trajectory prediction and collision risk assessment.

[0032] 3, The vehicle external environment detection device based on V2X of the application, dynamic environment modeling and intelligent decision optimization, the three-dimensional occupancy grid map generated by the control unit supports real-time updating, and the obstacle position and confidence level are labeled, and the trajectory prediction data (speed vector, motion intention analysis) of the traffic participants and the emergency event semantic information (traffic accident, road construction, etc.) broadcast by V2X are combined to realize dynamic environment modeling. For example, in the highway scene, the device obtains the accident information 1 km ahead through V2V communication, triggers the lane keeping assistance or adjusts the following distance in advance, and effectively avoids the chain collision.

[0033] 4, The vehicle external environment detection device based on V2X of the application, edge computing and low-delay response edge computing, the unit deploys a lightweight CNN model to process camera image data in real time (such as identifying vehicle lane changing intention), and performs feature-level fusion with V2X data. In the highway scene, this technology can predict potential cutting-in risks 1.5 seconds in advance, significantly improving the response speed of the active safety system.

[0034] 5, The vehicle external environment detection device based on V2X of the application, NLOS compensation and virtual perception field expansion, the NLOS compensation model reconstructs the virtual perception field of the blocked area through multipath signal analysis algorithm and RSU relay. For example, in a dense fog environment, the construction area information delivered by the RSU can mark the position of the conical bucket in the map, and the path planning module generates a detour trajectory accordingly, which makes up for the blind area defect of traditional sensors.

[0035] 6, The vehicle external environment detection device based on V2X of the application, the fusion result based on D-S evidence theory, the device triggers a three-level warning mechanism: HUD warning, sound alarm and automatic emergency braking (AEB), for example, when the predicted collision time is less than 2 seconds, AEB intervenes and realizes 6m / s 2 deceleration, which maximizes the reduction of accident damage. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 A general flowchart of a vehicle external environment detection device based on V2X is proposed for the application;

[0037] Figure 2 A collaborative calibration module flowchart of a vehicle external environment detection device based on V2X is proposed for the application;

[0038] Figure 3 A multi-source data fusion flowchart of a vehicle external environment detection device based on V2X is proposed for the application;

[0039] Figure 4 A NLOS compensation model flowchart of a vehicle external environment detection device based on V2X is proposed for the application;

[0040] Figure 5 A flow chart of a detection method of a vehicle external environment detection device based on V2X is provided. DETAILED DESCRIPTION

[0041] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all.

[0042] Referring to Figures 1-5 A vehicle external environment detection device based on V2X comprises:

[0043] A V2X transceiving device supporting DSRC and C-V2X dual-mode communication protocols;

[0044] A heterogeneous data interface connected with vehicle-mounted sensors, wherein the vehicle-mounted sensors comprise a camera, a millimeter wave radar, and a laser radar;

[0045] A cooperative calibration module for multi-source data space-time alignment, a multi-source data fusion processor based on Bayesian inference;

[0046] A control unit for generating a dynamic environment model and outputting a decision instruction, and the environment model is output by the control unit;

[0047] An edge computing unit for real-time processing of camera image data and a NLOS compensation model integrated with the device.

[0048] The cooperative calibration module comprises a dual-redundancy global clock synchronization mechanism based on NTP and PTP protocols, and a dynamic coordinate system conversion algorithm taking a road side unit (RSU) as a reference, wherein the dynamic coordinate system conversion algorithm supports real-time mapping of a WGS84 coordinate system and a local coordinate system.

[0049] The multi-source data fusion processor comprises a channel quality evaluation sub-module and an adaptive filter, wherein the channel quality evaluation sub-module is used to analyze the signal-to-noise ratio, packet loss rate, and transmission delay of V2X data packets, and the adaptive filter dynamically adjusts the fusion weight according to data freshness, signal strength, and channel quality.

[0050] The adaptive filter adopts a hybrid model of Kalman filtering and particle filtering, and introduces a confidence threshold to filter low-reliability data.

[0051] The environment model includes a real-time updated three-dimensional occupancy grid map, trajectory prediction data of traffic participants, and emergency semantic information based on V2X broadcast, wherein the three-dimensional occupancy grid map is used to mark obstacle positions and confidence levels, the trajectory prediction data of traffic participants includes speed vectors and motion intention analysis results, and the emergency semantic information includes traffic accidents, road construction, and sudden weather.

[0052] The edge computing unit deploys a lightweight convolutional neural network (CNN) model and performs feature-level fusion with V2X data.

[0053] The NLOS compensation model obtains reflection signals of the occluded area through RSU relaying, and reconstructs a virtual perception field based on geometric topological relationships.

[0054] The NLOS compensation model adopts a multipath signal analysis algorithm and dynamically optimizes compensation accuracy in combination with RSU deployment density.

[0055] The vehicle external environment detection method of the device includes the following steps:

[0056] Step S1: receiving and parsing BSM (basic safety message), SPAT (signal phase and timing), and MAP (map data) in V2X messages;

[0057] Step S2: performing spatiotemporal calibration on multi-source data to eliminate clock bias and coordinate system differences;

[0058] Step S3: performing multi-modal data fusion based on D-S evidence theory to generate environment perception results;

[0059] Step S4: triggering hierarchical warning according to the fusion results, including HUD prompt, sound alarm, and automatic emergency braking.

[0060] The spatiotemporal calibration step includes device synchronization using high-precision clock signals broadcast by RSUs and real-time coordinate system alignment of the local environment through SLAM (simultaneous localization and mapping) technology.

[0061] Embodiment 1: urban intersection scenario

[0062] Step 1: data collection

[0063] The host vehicle receives BSM data (including GPS position, speed, and heading angle) of adjacent vehicles through the V2X transceiver, and receives SPAT signal lamp countdown information sent by the road side unit (RSU) and V2P (vehicle-to-pedestrian) signals of pedestrians in the blind area.

[0064] Vehicle-mounted sensors (laser radar, millimeter wave radar, camera) synchronously collect local environment data, including obstacle distance, pedestrian profile, and traffic light status.

[0065] Step 2: Spatio-temporal alignment

[0066] The cooperative calibration module synchronizes the clocks between devices to sub-microsecond precision based on the RSU through the NTP protocol, and uses a dynamic coordinate system conversion algorithm to unify V2X data and laser radar point clouds to the WGS84 coordinate system (error < 5 cm).

[0067] SLAM technology constructs a local environment map in real time and aligns it with the global coordinate system, eliminating positioning deviations (such as Figure 2 indicated).

[0068] Step 3: Multi-source data fusion

[0069] The multi-source data fusion processor dynamically adjusts the weights based on channel quality evaluation results (signal-to-noise ratio > 20 dB, packet loss rate < 1%): in rainy and foggy weather, the millimeter wave radar data weight is set to 0.7, and the V2X data weight is set to 0.3.

[0070] The adaptive filter uses a hybrid model of Kalman filtering and particle filtering to filter low-reliability data (such as noise interference point clouds) with a confidence level below 80%.

[0071] Step 4: NLOS compensation and virtual perception

[0072] The NLOS compensation model obtains the reflected signals of the bus occluded area through RSU relaying, reconstructs the pedestrian position (confidence 90%) based on geometric topological relationships, and inputs the data into a three-dimensional occupancy grid map (such as Figure 3 indicated).

[0073] Step 5: Environment modeling and decision execution

[0074] The control unit generates a real-time updated three-dimensional map, labels pedestrians, vehicles, and signal light status, and calculates pedestrian crossing risk (collision probability > 95%) based on trajectory prediction algorithms.

[0075] When the predicted collision time is less than 2 seconds, trigger a three-level warning:

[0076] First level: HUD displays a red warning box, prompting the driver to brake urgently;

[0077] Second level: buzzer emits a sharp alarm sound (frequency 2 kHz);

[0078] Third level: automatic emergency braking (AEB) intervention, achieving a deceleration of 6 m / s 2 (as Figure 4 indicated).

[0079] Example 2: Highway scenario

[0080] Step 1: Long-distance information acquisition

[0081] The V2X communication receives broadcast information of a traffic accident 1 km ahead (including accident coordinates, number of vehicles, and lane occupancy status), and the cooperative mapping module maps the accident point to the local map (accuracy ±1 m).

[0082] Step 2: Multi-vehicle cooperative perception

[0083] Through V2V communication, the emergency braking state (deceleration >4 m / s 2 ) of vehicles in adjacent lanes is obtained, and after fusing the millimeter wave radar data of the vehicle, the following distance is adjusted to a safety threshold (minimum 50 m).

[0084] Step 3: Edge computing and risk prediction

[0085] The lightweight CNN model deployed by the edge computing unit identifies the lane-changing intention of the vehicle ahead in real time (such as turn signal activation, wheel deflection angle >5°), and predicts the potential risk of cutting in combination with V2X data.

[0086] When the risk level reaches the threshold, the lane keeping assist (LKA) is triggered 1.5 seconds in advance, and the steering wheel torque output is 3 N·m (as shown in Figure 5 ).

[0087] Example 3: Adverse weather scenario

[0088] Step 1: Dynamic adjustment of sensor weights

[0089] In a heavy fog environment, the camera is disabled due to insufficient visibility, and the control unit increases the weight of V2X data to 0.8 and reduces the weight of millimeter wave radar to 0.2, ensuring continuous perception.

[0090] Step 2: Virtual perception field expansion

[0091] The RSU relays the construction area information blocked by thick fog (such as cone coordinates, construction range), and the control unit marks the obstacle position in the map, and the path planning module generates a detour trajectory (lateral offset 2.5 m).

[0092] Step 3: Hierarchical warning optimization

[0093] Due to low visibility, the system adjusts the HUD warning box color to high-contrast yellow and reduces the buzzer frequency to 1 kHz to reduce interference.

[0094] Example 4: Tunnel environment scenario

[0095] Step 1: GPS signal compensation

[0096] When GPS signal is lost inside the tunnel, the device switches to SLAM technology combined with V2X data (such as the tunnel internal map provided by RSU) to achieve precise positioning (error <0.3m).

[0097] Step 2: Multi-modal data redundancy

[0098] Laser radar and millimeter wave radar data fusion, building a three-dimensional point cloud map inside the tunnel, labeling static obstacles (such as maintenance vehicles) and dynamic targets (such as motorcycles).

[0099] Step 3: Emergency braking coordination

[0100] If the vehicle in front suddenly brakes (deceleration >8m / s 2 ), the system broadcasts warning information through V2V, and the rear vehicle synchronously triggers AEB, forming a coordinated braking chain.

[0101] Example 5: Night driving scenario

[0102] Step 1: Low-light environment enhancement

[0103] The camera switches to infrared mode, and the edge computing unit enhances the image contrast through the CNN model to identify pedestrian reflective markers and animal outlines.

[0104] Step 2: V2X and sensor complementation

[0105] RSU provides night road construction information (such as temporary detour signs), which is fused with vehicle-mounted radar data to optimize path planning (such as speed limit adjustment to 30km / h).

[0106] Step 3: Anti-glare processing

[0107] The control unit dynamically adjusts the brightness of the HUD and the position of the warning box according to the light intensity of the oncoming vehicle (>2000 lumens) to avoid visual interference for the driver.

[0108] Device structure and working principle:

[0109] The core modules of the device include:

[0110] 1. V2X transceiver: supports DSRC (5.9GHz frequency band) and C-V2X (LTE / 5G) dual-mode communication, maximum coverage radius 1.5km.

[0111] 2. Cooperative calibration module: integrated dual-redundant clock synchronization mechanism (NTP error <1μs, PTP error <100ns) and SLAM algorithm to ensure spatiotemporal consistency.

[0112] 3. Multi-source data fusion processor: hybrid filtering model (Kalman filtering + particle filtering) is adopted, and the data update frequency is 100 Hz.

[0113] 4. NLOS compensation model: based on multi-path signal analysis (time delay spread < 50 ns), compensation accuracy reaches 85%.

[0114] 5. Edge computing unit: equipped with a lightweight CNN model (parameter quantity < 1M), and the processing delay is less than 10 ms.

[0115] The device workflow is as follows:

[0116] 1. Data acquisition -> 2. Spatio-temporal alignment -> 3. Data fusion -> 4. Environment modeling -> 5. Decision execution, full closed-loop control, to ensure real-time and reliability.

[0117] In the description of the present application, it should be understood that the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise" and the like indicate the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application.

[0118] In addition, the terms "first", "second" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise specifically limited.

[0119] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any skilled person in the art can make equivalent replacement or change according to the technical solution and the inventive concept of the present application within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A V2X-based vehicle external environment detection device, characterized in that, include: A V2X transceiver that supports both DSRC and C-V2X dual-mode communication protocols; A heterogeneous data interface for connecting to vehicle-mounted sensors, including cameras, millimeter-wave radar, and lidar; The collaborative calibration module for spatiotemporal alignment of multi-source data is a multi-source data fusion processor based on Bayesian inference; A control unit used to generate a dynamic environment model and output decision instructions, wherein the environment model is output by the control unit; NLOS compensation model for edge computing units and device integration for real-time processing of camera image data; The collaborative calibration module includes a dual-redundant global clock synchronization mechanism based on NTP and PTP protocols, and a dynamic coordinate system transformation algorithm based on the roadside unit (RSU). The dynamic coordinate system transformation algorithm supports real-time mapping between the WGS84 coordinate system and the local coordinate system. The multi-source data fusion processor includes a channel quality assessment submodule and an adaptive filter. The channel quality assessment submodule is used to analyze the signal-to-noise ratio, packet loss rate and transmission delay of V2X data packets, and the adaptive filter dynamically adjusts the fusion weights according to data freshness, signal strength and channel quality. The adaptive filter adopts a hybrid model of Kalman filtering and particle filtering, and introduces a confidence threshold to filter low-reliability data. The environmental model includes a real-time updated 3D occupancy grid map, trajectory prediction data of traffic participants, and emergency event semantic information based on V2X broadcast. The 3D occupancy grid map is used to mark the location and confidence level of obstacles, the trajectory prediction data of traffic participants includes velocity vectors and motion intention analysis results, and the emergency event semantic information includes traffic accidents, road construction, and sudden weather.

2. The vehicle external environment detection device based on V2X according to claim 1, characterized in that, The edge computing unit deploys a lightweight convolutional neural network (CNN) model and performs feature-level fusion with V2X data.

3. The vehicle external environment detection device based on V2X according to claim 1, characterized in that, The NLOS compensation model obtains the reflected signals of the occluded area through RSU relay and reconstructs the virtual sensing field based on geometric topological relationships.

4. The vehicle external environment detection device based on V2X according to claim 1, characterized in that, The NLOS compensation model employs a multipath signal analysis algorithm and dynamically optimizes the compensation accuracy by combining RSU deployment density.

5. A method for detecting the external environment of a vehicle based on the device described in any one of claims 1-4, characterized in that, Includes the following steps: Step S1: Receive and parse the BSM basic security message, SPAT signal phase and timing, and MAP map data in the V2X message; Step S2: Perform spatiotemporal calibration on the multi-source data to eliminate clock bias and coordinate system differences; Step S3: Perform multimodal data fusion based on DS evidence theory to generate environmental perception results; Step S4: Trigger tiered warnings based on the fusion results, including HUD prompts, audible alarms, and automatic emergency braking.

6. The method according to claim 5, characterized in that, The spatiotemporal calibration steps include synchronizing between devices using a high-precision clock signal broadcast by the RSU and aligning the local environment with a coordinate system in real time using SLAM synchronous positioning and mapping technology.

Images