Active safety control method for unmanned vehicle in unstructured environment

Abstract

The invention provides an active safety control method for an unmanned vehicle in an unstructured environment. The method comprises the following steps: acquiring millimeter wave radar detection data and laser radar detection data in a driving process of the unmanned vehicle, and preprocessing the millimeter wave radar detection data and the laser radar detection data; carrying out target detection and matching based on the preprocessed millimeter wave radar detection data and laser radar detection data, and judging whether a target obstacle exists or not; if the target obstacle exists, the relative speed of the target obstacle and the unmanned vehicle and a maneuvering driving path area of the unmanned vehicle are obtained, and a risk area is delimited based on the relative speed and a safe distance threshold value in the maneuvering driving path area; obstacle channel filtering is carried out on the risk area to obtain obstacle information in the risk area; and if the target obstacle exists in the risk area, emergency braking is carried out. The problem that in the prior art, an active safety control method of the unmanned vehicle in a non-structural environment is prone to causing false detection, and consequently practicability is poor is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned driving, and in particular relates to an active safety control method for an unmanned vehicle in an unstructured environment. Background Art

[0002] With the advancement of science and technology, the emergence of ground unmanned platforms has attracted widespread attention from the world. Among them, unmanned vehicles are the core platforms of ground unmanned platforms. Their application environment is expanding from ideal low-intensity environments to complex high-intensity environments, and the reliability and safety requirements of ground unmanned platforms are also increasing.

[0003] Most of the existing unmanned vehicles are based on remote control technology to achieve a platform application mode with remote control as the main and autonomy as the auxiliary. In the remote control system, the operator sends control instructions to control the unmanned platform to maneuver according to the environmental image information sent back by the unmanned vehicle. However, in actual use, due to the distortion when the camera collects the surrounding image information, it is difficult to achieve a one-to-one restoration of the environment using image processing technology; at the same time, it is difficult for the operator to accurately grasp the distance to the obstacle based on the image information alone, which can easily lead to misjudgment and create the risk of collision. Methods based on wireless communication have large bandwidth limitations and transmission delays. Large delays will affect the safety of unmanned vehicles;

[0004] As the demand for autonomous driving of unmanned vehicles increases, the active safety maneuvering technology of unmanned vehicles has been developed to a certain extent in the relatively ideal urban structured environment. Sensors can be used to detect moving vehicles and pedestrians. However, in unstructured environments, there are many interference factors such as weeds, dust, and branches, which can easily cause sensor misdetection and frequent emergency braking, resulting in low practicality of the active safety system. Summary of the invention

[0005] In view of the above analysis, the present invention aims to provide an active safety control method for unmanned vehicles in unstructured environments, which is used to solve the problem that the active safety control method for unmanned vehicles in unstructured environments in the prior art is prone to false detection, resulting in poor practicality.

[0006] The purpose of the present invention is mainly achieved through the following technical solutions:

[0007] On the one hand, the present invention provides an active safety control method for an unmanned vehicle in an unstructured environment, the method comprising:

[0008] Acquire the millimeter wave radar detection data and lidar detection data of the unmanned vehicle during driving and perform preprocessing;

[0009] Perform target detection and matching based on the preprocessed millimeter-wave radar detection data and lidar detection data to determine whether there are target obstacles;

[0010] If there is a target obstacle, the relative speed between the target obstacle and the unmanned vehicle and the maneuvering path area of ​​the unmanned vehicle are obtained, and the risk area is delineated based on the relative speed and the safety distance threshold in the maneuvering path area;

[0011] Obstacle channel filtering is performed on the risk area to obtain obstacle information in the risk area; if there is a target obstacle in the risk area, emergency braking is performed.

[0012] Further, the millimeter wave radar detection data is preprocessed, including:

[0013] Extracting reflection intensity information from the millimeter-wave radar detection data, filtering out detection data with reflection intensity less than a first threshold, and obtaining valid millimeter-wave radar detection data;

[0014] An ID is created for each valid millimeter-wave radar detection data and cycle monitoring is performed. The detection data that persists in a preset number of cycles is used as pre-processed millimeter-wave radar detection data for target detection and matching.

[0015] Further, the laser radar detection data is preprocessed, including: continuously reducing the voxel volume from near to far, and downsampling the environmental point cloud data obtained by the laser radar detection;

[0016] Based on the downsampled environmental point cloud data, target detection is performed using the following method:

[0017] Based on the downsampled environmental point cloud data, grid division is performed with a preset size, and obstacle judgment is performed on each grid through the height threshold method. The grids with obstacles are marked as 1, and the grids without obstacles are marked as 0. The preprocessed lidar target detection data converted from three-dimensional space to two-dimensional plane is obtained.

[0018] Furthermore, downsampling the environmental point cloud data obtained by the laser radar detection includes:

[0019] Loading the environmental point cloud data, and taking the position of the laser radar as a reference point, calculating the distance of each point in the point cloud data relative to the reference point;

[0020] According to the distance of each point relative to the reference point, the voxel size of each position in the point cloud space is obtained based on a preset decreasing function; and according to the voxel size, a voxel grid is created in the point cloud space;

[0021] The point cloud data in each voxel is aggregated based on the voxel grid to obtain downsampled point cloud data.

[0022] Furthermore, the following method is used to perform target matching based on the preprocessed millimeter wave radar detection data and the laser radar detection data to determine whether there is an obstacle:

[0023] If there is a laser radar detection target within the preset range around the millimeter wave radar detection target, it is determined that there is a possible obstacle at that location; the possible obstacle includes a target obstacle that needs to be avoided and a pseudo obstacle that does not need to be avoided;

[0024] If the size of the possible obstacle is greater than the second threshold, it is determined to be a target obstacle;

[0025] If the size of the possible obstacle is not greater than the second threshold, the millimeter-wave radar reflection intensity is judged; if the millimeter-wave radar reflection intensity is greater than the third threshold, the possible obstacle is determined to be a target obstacle; if the millimeter-wave radar reflection intensity is not greater than the third threshold, the possible obstacle is determined to be a pseudo-obstacle.

[0026] Furthermore, based on the curvature information fed back by the chassis of the unmanned vehicle, the maneuvering path area of ​​the unmanned vehicle is obtained by the following formula:

[0027] (rw / 2) 2 <x 2 +y 2 <(r+w / 2) 2 ;

[0028] Where (x, y) is the coordinate point in the maneuvering area, r is the turning radius, which is the inverse of the curvature, and w is the width of the unmanned vehicle.

[0029] Further, the risk area includes a collision warning area and a collision area; the collision warning area and the collision area are obtained based on the relative speed and different safety distance thresholds;

[0030] If there is a target obstacle in the collision warning area, the unmanned vehicle brakes in a speed percentage braking manner;

[0031] If there is a target obstacle in the collision area, the unmanned vehicle will use maximum acceleration braking to perform emergency braking.

[0032] Furthermore, the performing obstacle channel filtering on the risk area includes:

[0033] The continuous existence time of the target obstacle in the collision warning area and the collision area is monitored. If the continuous existence time is greater than the time threshold, it is determined that there is a target obstacle in the area, and braking is performed through the corresponding braking method.

[0034] Furthermore, the collision warning area is obtained by the following formula:

[0035] l pre =dis safe +v 2 / 2a;

[0036] w pre =w / 2+0.5;

[0037] The collision area is obtained by the following formula:

[0038] l col =1.0+v 2 / 2a;

[0039] w col =w / 2;

[0040] Among them, dis safe is the safety distance threshold of the collision warning area, v is the current relative speed, l pre is the distance of the collision warning area of ​​the center line of the maneuvering area in front of the unmanned vehicle, which is the arc length distance, w pre is the left and right width of the unmanned vehicle collision warning area, w is the width of the unmanned vehicle, and a is the maximum braking acceleration; l col is the distance of the collision area from the center line of the maneuvering area in front of the unmanned vehicle, also the arc length distance, w col It is the left and right width distance of the collision area of ​​the unmanned vehicle. The safety distance threshold of the collision area is 1.0.

[0041] Furthermore, the unmanned vehicle decelerates by speed percentage braking, including:

[0042] Based on the current speed of the unmanned vehicle, generating a desired speed with a preset percentage factor;

[0043] Controlling the unmanned vehicle to decelerate and travel at a desired speed based on a preset interval period;

[0044] The desired speed is expressed as:

[0045] v desired =α·v cur ;

[0046] Among them, v desired is the expected speed, v cur is the current speed, and α is the percentage coefficient.

[0047] Beneficial effects of this technical solution:

[0048] 1. The present invention solves the motor driving path area according to the vehicle driving state, and divides the collision warning area and the collision area based on the motor driving path area, instead of using simple linear braking, to ensure that the vehicle can still achieve road following performance during braking. At the same time, the collision area is adaptively adjusted by channel filtering, which reduces the amount of data processing, has good stability and real-time performance, and improves real-time performance while reducing hardware configuration requirements;

[0049] 2. The present invention proposes a millimeter wave radar detection data preprocessing method that combines reflection intensity filtering and cyclic period monitoring filtering, which effectively obtains effective data of the millimeter wave radar and improves the usability of the millimeter wave radar in unstructured environments; and designs a two-dimensional mapping method based on voxel smoothing downsampling and height threshold statistics, which can effectively reduce the density of close-range point clouds and retain long-range point cloud information. On the basis of converting three-dimensional point clouds into two-dimensional information with less data volume. By integrating millimeter wave radar and laser radar, the accuracy of obstacle detection in unstructured scenes is improved, and the practicality of the invention is enhanced. It does not rely on a good environment in an urban environment, can adapt well to complex and changeable unstructured scenes, and reduces the probability of false detection.

[0050] Other features and advantages of the present invention will be described in the following description, and part of them will become obvious from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The accompanying drawings are only used for the purpose of illustrating specific embodiments and are not to be considered as limiting the present invention. In the entire drawings, the same reference symbols represent the same components;

[0052] Figure 1 is a flow chart of an active safety control method for an unmanned vehicle in an unstructured environment according to an embodiment of the present invention;

[0053] Figure 2 is a schematic diagram of cycle monitoring filtering of millimeter wave radar detection data according to an embodiment of the present invention;

[0054] Figure 3 is a schematic diagram of constructing a two-dimensional map of laser radar detection results according to an embodiment of the present invention;

[0055] Figure 4 is a schematic diagram of obstacle detection in a typical dust-raising environment according to an embodiment of the present invention;

[0056] Figure 5 is a schematic diagram of millimeter wave radar detection data and laser radar detection data in a typical scene of weeds and branches in an embodiment of the present invention;

[0057] Figure 6 is a schematic diagram of a maneuvering driving path area of ​​an unmanned vehicle according to an embodiment of the present invention;

[0058] Figure 7 It is a schematic diagram of risk area division according to an embodiment of the present invention. DETAILED DESCRIPTION

[0059] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the implementation cases of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0060] An embodiment of the present invention provides an active safety control method for an unmanned vehicle in an unstructured environment. Figure 1 As shown, the method comprises the following steps:

[0061] Step S1: obtaining millimeter wave radar detection data and laser radar detection data during the driving process of the unmanned vehicle, and performing preprocessing;

[0062] Specifically, in order to achieve active safety control of unmanned vehicles in unstructured environments, this embodiment improves the accuracy of obstacle detection in unstructured scenarios by integrating millimeter-wave radar and lidar. The millimeter-wave radar and lidar in this embodiment are arranged at the front end of the unmanned vehicle with the upper and lower central axes aligned, so as to obtain millimeter-wave radar detection data and lidar detection data for obstacle detection.

[0063] Among them, the millimeter-wave radar uses sensors to emit electromagnetic waves to scan the road, and obtains information such as target distance and azimuth by performing signal analysis on the received reflected waves, thereby obtaining the original millimeter-wave radar detection data. However, in an unstructured off-road environment, due to interference from weeds, dust, branches, etc., the detection data has a lot of noise and cannot be used directly. Specific preprocessing methods are required to extract valid data. This embodiment designs a millimeter-wave radar detection data preprocessing method that combines reflection intensity filtering and cycle monitoring filtering to filter out noise data, such as Figure 2 As shown, it includes: extracting reflection intensity information from the millimeter-wave radar detection data, filtering out detection data with reflection intensity less than a first threshold, and obtaining valid millimeter-wave radar detection data; establishing an ID for each valid millimeter-wave radar detection data and performing cycle monitoring, and using the detection data that persists in a preset number of cycles as pre-processed millimeter-wave radar detection data for target detection and matching;

[0064] More specifically, the reflection intensity of the millimeter-wave radar used in this embodiment is -1 to 6 levels, representing the millimeter-wave reflection intensity from weak to strong. First, the reflection intensity information in the millimeter-wave radar detection data is extracted, the validity of the reflection intensity is judged, and the target with a reflection intensity of -1 is filtered out; then, the remaining millimeter-wave radar detection data is continuously tracked and recorded through cycle monitoring, and a corresponding ID number is established for each target. If a certain data continues to exist in 5 cycle periods, it is considered to be obstacle target data, which can be used as pre-processed millimeter-wave radar monitoring data for subsequent target detection and matching.

[0065] Furthermore, the laser radar uses sensors to scan the driving road to obtain environmental point cloud information. In an unstructured environment, the road surface does not have the standardization of civilian application scenarios, so the road surface is usually difficult to extract. This embodiment designs a two-dimensional mapping method based on a combination of voxel smoothing downsampling and height threshold statistics to convert the three-dimensional point cloud into two-dimensional information with a small amount of data for fusion matching detection with millimeter-wave radar detection data. Since the laser beam gradually becomes sparse from near to far, in order to retain valid data as much as possible while reducing the amount of data, this embodiment uses voxel smoothing to preprocess the data. First, the environmental point cloud data is loaded, and the location of the laser radar is used as a reference point to calculate the distance of each point in the point cloud data relative to the reference point; according to the distance of each point relative to the reference point, the voxel size of each position in the point cloud space is obtained based on a preset decreasing function; preferably, in this example, based on the actual application scenario, only the point cloud information within 50 meters is considered, and the decreasing function is set to Among them l p represents the side length of the voxel grid at point p, represents the maximum value of the side length of the voxel grid, which is 0.2 meters in this embodiment. represents the minimum value of the side length of the voxel grid, which is 0.1 m in this embodiment. p Represents the distance of point p relative to the reference point. Based on the processed environmental point cloud data, according to the voxel size, a voxel grid is created in the point cloud space; based on the voxel grid, the point cloud data in each voxel is aggregated to obtain the downsampled point cloud data. This embodiment uses the downsampling method to ensure that the number of high-density point clouds in the near distance is effectively reduced, and the low-density point cloud data in the far distance is effectively retained. Based on the downsampled point cloud data, the grid is divided into a preset size, and the height information of the point cloud data in the same grid is counted. The grid whose maximum height and minimum height difference is greater than the height threshold is judged to have an obstacle. Otherwise, there is no obstacle. This embodiment corresponds to a light unmanned platform (3 tons), so the set height threshold is 30 cm. The grid with obstacles is marked as 1, and the grid without obstacles is marked as 0, and the pre-processed laser radar detection data converted from three-dimensional space to two-dimensional plane is obtained, such as Figure 3 In particular, the grid size is set to 20 cm × 20 cm in this embodiment, and after obstacle marking, the laser radar target detection data converted from three-dimensional space to two-dimensional plane is obtained, which is convenient for subsequent final target detection and matching.

[0066] Step S2: Target detection and matching are performed based on the preprocessed millimeter wave radar detection data and the laser radar detection data to determine whether there is a target obstacle;

[0067] Specifically, the present embodiment determines whether there is a target obstacle by the following method: if there is a laser radar detection target within a preset range around the millimeter wave radar detection target, it is determined that there is a possible obstacle at the location; the possible obstacles include target obstacles that need to be avoided and pseudo obstacles that do not need to be avoided; if the size of the possible obstacle is greater than the second threshold, it is determined to be a target obstacle; if the size of the possible obstacle is not greater than the second threshold, the millimeter wave radar reflection intensity is determined; if the millimeter wave radar reflection intensity is greater than the third threshold, the possible obstacle is determined to be a target obstacle, and if the millimeter wave radar reflection intensity is not greater than the third threshold, the possible obstacle is determined to be a pseudo obstacle; from actual experience, it can be known that in a non-structural environment, obstacles with an actual size of no more than 60cm*60cm*60cm are usually obstacles such as weeds and branches that do not affect the driving of the vehicle. Therefore, the second threshold in the present embodiment is set to 60cm*60cm*60cm. The present embodiment uses a millimeter wave radar with a reflection intensity of -1 to 6 levels. According to actual detection experience, the third threshold of the millimeter wave radar reflection intensity is set to 0.

[0068] This embodiment matches the millimeter-wave radar detection results with the laser radar detection results, and further confirms the obstacle type and size to determine whether there is a target obstacle that needs to be avoided, avoiding sensor misdetection caused by many interference factors such as weeds, dust, branches, etc. in unstructured environments, and improving the actual application efficiency of the system. Since the millimeter-wave radar and the laser radar are installed in an aligned upper and lower central axis, there is no need for projection transformation on the two-dimensional map, and target matching can be performed directly. Taking into account the error factor, if there is a laser radar detection target within 40 cm around the target position in the millimeter-wave radar detection result, it is considered that there is an obstacle at that position.

[0069] In unstructured environments, millimeter-wave radars are generally difficult to detect dust interference targets. Even if dust information is detected occasionally in a frame of data, the aforementioned cycle monitoring and filtering method can effectively filter out such information and reduce false detections. Figure 4As shown, it can be seen that in the dusty environment, after being processed by this method, the lidar detects a large number of obstacles and gives obstacle prompts, but no obstacles appear in the millimeter-wave radar detection results in the area, so the dust targets are effectively filtered out.

[0070] For weeds and branches, this embodiment makes full use of the millimeter wave radar's sensitivity to metals, and filters them out by taking into account the weak reflection intensity of millimeter waves and the size of obstacles. Figure 5 As shown, it can be seen that a large number of obstacles such as weeds and branches appear in the scene area. The three-dimensional scanning of the obstacles by the lidar gives obstacle prompts, but the millimeter-wave radar is unstable and gives local detection results. At the same time, the millimeter-wave radar reflection intensity is -1 or 0, and the signal is weak, indicating that there may be obstacles. However, combined with the millimeter-wave radar reflection intensity, it can be judged that the obstacles at the millimeter-wave obstacle result do not meet the characteristics of hard obstacles (for hard obstacles, the millimeter-wave reflection intensity is higher). Obstacles are not shown in the detection results, and obstacles such as weeds and branches are effectively filtered out.

[0071] Step S3: If there is a target obstacle, the relative speed between the target obstacle and the unmanned vehicle and the maneuvering path area of ​​the unmanned vehicle are obtained, and the risk area is delineated based on the relative speed and the safety distance threshold in the maneuvering path area;

[0072] Specifically, after the obstacle is detected, the relative speed between the obstacle and the unmanned vehicle is first obtained based on the relative displacement and time difference between the two frames of data. Since the longitudinal speed is larger and the lateral speed is smaller during the vehicle's movement, the obtained speed is fitted to the longitudinal speed, as shown in the following formula:

[0073] v = Δl / Δt;

[0074] Among them, v represents the relative speed between the obstacle and the unmanned vehicle, Δl represents the relative displacement between two frames of data, and Δt represents the time difference between two frames of data.

[0075] Furthermore, combined with the curvature information 1 / r fed back by the chassis, the maneuvering path area of ​​the unmanned vehicle is obtained, such as Figure 6 As shown. Taking the current position of the unmanned vehicle as the coordinate origin, its maneuvering path area is expressed as:

[0076] (rw / 2) 2 <x 2 +y 2 <(r+w / 2) 2 ;

[0077] Among them, (x, y) is the coordinate point in the maneuvering area, r is the turning radius (the inverse of the curvature), and w is the width of the unmanned vehicle.

[0078] Based on the unmanned vehicle's maneuvering path area, the risk area is delineated according to the relative speed and safety distance threshold. This embodiment proposes a collision warning area and collision area division method based on the maneuvering path area. The risk area division result is as follows: Figure 7 As shown, the risk area is divided into a collision warning area and a collision area based on the relative speed and different safety distance thresholds, so that different braking methods can be adopted for different risk areas, reducing frequent emergency stops caused by non-structural environments and improving the safety of unmanned vehicle maneuvering.

[0079] Specifically, the collision warning area in the risk area defined in this embodiment is obtained by the following formula:

[0080] l pre =dis safe +v 2 / 2a;

[0081] w pre =w / 2+0.5;

[0082] The collision area is obtained by the following formula:

[0083] l col =1.0+v 2 / 2a;

[0084] w col =w / 2;

[0085] Among them, dis safe is the safety distance threshold of the collision warning area, v is the current relative speed, l pre is the distance of the collision warning area of ​​the center line of the maneuvering area in front of the unmanned vehicle, which is the arc length distance, w pre is the left and right width of the unmanned vehicle collision warning area, w is the width of the unmanned vehicle, and a is the maximum braking acceleration; l col is the distance of the collision area from the center line of the maneuvering area in front of the unmanned vehicle, also the arc length distance, w col It is the left and right width distance of the collision area of ​​the unmanned vehicle. The safety distance threshold of the collision area is 1.0.

[0086] Step S4: performing obstacle channel filtering on the risk area to obtain obstacle information in the risk area; if there is a target obstacle in the risk area, emergency braking is performed.

[0087] Specifically, after the risk area is delineated, the obstacle channel filtering is performed on the risk area by the following method:

[0088] The continuous existence time of the target obstacle in the collision warning area and the collision area is monitored. If the continuous existence time is greater than the time threshold (set to 0.2 seconds in this embodiment), it is determined that there is a target obstacle in the area, and braking is performed through the corresponding braking method.

[0089] Preferably, if there is a target obstacle in the collision warning area, the unmanned vehicle brakes using a speed percentage braking method; if there is a target obstacle in the collision area, the unmanned vehicle uses a maximum acceleration braking method for emergency braking.

[0090] In an unstructured environment, there are often uncertain factors such as curves and ups and downs. In order to improve the braking safety of unmanned vehicles and reduce the risk of rollover, this embodiment proposes a safe braking control method based on a combination of speed percentage braking and maximum acceleration braking; when the target obstacle is in the collision warning area, the unmanned vehicle will decelerate by speed percentage braking, that is, based on the current speed of the unmanned vehicle, the expected speed is generated by a preset percentage coefficient; based on a preset interval period, the unmanned vehicle is controlled to decelerate at the expected speed. In this embodiment, the percentage coefficient is set to 0.99 and the interval period is set to 100ms;

[0091] The expected speed is expressed as:

[0092] v desired =α·v cur ;

[0093] Among them, v desired is the expected speed, v cur is the current speed, and α is the percentage coefficient.

[0094] When an obstacle is within the collision zone, the unmanned vehicle will use maximum acceleration braking to perform emergency braking to ensure that the vehicle can stop safely.

[0095] In particular, the present invention is also provided with an unmanned vehicle remote control system, which sends an alarm signal to the unmanned vehicle remote control system when an obstacle is detected and when the obstacle is approaching, reminding the operator to pay attention to the risk of collision. If the operator does not take measures, the corresponding braking method is actively used to brake. The present invention combines the active safety control and remote control methods of unmanned vehicles, reduces the risks caused by non-intuitive operations based on the remote control system, reduces the driving burden of the operator, and improves the stability of the safe operation of the unmanned vehicle.

[0096] In summary, the active safety control method for unmanned vehicles in a non-structured environment provided by the embodiment of the present invention solves the maneuvering path area according to the vehicle driving state, and divides the collision warning area and the collision area based on the maneuvering path area, instead of using simple linear braking, to ensure that the vehicle can still achieve road following during braking. At the same time, the collision area is adaptively adjusted by channel filtering, which reduces the amount of data processing, has good stability and real-time performance, and improves the real-time performance while reducing the hardware configuration requirements; and the present invention proposes a millimeter wave radar detection data preprocessing method combining reflection intensity filtering and cycle monitoring filtering, effectively obtains millimeter wave radar effective data, and improves the usability of millimeter wave radar in unstructured environments; and designs a two-dimensional mapping method based on voxel smoothing downsampling and height threshold statistics, which can effectively reduce the density of short-range point clouds and retain long-range point cloud information. On the basis of converting three-dimensional point clouds into two-dimensional information with less data volume. By integrating millimeter wave radar and laser radar, the accuracy of obstacle detection in unstructured scenes is improved, and the practicality of the invention is enhanced. The invention does not rely on the good environment in the urban environment, can adapt well to complex and changeable unstructured scenes, and reduces the probability of false detection.

[0097] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0098] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. An active safety control method for an unmanned vehicle in an unstructured environment, characterized in that: include: Acquire the millimeter wave radar detection data and lidar detection data of the unmanned vehicle during driving and perform preprocessing; Perform target detection and matching based on the preprocessed millimeter-wave radar detection data and lidar detection data to determine whether there are target obstacles; If there is a target obstacle, the relative speed between the target obstacle and the unmanned vehicle and the maneuvering path area of ​​the unmanned vehicle are obtained, and the risk area is delineated based on the relative speed and the safety distance threshold in the maneuvering path area; Performing obstacle channel filtering on the risk area to obtain obstacle information in the risk area; If there is a target obstacle in the risk area, emergency braking is performed.

2. The active safety control method for unmanned vehicles in unstructured environments according to claim 1 is characterized in that: Preprocessing the millimeter wave radar detection data includes: Extracting reflection intensity information from the millimeter-wave radar detection data, filtering out detection data with reflection intensity less than a first threshold, and obtaining valid millimeter-wave radar detection data; An ID is created for each valid millimeter-wave radar detection data and cycle monitoring is performed. The detection data that persists in a preset number of cycles is used as pre-processed millimeter-wave radar detection data for target detection and matching.

3. The active safety control method for unmanned vehicles in unstructured environments according to claim 2 is characterized in that: Preprocessing the laser radar detection data includes: continuously reducing the voxel volume from near to far, and downsampling the environmental point cloud data obtained by the laser radar detection; Based on the downsampled environmental point cloud data, target detection is performed using the following method: Based on the downsampled environmental point cloud data, grid division is performed with a preset size, and obstacle judgment is performed on each grid through the height threshold method. The grids with obstacles are marked as 1, and the grids without obstacles are marked as 0, and the lidar target detection data converted from three-dimensional space to two-dimensional plane is obtained.

4. The active safety control method for unmanned vehicles in unstructured environments according to claim 3 is characterized in that: Downsampling the environmental point cloud data obtained by the laser radar detection includes: Loading the environmental point cloud data, and taking the position of the laser radar as a reference point, calculating the distance of each point in the point cloud data relative to the reference point; According to the distance of each point relative to the reference point, the voxel size of each position in the point cloud space is obtained based on a preset decreasing function; and according to the voxel size, a voxel grid is created in the point cloud space; The point cloud data in each voxel is aggregated based on the voxel grid to obtain downsampled point cloud data.

5. The active safety control method for unmanned vehicles in unstructured environments according to claim 4 is characterized in that: The following method is used to perform target matching based on the preprocessed millimeter wave radar detection data and the laser radar detection data to determine whether there is an obstacle: If there is a laser radar detection target within the preset range around the millimeter wave radar detection target, it is determined that there is a possible obstacle at that location; the possible obstacle includes a target obstacle that needs to be avoided and a pseudo obstacle that does not need to be avoided; If the size of the possible obstacle is greater than the second threshold, it is determined to be a target obstacle; If the size of the possible obstacle is not greater than the second threshold, the millimeter-wave radar reflection intensity is judged; if the millimeter-wave radar reflection intensity is greater than the third threshold, the possible obstacle is determined to be a target obstacle; if the millimeter-wave radar reflection intensity is not greater than the third threshold, the possible obstacle is determined to be a pseudo-obstacle.

6. The active safety control method for unmanned vehicles in unstructured environments according to claim 1, characterized in that: Based on the curvature information fed back by the chassis of the unmanned vehicle, the maneuvering path area of ​​the unmanned vehicle is obtained by the following formula: (r-w / 2) 2 <x 2 +y 2 <(r+w / 2) 2 ; Where (x, y) is the coordinate point in the maneuvering area, r is the turning radius, which is the inverse of the curvature, and w is the width of the unmanned vehicle.

7. The active safety control method for unmanned vehicles in unstructured environments according to claim 1, characterized in that: The risk area includes a collision warning area and a collision area; the collision warning area and the collision area are obtained based on the relative speed and different safety distance thresholds; If there is a target obstacle in the collision warning area, the unmanned vehicle brakes in a speed percentage braking manner; If there is a target obstacle in the collision area, the unmanned vehicle will use maximum acceleration braking to perform emergency braking.

8. The active safety control method for unmanned vehicles in unstructured environments according to claim 7, characterized in that: The performing obstacle channel filtering on the risk area includes: The continuous existence time of the target obstacle in the collision warning area and the collision area is monitored. If the continuous existence time is greater than the time threshold, it is determined that there is a target obstacle in the area, and braking is performed through the corresponding braking method.

9. The active safety control method for unmanned vehicles in unstructured environments according to claim 7, characterized in that: The collision warning area is obtained by the following formula: <h2 style=";text-align:left;direction:ltr">l<h2 style=";text-align:left;direction:ltr"> pre <h2 style=";text-align:left;direction:ltr"> =dis<h2 style=";text-align:left;direction:ltr"> safe <h2 style=";text-align:left;direction:ltr"> +v<h2 style=";text-align:left;direction:ltr"> 2 <h2 style=";text-align:left;direction:ltr"> / 2a; In pre =w / 2+0.5; The collision area is obtained by the following formula: l col =1.0+v 2 / 2a; In col =w / 2; Among them, dis safe is the safety distance threshold of the collision warning area, v is the current relative speed, l pre is the distance of the collision warning area to the center line of the maneuvering area in front of the unmanned vehicle, w pre is the left and right width of the unmanned vehicle collision warning area, w is the width of the unmanned vehicle, and a is the maximum braking acceleration; l col is the distance of the collision area from the center line of the maneuvering area in front of the unmanned vehicle, w col It is the left and right width distance of the collision area of ​​the unmanned vehicle. The safety distance threshold of the collision area is 1.

0.

10. The active safety control method for unmanned vehicles in unstructured environments according to claim 4, characterized in that: The unmanned vehicle decelerates by speed percentage braking, including: Based on the current speed of the unmanned vehicle, generating a desired speed with a preset percentage factor; Controlling the unmanned vehicle to decelerate and travel at a desired speed based on a preset interval period; The desired speed is expressed as: v desired =α·v cur ; Among them, v desired is the expected speed, v cur is the current speed, and α is the percentage coefficient.

Images