Intelligent auxiliary braking method for road driving skill test based on environment perception

Abstract

The application provides an environmental perception-based intelligent auxiliary braking method for road driving skill examination, which can collect the running state of the examination vehicle and the surrounding traffic environment, compare the collected running state and the surrounding traffic environment with preset safety boundary conditions, judge whether the vehicle is in a safe running state, support automatic intervention braking for dangerous driving situations, and avoid accidents or reduce the severity of accidents.

Description

Technical Field

[0001] This invention relates to the field of intelligent traffic control technology, specifically to an intelligent assisted braking method for road driving skills testing based on environmental perception. Background Technology

[0002] With the widespread use of automobiles, the demand for driver's licenses is constantly increasing. The road driving skills test is a mandatory subject for obtaining a driver's license and a crucial step in comprehensively assessing basic driving skills and safe driving awareness. Current road driving skills tests are conducted on public roads used by other vehicles and generally employ a combination of a road driving skills testing system (hereinafter referred to as the "test system") and in-vehicle human evaluation. The test system focuses on assessing basic driving skills, such as the use of turn signals, braking and deceleration, gear shifting, and lane marking; human evaluation primarily assesses safe driving awareness and abilities, including driving operation, observation while driving, and the ability to identify dangerous situations caused by improper operation.

[0003] However, due to the relatively complex traffic conditions during road tests and the limited driving skills and safety awareness of test takers, there is a certain probability of dangerous situations arising during the test. In such cases, the examiner mainly relies on stepping on the auxiliary brake pedal of the test vehicle to apply emergency braking and avoid accidents. However, due to limitations such as attention and reaction time, it is impossible for a human to react promptly to all dangerous situations at all times, thus road test-related traffic accidents occur frequently. Summary of the Invention

[0004] To address the issues of delayed identification of dangerous situations and slow braking by human examiners during road tests, this invention provides an intelligent assisted braking method for road driving skills testing based on environmental perception. This method identifies the vehicle's operating status in real time and automatically intervenes to brake in a timely manner once a dangerous driving situation occurs, thereby preventing accidents or reducing their severity.

[0005] The technical solution of this invention is as follows: an intelligent assisted braking method for road driving skills testing based on environmental perception, characterized by comprising the following steps: S1: Install a decision-making module, road test data acquisition and processing equipment, and an execution module on the test vehicle; The road test data acquisition and processing equipment includes: lidar, millimeter-wave radar and short-range lidar installed on the vehicle; The lidar collects three-dimensional information around the vehicle, identifies surrounding background objects and traffic environment through calculation and processing, and outputs point cloud data, specifically including: object position, velocity and contour. The near-range lidar is used to collect distance and motion data of background objects in the lane where the test vehicle is located and in adjacent lanes; the installation positions of the near-range lidar include: the front left, front right, rear left, and rear right of the vehicle; The millimeter-wave radar is installed in front of and behind the vehicle to collect background objects and motion data in front of and behind the vehicle. The execution module brakes the test vehicle based on the instructions from the decision module; The decision-making module performs calculations based on the data collected by the road test data acquisition and processing equipment, identifies safety boundary conditions, and issues instructions to the execution module based on the calculation results; S2: Based on the road test data acquisition and processing equipment, collect the position and motion data of the test vehicle and background objects; S3: Calculate the relative position and relative speed between the test vehicle and the background objects at the current vehicle speed; If there are background objects in front of the test vehicle in the same lane, proceed to step S4; If the test vehicle changes lanes, proceed to step S5; Otherwise, repeat steps S2 to S3. S4: Compare the speed V of the test vehicle. 考试车 Speed ​​V of the vehicle in front 背景物 ; If the following safety boundary conditions are met simultaneously, then proceed to step S7; V 考试车 >V 背景物 And Δt≤3s; Where △t is the estimated time for the test vehicle to collide with the background object in front, △t = △s / (V 考试车 -V 背景物 ), △s is the relative distance between the background object in the same lane ahead and the test vehicle; Otherwise, repeat steps S2 to S3. S5: Detect the area behind the test vehicle in the left and right adjacent lanes. If there are background objects, proceed to step S6; otherwise, repeat steps S2 to S3. S6: Execute step S7 when any one of the following safety boundary conditions is met: Condition 1: The relative distance Δs between the background object and the test vehicle is less than or equal to 3m; Condition 2: The relative distance Δs between the background object and the test vehicle satisfies: 3 < Δs ≤ 30m, and Δt ≤ 3s; S7: Issue a vehicle braking command to the execution module.

[0006] Its further features are: It also includes the following steps: a1: A positioning antenna is installed on the test vehicle to provide real-time location coordinates of the test vehicle; a2: Collect the location coordinates and on-board signal status of the test vehicle, and overlay the data of the test vehicle at different times with time as the axis to generate driving trajectory data; a3: Integrate driving trajectory data with high-precision electronic maps, overlay vehicle motion data, the lane where the test vehicle is located, and the position of the vehicle body outline points on the electronic map of the test route, and combine the traffic data collected by the map with the test items to monitor the operation status of the test vehicle in real time. The vehicle motion status data includes: vehicle speed, driving direction, acceleration, deceleration or constant speed. The test vehicle's operating status includes: the test vehicle's location and driving behavior, wherein the driving behavior includes: lane changing, lane crossing, and turning; In step S3, the relative position and relative speed between the test vehicle and the background object at the current vehicle speed are calculated, specifically including the following steps: b1: Based on the three-dimensional information around the vehicle collected by the lidar, point cloud data is formed; Each point in the point cloud data includes: the point's spatial coordinates x, y, z, and reflection intensity value; b2: Perform a first filter on the point cloud data to remove interference in the test road section environment and obtain the data after the first filter; Specifically, the following steps are included: b21: Construct an interference filtering model based on the types of interference present in the test route; b22: Using the test vehicle as the center, specify the filtering radius and construct the filtering range; b23: Based on the interference filtering model, remove the data outside the filtering range and the interference within the filtering range to obtain the data after the first filtering. b3: Perform a second filtering on the point cloud data to remove the interfering point cloud data of the test road surface, and obtain the data after the second filtering; Specifically, the following steps are included: b31: Create a spatial reference plane: Using the lidar of the test vehicle as the origin of the coordinate system, simulate a reference plane parallel to the ground; b32: Read the distance threshold preset in the system; The distance threshold includes: the highest position above the lidar and the lowest position below the lidar; b33: Read the data after the first filtering and calculate the distance from each point to the spatial reference plane; Points exceeding the highest and lowest positions are filtered out to obtain the second filtered data; b4: Generate a 3D point cloud view of the background objects within the test route; Specifically, the following steps are included: b41: Read the preset clustering threshold, maximum point cloud count, and minimum point cloud count; The clustering threshold represents the distance between points within the same cluster, and is a range value. The maximum and minimum point cloud counts indicate that only clusters with a number of points between the maximum and minimum point cloud counts are considered valid clustering results. b42: Read the data after the second filtering, and record it as: data of points to be clustered; b42: Using the clustering threshold, calculate the distance between points in the data to be clustered to obtain a point cloud cluster set; b43: For all the point cloud cluster sets, determine the points in the point cloud cluster sets whose number of points is between the maximum and minimum number of point clouds, and release them as the data to be clustered; b44: Repeat steps b42~b43 until the number of points in all the point cloud cluster sets obtained each time is between the maximum number of points and the minimum number of points. b45: Each of the point cloud clusters represents a three-dimensional point cloud view of the corresponding background object; b5: Compress the three-dimensional point cloud view of the background object into a two-dimensional space; Specifically, the following steps are included: b51: Construct a plane parallel to the ground as a two-dimensional reference plane, wherein the two-dimensional reference plane is not higher than the height of the test vehicle; b52: Obtain each point cloud cluster set, find the outermost point of each point cloud cluster set mapped on the two-dimensional reference plane, and use it as the two-dimensional feature point of the corresponding background object. The remaining points are recorded as non-feature points. b53: Filter out all the non-feature points and display the two-dimensional feature points of each background object on a two-dimensional plane to obtain a two-dimensional planar image of each background object. b6: Integrate the 2D plan view of the background object, the test vehicle positioning data, and the high-precision map. Objects within the lane are marked as background objects for calculation, while background objects outside the lane are deleted as interference. By combining the time information, the real-time relative positional relationship between the test vehicle and the background object can be obtained, and then the relative distance between the test vehicle and the background object can be calculated. b7: Correlate the data of the two frames before and after the background object with time as the axis to obtain the continuous motion trajectory information of the background object, and then calculate the relative speed between the background object and the test vehicle. The execution module includes: a mechanical structure and an electronic structure; The road test data acquisition and processing equipment also includes: an on-board signal acquisition device and a differential positioning device; The differential positioning device includes a positioning base station and a positioning mobile station, used to collect the location information of the test vehicle, and calculate the current vehicle motion data and operating status by combining the current GPS position and direction with a pre-loaded vehicle contour model. The on-board data acquisition unit includes the vehicle ECU, which is used to analyze and convert vehicle signals.

[0007] This application provides an intelligent assisted braking method for road driving skills testing based on environmental perception. It can collect data on the test vehicle's own operating status and the surrounding traffic environment, compare this data with preset safety boundary conditions, and determine whether the vehicle is in a safe operating state. For dangerous driving situations, it supports automatic braking intervention to avoid accidents or reduce their severity. The safety boundary conditions in this method are clear and explicit. Based on the relative distance and speed between the test vehicle and the background vehicle or pedestrian, safety boundary conditions are set for the test vehicle's operating status to ensure that the test vehicle will not collide with other vehicles. This method is particularly suitable for the safety protection of less experienced drivers in test vehicles. Attached Figure Description

[0008] Figure 1 A schematic diagram showing the installation and deployment locations of sensors on the test vehicle; Figure 2 A schematic diagram showing the installation and deployment locations of system modules on the test vehicle; Figure 3 This is an illustration of a following vehicle scenario; Figure 4 This is a schematic diagram of a lane changing scenario; Figure 5 This is a schematic diagram of the system operation of the test vehicle based on this method; Figure 6 Example of rendering a point cloud data from a LiDAR system; Figure 7 This is an example of the point cloud data after the first filtering. Figure 8 This is an example of the point cloud data after the second filtering. Figure 9 This is an example of a 3D point cloud view of the target object after clustering. Figure 10 An example of a 3D point cloud view of the clustered test route; Figure 11 This is an example of two-dimensional point cloud data results within a lane. Detailed Implementation

[0009] like Figures 1-2As shown, the hardware used in this method includes: a decision module, a road test data acquisition and processing device, and an execution module.

[0010] The road test data acquisition and processing equipment includes: radar sensors mounted on the vehicle beam, an onboard signal acquisition unit, and a differential positioning device. By acquiring the test vehicle's own coordinate position and onboard signals, and simultaneously collecting motion data, speed, and distance data of traffic participants around the test vehicle, the motion data of the test vehicle and background objects are compared to obtain key evaluation information such as relative speed, relative distance, and operational data. The vehicle's motion data in this method includes: vehicle speed, acceleration, and direction of travel.

[0011] like Figure 1 As shown, the radar sensors installed on the vehicle include: lidar, millimeter-wave radar, and short-range lidar.

[0012] LiDAR collects three-dimensional information about the vehicle's surroundings, calculates and processes it to identify background objects and traffic environment, and outputs point cloud data, specifically including object position, velocity, and contour.

[0013] Near-range lidar is used to collect distance and motion data of background objects in the lane where the test vehicle is located and in adjacent lanes; the installation positions of near-range lidar include: the front left, front right, rear left and rear right of the vehicle.

[0014] Depending on the specific circumstances of the examination venue, the background objects in this method are all things that may appear on the examination road surface, including: motor vehicles, non-motor vehicles, people, animals, or fixed objects, or combinations of these.

[0015] Millimeter-wave radar is installed in front of and behind the vehicle to collect background objects and motion data in front of and behind the vehicle.

[0016] The differential positioning equipment includes a positioning base station and a positioning mobile station, which are used to collect the location information of the test vehicle. By combining the current GPS position and direction with a pre-loaded vehicle contour model, the current vehicle motion data and operating status are calculated.

[0017] Vehicle motion data includes: vehicle speed, direction of travel, acceleration, deceleration, or constant speed. The operating status of the test vehicle includes: the location of the test vehicle and the driving behavior of the vehicle, including: lane changing, lane crossing and turning; The on-board data acquisition unit includes the vehicle's ECU (engine control unit), which is used to analyze and convert vehicle signals.

[0018] like Figure 2As shown, the decision-making module consists of an industrial control computer, a switch, and a high-precision electronic map, all integrated into a chassis located in the trunk of the test vehicle. The industrial control computer is used to support system operation, data processing, and storage of the electronic map.

[0019] The high-precision electronic map consists of two parts: a vehicle model and a test route model. The vehicle model mainly includes the vehicle's outline coordinate data. The test route model includes the test locations for road test items and traffic facility data for the test sections. Test items include the location information of starting, going straight at intersections, turning left at intersections, turning right at intersections, changing lanes, overtaking, passing school zones, passing bus stops, and passing pedestrian crossings within the test section. Traffic data for the test section includes traffic information such as dashed / solid lines, speed limit signs, and lane directions. Traffic information is stored on the electronic map using coordinate data. The system retrieves road section information from the map and calls the corresponding safety boundary algorithm model. A switch is used for transmitting positioning data, vehicle signals, sensor data, and system commands.

[0020] The execution module comprises both mechanical and electronic structures. The mechanical execution module consists of a motor and a linkage mechanism, installed at the secondary brake pedal location on the test vehicle. After the decision module issues a braking command, the motor starts running, simulating manual depressing of the secondary brake pedal via the linkage mechanism to brake the test vehicle. The electronic execution module, referencing the control principles of the Automotive Emergency Braking (AEB) system, sends hydraulic pressure commands to the vehicle's electronic stability control system. Through coordinated operation between various vehicle systems, it instantly cuts off or reduces power and simultaneously activates the vehicle's braking system. Currently, most test vehicles, considering cost, employ mechanical execution modules for vehicle braking.

[0021] The execution module brakes the test vehicle based on instructions from the decision-making module. The decision-making module performs calculations based on data collected by the road test data acquisition and processing equipment, and issues instructions to the execution module based on the calculation results.

[0022] This application includes an intelligent assisted braking method for road driving skills testing based on environmental perception, which includes the following steps: S1: Install a decision-making module, road test data collection and processing equipment, and an execution module on the test vehicle.

[0023] S2: Based on the road test data acquisition and processing equipment, collect position and motion data of the test vehicle and background objects.

[0024] In this method, the communication protocols and contents of satellite positioning data and vehicle signal data required by the system are defined according to application requirements, which facilitates data transmission and data recognition by the decision-making module.

[0025] The differential satellite positioning data protocol is $KSXT, where (satellite time), (longitude), (latitude), (altitude), (azimuth), (pitch), (velocity angle), (velocity), (roll), (satellite positioning status), (satellite orientation status), (number of available satellites for front antenna), (number of available satellites for rear antenna), (eastward position coordinates), (northward position coordinates), (skyward position coordinates), (eastward velocity), (northward velocity), (skyward velocity), and 0FFFFFFFF (checksum bit). The vehicle signal data protocol is hexadecimal, two bytes long, containing the following information: seatbelt (0-not fastened, 1-fastened), engine status (0-off, 1-started), left turn (0-off, 1-on), right turn (0-off, 1-on), parking brake (0-released, 1-engaged), service brake (0-no signal, 1-signal present), clutch (0-no signal, 1-signal present), and auxiliary brake (0-no signal, 1-signal present). The second character is from low to high: Neutral (000), 1st gear (001), 2nd gear (010), 3rd gear (011), 4th gear (100), 5th gear (101), 6th gear (110), and Reverse (111).

[0026] After acquiring the motion data of the test vehicle and the background object, the decision module calculates the relative distance and relative speed between the test vehicle and the background object at the current vehicle speed, and determines whether the safety boundary has been reached. Once the safety boundary is reached and the triggering condition is met, an execution command is generated and pushed to the execution module to achieve vehicle braking.

[0027] The execution instruction format of the decision module is as follows: $DTER, Exam Item, Scenario Type, Indicator A Type, Indicator A Value, Indicator B Type, Indicator B Value, Indicator C Type, Indicator C Value, Reference Evaluation Result, Execution Requirements, *END. The data frame header is "$DTER", the end character is "*END", and each data bit is separated by a comma.

[0028] The "Scenario Type" data field is used by the system to determine the scenario type based on the current test route, test item, and driving environment. Since safety boundary conditions differ across scenarios, each message structure reserves three evaluation indicator information fields, including three indicator type data fields and three indicator value data fields; the last two evaluation result information fields refer to the evaluation results and execution requirements.

[0029] Scenario types: {Following another vehicle; other road users intruding; Crossing a pedestrian crossing / bus stop; Passing a school; Going straight at an intersection; Turning at an intersection; Changing lanes; Overtaking}.

[0030] Indicator types: {Forward relative distance; Left relative distance; Right relative distance; Test vehicle speed; Speed ​​difference with traffic participants; Acceleration of traffic participants; Type of traffic participants}.

[0031] Reference evaluation result: {Pass; Fail}. Implementation requirements: {None; Braking and deceleration}.

[0032] S3: Calculate the relative position and relative speed between the test vehicle and the background objects at the current vehicle speed; If there are background objects in front of the test vehicle in the same lane, proceed to step S4; If the test vehicle changes lanes, proceed to step S5; Otherwise, repeat steps S2 to S3.

[0033] In calculating the relative position and relative speed between the test vehicle and background objects at the current speed, the lidar acquires information about objects around the test vehicle, calculates the distance between the lidar and each object, and obtains the three-dimensional coordinates of each point relative to the lidar. Without losing environmental information, based on key parameters, a filtering algorithm is used to filter out traffic information of non-key targets, retaining information about key targets. Key parameters include: object shape, size, and location; key target information includes: background objects in front of the vehicle in the same lane and background objects behind the vehicle in the adjacent lane; non-key targets include: environmental elements and background objects in non-adjacent lanes, with environmental elements including: trees, houses, and green belts; this application's technical solution collects data on the test vehicle's own operation and the surrounding traffic environment, compares it with preset safety boundary conditions, and determines whether the vehicle is in a safe operating state. For dangerous driving situations, it supports automatic braking intervention to avoid accidents or reduce their severity. It also supports data transmission to the testing system to achieve intelligent evaluation of some manually deducted points.

[0034] In step S3, the relative position and relative speed between the test vehicle and the background object at the current vehicle speed are calculated, which specifically includes the following steps.

[0035] b1: Point cloud data is generated based on the three-dimensional information of the vehicle's surroundings collected by LiDAR. Each point in the point cloud data includes: the point's spatial coordinates x, y, z, and reflection intensity value.

[0036] like Figure 6As shown, during a road test, a single frame of LiDAR point cloud data contains approximately 30,000 points. However, to represent the trajectories of the test vehicle and background objects on a high-precision map, many frames are required. Furthermore, to achieve real-time computation, the computation speed must be increased. To ensure the practicality of this method and to implement it even on test vehicles with lower performance hardware, data filtering is necessary to remove redundant data and retain only valid data, thereby reducing computational load. During computation, data from non-target areas is removed without losing environmental information, leaving only the point cloud of the target objects. Filtering objects include trees and lampposts along the roadside; reducing the number of point clouds shortens system processing time and reduces system load.

[0037] b2: Perform the first filtering on the point cloud data to remove interference in the test road section environment and obtain the data after the first filtering. Specifically, the following steps are included: b21: Construct an interference filtering model based on the types of interference present in the test route; The interference objects in this method refer to fixed-position objects outside the road surface, such as trees and streetlights. During driving tests, fixed-position objects typically do not intentionally cause accidents. Therefore, this application primarily addresses anomalies related to common road objects within the test lane, such as people, vehicles, and large animals. If fixed objects on the test road affect driving, they will also be identified as background objects with a speed of 0 and processed accordingly. Assisted braking methods for objects outside the lane are handled using other methods and are not within the scope of this method.

[0038] b22: Using the test vehicle as the center, specify the filtering radius and construct the filtering range; The filtering range is adaptively set according to different exam types. This method only retains data on common background objects on the road within the filtering range.

[0039] The types of interference that may appear on the test route are limited, so in practical applications, a interference filtering model is constructed for each type of interference based on a hexahedron. The model size (side length of the hexahedron) and the filtering range (with the test vehicle as the origin of the coordinate system) are key parameters for point cloud filtering. The interference filtering models include: lamppost models, roadside shrub models, tree models, building models, etc.; for example, the length, width, and height of the roadside shrub model are set to (0.5, 0.5, 0.5), the length, width, and height of the lamppost model are (0.15, 0.15, 10), and the filtering range is set to a spherical range with a radius of 30m.

[0040] b23: Based on the interference filtering model, data outside the filtering range and interference within the filtering range are removed to obtain the data after the first filtering.

[0041] Specifically, such as Figure 7 As shown, Figure 6 The image shows the result after the data has undergone the first filtering.

[0042] When removing interference, the scanned point cloud data is clustered based on existing clustering algorithms to obtain a set of suspected interference. Each set of suspected interference is then compared in size and shape with all interference filtering models. If the similarity meets a preset threshold range, it is identified as interference of the corresponding type and removed.

[0043] b3: Perform a second filtering on the point cloud data to remove point cloud data containing interference objects on the road surface in the examination room, and obtain the data after the second filtering; Point clouds acquired by LiDAR are composed of a large number of points. To reconstruct an object from these scattered points, aggregation is required based on the position and distance relationships of the points. However, in point cloud data, there is an intersection between ground points and target object points, necessitating the segmentation and filtering of the ground point cloud before reconstructing the object outline.

[0044] In this method, the location of the lidar is known. A reference plane is set based on the lidar. Then, by setting a minimum position, point clouds located below the lidar and with a vertical distance greater than the minimum position value are considered as objects that will not pose a danger to the operation of the test vehicle, such as road edge stones on the ground and point cloud data of the ground itself. At the same time, a maximum position is set, and data located above the lidar and with a vertical distance greater than the maximum position value are considered as objects such as trees and buildings that were not cleared in step b2.

[0045] Specifically, the following steps are included: b31: Create a spatial reference plane: Using the lidar of the test vehicle as the origin of the coordinate system, simulate a reference plane parallel to the ground; b32: Read the distance threshold preset in the system; The distance threshold includes: the highest position above the lidar and the lowest position below the lidar; b33: Read the data after the first filtering and calculate the distance from each point to the spatial reference plane; Points exceeding the highest and lowest positions are filtered out to obtain the data after the second filtering.

[0046] Specifically, such as Figure 8 As shown, Figure 7 The result of the data after the second filtering. Assuming the highest position is set to 3m and the lowest position is set to 1.5m, that is, taking the reference plane where the lidar is located as the standard, all point cloud data that are 3m higher than the reference plane are deleted; all data that are 1.5m lower than the reference plane are cleared, thus completing the removal of ground and air interference.

[0047] b4: Generate a 3D point cloud view of the background objects in the examination room; Specifically, the following steps are included: b41: Read the preset clustering threshold, maximum point cloud count, and minimum point cloud count; The clustering threshold represents the distance between points within the same cluster; it is a range of values. The maximum and minimum point cloud counts indicate that only clusters with a point count between the maximum and minimum point cloud counts are considered valid clustering results. b42: Read the data after the second filtering, denoted as: data of points to be clustered; b42: Using a clustering threshold, calculate the distance between points in the data to be clustered to obtain a point cloud cluster set; b43: Judge all point cloud cluster sets, and release the points in the point cloud cluster sets whose number of points is between the maximum and minimum point cloud numbers, and record them as the data to be clustered; In a typical road test environment, the distance threshold is generally between 0 and 1m, the maximum number of point clouds in a cluster does not exceed the total number of point clouds, and the value is 1000; the minimum number of point clouds in a cluster needs to be set according to the target object, and the value is 20. b44: Repeat steps b42~b43 until the number of points in all the point cloud cluster sets obtained each time is between the maximum number of points and the minimum number of points. b45: Each point cloud cluster represents a 3D point cloud view of its corresponding background object.

[0048] like Figure 9 As shown, this is a 3D point cloud view after clustering. To make it clear, different point cloud clusters are marked with rectangles. Figure 10 This is a 3D point cloud view of the clustered test road sections.

[0049] b5: Compresses the 3D point cloud view of the background object into a 2D space; Specifically, the following steps are included: b51: Construct a plane parallel to the ground as a two-dimensional reference plane, which shall not be higher than the height of the test vehicle; b52: Obtain each point cloud cluster set, find the outermost point of each point cloud cluster set mapped onto the two-dimensional reference plane, and use it as the two-dimensional feature point of the corresponding background object. The remaining points are recorded as non-feature points. b53: Filter out all non-feature points and display the two-dimensional feature points of each background object on a two-dimensional plane to obtain a two-dimensional planar image of each background object.

[0050] b6: The 2D planar image of the background object, the test vehicle's positioning data, and the high-precision map are fused. 2D objects within the lane are considered background objects that could affect the test vehicle's safety and need to be included in the calculation. Objects outside the lane are filtered out. Combined with time information, the real-time relative positional relationship between the test vehicle and the background objects is obtained. By combining the ranging results from the short-range lidar and millimeter-wave radar installed on the test vehicle, the relative distance between the test vehicle and the background objects is calculated. See the specific rendering for reference. Figure 11 .

[0051] b7: Correlate the data of two consecutive frames of the background object along the time axis to obtain the continuous motion trajectory information of the background object, and then calculate the relative speed between the background object and the test vehicle. In actual calculation, after finding the background object involved in the calculation through b1~b7, the actual speed of the background object can be obtained by combining the short-range lidar and millimeter-wave radar installed on the test vehicle that can realize speed measurement. The relative speed can be obtained by calculating the actual speed of the background object and the speed of the test vehicle.

[0052] This method is a technical solution designed for road tests, focusing primarily on the situation in the lanes ahead and to the left and right. During data collection, the data undergoes two filtering operations to mask barriers, trees, and vehicles from other lanes, removing irrelevant data before subsequent calculations. This ensures that the point cloud recognition of this method more accurately identifies background objects (people or vehicles) that could affect the safety of the test vehicle. Compared to other methods, this method is simpler, more accurate, more efficient, and has a faster system response, making it particularly suitable for reducing vehicle accidents during road tests. Furthermore, the irrelevant data removal process is based entirely on data collected by radar sensors, eliminating the need for high-cost components. The subsequent calculations for safety boundaries are also less demanding, thus requiring lower hardware performance from the system. The relevant calculations can be completed using lower-cost radar, making it more suitable for use in cost-constrained testing systems.

[0053] like Figure 3 As shown, when an object appears in front of the test vehicle, the system assesses vehicle safety based on the relative distance and speed between the test vehicle and the background object. The system detects the real-time relative distance and speed between the background object and the test vehicle. When a dangerous situation arises between the test vehicle and the object, the system intervenes with braking to prevent an accident or mitigate its severity.

[0054] S4: Compare the speed V of the test vehicle. 考试车 Speed ​​V of the vehicle in front 背景物 ; If the following safety boundary conditions are met simultaneously, then proceed to step S7; V 考试车 >V 背景物 And Δt≤3s; Where △t is the estimated time for the test vehicle to collide with the background object in front, △t = △s / (V 考试车 -V 背景物 ), △s is the relative distance between the background object in the same lane ahead and the test vehicle; Otherwise, repeat steps S2 to S3.

[0055] like Figure 4 As shown, when the test vehicle changes lanes, the system identifies background objects behind the test vehicle in the adjacent lanes to the left and right.

[0056] S5: Detect the area behind the test vehicle in the left and right adjacent lanes. If there are background objects, proceed to step S6; otherwise, repeat steps S2 to S3.

[0057] S6: Execute step S7 when any one of the following safety boundary conditions is met: Condition 1: The relative distance Δs between the background object and the test vehicle is less than or equal to 3m; Condition 1 ensures that there is a background object behind the test vehicle in the adjacent lanes on the left and right, and the relative distance △s between the background object and the test vehicle is less than or equal to 3m, then no accident will occur; Condition 2: The relative distance Δs between the background object and the test vehicle satisfies: 3 < Δs ≤ 30m, and Δt ≤ 3s; Condition 2 ensures that there is a background object behind the test vehicle in the left and right adjacent lanes, and the relative distance between the background object and the test vehicle is between 3m and 30m. When the ratio of the relative distance between the background object and the test vehicle to the relative speed is less than 3s, an accident will not occur.

[0058] The safety boundary conditions in this application are threshold conditions derived from my country's road traffic safety laws and regulations and relevant policy documents of the Ministry of Public Security regarding driver's license examinations, combined with actual vehicle test results. The road driving skills test consists of 16 test items, including: vehicle preparation, starting, straight driving, gear shifting, lane changing, parking at the side of the road, driving straight through an intersection, turning left at an intersection, turning right at an intersection, crossing a pedestrian crossing, passing a school zone, passing a bus stop, meeting oncoming traffic, overtaking, U-turns, and night driving. Vehicle preparation is a test item performed when the vehicle is stationary and is unrelated to the traffic environment; the remaining items are performed while the test vehicle is in motion. Based on the safety officer's assessment of the test vehicle's operational safety during the road driving skills test, the system functions and safety thresholds are designed. The system detects positional relationships covering the lanes in front of the test vehicle and adjacent lanes to the left / right. The core parameters are mainly the longitudinal and lateral relative distances between the test vehicle and background objects; only the longitudinal distance is used within the same lane. Adjacent lanes are not necessarily longitudinal distances between the front and rear of the vehicle; in practical applications, lateral distance measurements are used in the calculation and correction of adjacent lane distance measurements. Based on common accident scenarios during road tests and the operational characteristics of road tests, this patent is applicable to hazard avoidance scenarios in three directions: front, left rear, and right rear of the test vehicle.

[0059] S7: Issues a vehicle braking command to the execution module.

[0060] This method also includes the following steps: a1: A positioning antenna is installed on the test vehicle to provide real-time location coordinates of the test vehicle; a2: such as Figure 5 As shown, the test vehicle's position coordinates and onboard signal status are collected based on the vehicle-mounted signal acquisition device and differential positioning equipment. The test vehicle's data at different times are superimposed on time to generate driving trajectory data. a3: By integrating driving trajectory data with high-precision electronic maps, vehicle movement data, the lane where the test vehicle is located, and the position of the vehicle body outline points are overlaid on the electronic map of the test route. Combined with traffic data collected from the map and test items, the operating status of the test vehicle is monitored in real time.

[0061] By using the technical solution of this invention, the identification and judgment of dangerous situations under typical road test scenarios can be realized. This method collects data on the operation of the test vehicle and the surrounding traffic environment, compares it with preset safety boundary conditions, and determines whether the vehicle is in a safe operating state. For dangerous driving situations, it supports intervention of auxiliary braking and pushes the data to the testing system, making road driving skills testing more intelligent and safer.

Claims

1. A road driving skills test intelligent assisted braking method based on environmental perception, characterized in that, It includes the following steps: S1: Install a decision-making module, road test data acquisition and processing equipment, and an execution module on the test vehicle; The road test data acquisition and processing equipment includes: lidar, millimeter-wave radar and short-range lidar installed on the vehicle; The lidar collects three-dimensional information around the vehicle, identifies surrounding background objects and traffic environment through calculation and processing, and outputs point cloud data, specifically including: object position, velocity and contour. The near-range lidar is used to collect distance and motion data of background objects in the lane where the test vehicle is located and in adjacent lanes; the installation positions of the near-range lidar include: the front left, front right, rear left, and rear right of the vehicle; The millimeter-wave radar is installed in front of and behind the vehicle to collect background objects and motion data in front of and behind the vehicle. The execution module brakes the test vehicle based on the instructions from the decision module; The decision-making module performs calculations based on the data collected by the road test data acquisition and processing equipment, identifies safety boundary conditions, and issues instructions to the execution module based on the calculation results; S2: Based on the road test data acquisition and processing equipment, collect the position and motion data of the test vehicle and background objects; S3: Calculate the relative position and relative speed between the test vehicle and the background objects at the current vehicle speed; If there are background objects in front of the test vehicle in the same lane, proceed to step S4; If the test vehicle changes lanes, proceed to step S5; Otherwise, repeat steps S2 to S3. S4: Compare the speed V of the test vehicle 考试车 Speed ​​V of the vehicle in front 背景物 ; If the following safety boundary conditions are met simultaneously, then proceed to step S7; V 考试车 >V 背景物 And Δt≤3s; Where △t is the estimated time for the test vehicle to collide with the background object in front, △t = △s / (V 考试车 -V 背景物 ), △s is the relative distance between the background object in the same lane ahead and the test vehicle; Otherwise, repeat steps S2 to S3. S5: Detect the area behind the test vehicle in the left and right adjacent lanes. If there are background objects, proceed to step S6; otherwise, repeat steps S2 to S3. S6: Execute step S7 when any one of the following safety boundary conditions is met: Condition 1: The relative distance Δs between the background object and the test vehicle is less than or equal to 3m; Condition 2: The relative distance Δs between the background object and the test vehicle satisfies: 3 < Δs ≤ 30m, and Δt ≤ 3s; S7: Issue a vehicle braking command to the execution module; The execution module includes: a mechanical structure and an electronic structure; The mechanical structure's execution module consists of a motor and a linkage mechanism, installed at the auxiliary brake pedal position on the test vehicle. After the decision module issues a braking command, the motor starts running and simulates manually pressing the auxiliary brake pedal through the linkage mechanism to brake the test vehicle. In step S3, the relative position and relative speed between the test vehicle and the background object at the current vehicle speed are calculated, specifically including the following steps: b1: Based on the three-dimensional information around the vehicle collected by the lidar, point cloud data is formed; Each point in the point cloud data includes: the point's spatial coordinates x, y, z, and reflection intensity value; b2: Perform a first filter on the point cloud data to remove interference in the test road section environment and obtain the data after the first filter; Specifically, the following steps are included: b21: Construct an interference filtering model based on the types of interference present in the test route; b22: Using the test vehicle as the center, specify the filtering radius and construct the filtering range; b23: Based on the interference filtering model, remove the data outside the filtering range and the interference within the filtering range to obtain the data after the first filtering. Each set of suspected interference objects is compared with all the interference object filtering models in terms of size and shape. If the similarity meets the preset threshold range, it is judged as the corresponding type of interference object and removed. b3: Perform a second filtering on the point cloud data to remove the interfering point cloud data of the test road surface, and obtain the data after the second filtering; Specifically, the following steps are included: b31: Create a spatial reference plane: Using the lidar of the test vehicle as the origin of the coordinate system, simulate a reference plane parallel to the ground; b32: Read the distance threshold preset in the system; The distance threshold includes: the highest position above the lidar and the lowest position below the lidar; b33: Read the data after the first filtering and calculate the distance from each point to the spatial reference plane; Points exceeding the highest and lowest positions are filtered out to remove ground and air interference, thus obtaining the second filtered data. b4: Generate a 3D point cloud view of the background objects within the test route; Specifically, the following steps are included: b41: Read the preset clustering threshold, maximum point cloud count, and minimum point cloud count; The clustering threshold represents the distance between points within the same cluster, and is a range value. The maximum and minimum point cloud counts indicate that only clusters with a number of points between the maximum and minimum point cloud counts are considered valid clustering results. b42: Read the data after the second filtering, and record it as: data of points to be clustered; b42: Using the clustering threshold, calculate the distance between points in the data to be clustered to obtain a point cloud cluster set; b43: For all the point cloud cluster sets, determine the points in the point cloud cluster sets whose number of points is between the maximum and minimum number of point clouds, and release them as the data to be clustered; b44: Repeat steps b42~b43 until the number of points in all the point cloud cluster sets obtained each time is between the maximum number of points and the minimum number of points. b45: Each of the point cloud clusters represents a three-dimensional point cloud view of the corresponding background object; b5: Compress the three-dimensional point cloud view of the background object into a two-dimensional space; Specifically, the following steps are included: b51: Construct a plane parallel to the ground as a two-dimensional reference plane, wherein the two-dimensional reference plane is not higher than the height of the test vehicle; b52: Obtain each point cloud cluster set, find the outermost point of each point cloud cluster set mapped on the two-dimensional reference plane, and use it as the two-dimensional feature point of the corresponding background object. The remaining points are recorded as non-feature points. b53: Filter out all the non-feature points and display the two-dimensional feature points of each background object on a two-dimensional plane to obtain a two-dimensional planar image of each background object. b6: Integrate the 2D plan view of the background object, the test vehicle positioning data, and the high-precision map. Objects within the lane are marked as background objects for calculation, while background objects outside the lane are deleted as interference. By combining the time information, the real-time relative positional relationship between the test vehicle and the background object can be obtained, and then the relative distance between the test vehicle and the background object can be calculated. b7: Correlate the data of the two frames before and after the background object with time as the axis to obtain the continuous motion trajectory information of the background object, and then calculate the relative speed between the background object and the test vehicle.

2. The intelligent assisted braking method for road driving skills testing based on environmental perception as described in claim 1, characterized in that: It also includes the following steps: a1: A positioning antenna is installed on the test vehicle to provide real-time location coordinates of the test vehicle; a2: Collect the location coordinates and on-board signal status of the test vehicle, and overlay the data of the test vehicle at different times with time as the axis to generate driving trajectory data; a3: Integrate driving trajectory data with high-precision electronic maps, overlay vehicle motion data, the lane where the test vehicle is located, and the position of the vehicle body outline points on the electronic map of the test route, and combine the traffic data collected by the map with the test items to monitor the operation status of the test vehicle in real time. The vehicle motion status data includes: vehicle speed, driving direction, acceleration, deceleration or constant speed. The test vehicle's operating status includes: the test vehicle's location and driving behavior, wherein the driving behavior includes: lane changing, lane crossing, and turning.

3. The intelligent assisted braking method for road driving skills testing based on environmental perception as described in claim 1, characterized in that: The road test data acquisition and processing equipment also includes: an on-board signal acquisition device and a differential positioning device; The differential positioning device includes a positioning base station and a positioning mobile station, used to collect the location information of the test vehicle, and calculate the current vehicle motion data and operating status by combining the current GPS position and direction with a pre-loaded vehicle contour model. The on-board data acquisition unit includes the vehicle's ECU, which is used to analyze and convert vehicle signals.

Images