Outdoor robot road surface evaluation method

Abstract

The application provides an outdoor robot road surface evaluation method, first, the laser radar x-axis is ensured to form a certain angle with the horizontal ground, then a plane alpha is obtained by the points on the ground shot by the 3D laser radar, the plane alpha in the ideal case is obtained from the geometric relationship between the 3D laser radar and the robot * ; a straight line b is obtained by the point cloud on the ground shot by the 2D laser radar, the straight line b in the ideal case is obtained from the geometric relationship between the 2D laser radar and the robot * ; then, the road condition evaluation is realized through logical judgment. The application reuses the existing 3D laser radar for mapping and the 2D laser radar sensor for obstacle detection on the vehicle body, uses the 3D and 2D laser radars of the laser radar to perceive the road condition information of the target area in the robot coordinate system, reduces the detection blind area and improves the detection accuracy through a certain sensor layout method, greatly reduces the robot falling risk, and improves the safety performance of the robot in the outdoor complex scene.

Description

Technical Field

[0001] This invention relates to the field of robotics, specifically to an outdoor robot road surface evaluation method. Background Technology

[0002] With the implementation of robotics technology, a large number of robots are being used in outdoor scenarios such as inspection and patrol, garbage collection, washing and sweeping, and epidemic prevention and disinfection. Since outdoor scenarios have more complex road conditions than indoor scenarios, such as slopes, openwork manhole covers, steps, and roadside curbs, how to ensure the stable and safe operation of robots in the above complex scenarios has become an important indicator for measuring the safety performance of a robot. Among them, the reliability evaluation of road conditions is the key to ensuring this indicator.

[0003] In existing technologies, there are two main methods for evaluating outdoor robot road surfaces:

[0004] One approach is to acquire point cloud data within the target area; classify the point cloud data to obtain classification results; determine the drop environment point cloud data based on the classification results; acquire historical drop environment point cloud data; generate a local map based on the drop environment point cloud data and historical drop environment point cloud data; and determine the robot's travel path based on the local map.

[0005] Another approach is to utilize radar 3D point cloud information and apply it to the road conditions of the target area to be detected based on the robot's chassis coordinate system through spatial coordinate transformation. By assessing the road conditions in this area, the robot can pass through the area, thereby reducing the risk of the robot falling.

[0006] The two methods mentioned above can provide some fall protection, but their low reliability in complex scenarios can cause significant errors in the characteristics of the target area perceived by the 3D point cloud. This can lead to the failure of functions such as fall protection and road surface perception, resulting in the robot going out of control. Summary of the Invention

[0007] To address the problems of existing technologies, this invention provides an outdoor robot road surface evaluation method. By reusing the vehicle's existing 3D LiDAR for mapping and 2D LiDAR for obstacle detection, the 3D and 2D LiDAR sensors are used to perceive road condition information of the target area in the robot's coordinate system. By employing a specific sensor layout method, the detection blind spot is reduced and the detection accuracy is improved, greatly reducing the risk of robot falls and enhancing the safety performance of the robot operating in complex outdoor scenarios.

[0008] This invention provides an outdoor robot road surface evaluation method, comprising the following steps:

[0009] 1) Install a lidar. Here, the x-axis of the 3D lidar forms a certain angle θ with the horizontal ground, and some point clouds are scattered on the ground; the x-axis of the 2D lidar forms a certain angle with the horizontal plane In the case of no occlusion, the point clouds within a certain forward field of view will fall on the ground;

[0010] 2) Obtain the plane α through the points of the 3D lidar that hit the ground. The plane α in the ideal case obtained from the geometric relationship between the 3D lidar and the robot * ;

[0011] 3) Obtain the line b through the point clouds of the 2D lidar that fall on the ground. The line b in the ideal case obtained from the geometric relationship between the 2D lidar and the robot * ;

[0012] 4) Achieve the evaluation of the road condition through the following logical judgment:

[0013]

[0014] When Jump to step a. The road condition ahead is poor and impassable;

[0015] When Jump to step b. Use α * As the plane equation, judge the positional relationship between the line b and the plane α * :

[0016] When |d| < T, the line is on the plane, and it is determined that the road condition is good and normal passage is possible;

[0017] When |d| > T and d < 0, the line is below the plane, and it is determined that there is a risk of road surface drop and passage is prohibited;

[0018] When |d| > T and d > 0, the line is above the plane, and it is determined that there is a road surface protrusion and passage is prohibited / slowed down;

[0019] When Jump to step c. Use b * As the line equation, judge the positional relationship between the line b * and the plane α:

[0020] When |d| < T, the line is on the plane, and it is determined that the road condition is good and normal passage is possible;

[0021] When |d| > T && d < 0, the line is below the plane, and it is determined that there is a risk of road surface drop and passage is prohibited;

[0022] When |d| > T && d > 0, the line is above the plane, and it is determined that there is a road surface protrusion and passage is prohibited / slowed down;

[0023] When Jump to step d to determine the positional relationship between line b and plane α:

[0024] When |d| < T, b lies on a, and at this time, it is necessary to further judge the relationship between α and α * :

[0025] When the distance |d| between α and α * is less than T, it is determined that the road condition is good and normal passage is allowed;

[0026] When the distance |d| between α and α * is greater than T && d < 0, it is determined that there is a risk of road surface drop and passage is prohibited;

[0027] When the distance |d| between α and α * is greater than T && d > 0, it is determined that the road condition is good, it is determined that there is a protrusion on the road surface, and passage is prohibited / slowed down;

[0028] When |d| > T and d < 0, the line is below the plane, and it is determined that there is a risk of road surface drop and passage is prohibited;

[0029] When |d| > T and d > 0, the line is above the plane, and it is determined that there is a protrusion on the road surface, and passage is prohibited / slowed down;

[0030] In the above formula, T is the distance threshold between the line and the plane equation, and d is the distance between the line and the plane;

[0031] 5) Feed back the evaluated effect to the application layer, and the application layer performs comprehensive processing in combination with the perception and road surface evaluation results.

[0032] The plane equation α in step 2) is obtained through the following steps:

[0033] 2.1) Record the current iteration number M = 0, randomly select m points from n radar point cloud data, and use these m points to calculate and fit a plane α M , if fitting fails, reselect the point cloud;

[0034] 2.2) Traverse the remaining point cloud x ∈ (n - m), and judge whether x is on the plane α M , if not satisfied, re-traverse the remaining point cloud, if satisfied, add it to the plane inlier list;

[0035] 2.3) Judge whether all the point clouds have been traversed. If not satisfied, return to step 2.2). If satisfied, record the plane equation α M and the corresponding inlier number x M to a list L a ;

[0036] 2.4) Record the iteration number M = M + 1, and judge whether M > K is satisfied. If satisfied, then in L aFind the plane equation α with the most interior points; if it is not satisfied, return to step 2.2.

[0037] The equation of the line b in step 3) is obtained through the following steps:

[0038] 3.1) Record the iteration number M = 0 at this point. Randomly select m points from the n radar point cloud data and use these m points to calculate and fit a straight line b. M If a fit cannot be achieved, a new point cloud should be selected.

[0039] 3.2) Traverse the remaining point cloud x∈(nm), and determine whether x lies on the line b. M If the condition is not met, the remaining point cloud is re-traversed; if the condition is met, the point is added to the list of points within the line.

[0040] 3.3) Determine if all point clouds have been traversed. If not, return to step 3.2). If so, record the plane equation b. M and the corresponding number of inner palaces y M To a list L b middle;

[0041] 3.4) Record the iteration number M = M + 1, and determine whether M > K is satisfied. If satisfied, record the plane equation b. M and the corresponding number of inner palaces y M To a list L b If the condition is not met, return to step 3.2).

[0042] The selection of the iteration number k is based on the following: each iteration uses n points to calculate the model. When the iteration number k is large enough, the probability of obtaining the correct model from the sampled n points is (1-ω). n ) k The probability of success becomes very small. The higher the confidence level, the more likely it is that at least one set of correct model parameters can be obtained in k iterations. The probability of success in at least one of the k sampling iterations is 1-(1-ω). n ) k Then the value of k that satisfies a certain success rate z is the iteration termination condition we need. The loop exits when the actual number of iterations exceeds k.

[0043] The beneficial effects of this invention are as follows:

[0044] 1. This method reuses existing sensors, reducing the design and installation costs of additional sensors;

[0045] 2. This method has high versatility and provides excellent reference for sensor layout;

[0046] 3. This method cleverly adds 2D point cloud data to the road surface detection, increasing the detection area and reducing blind spots;

[0047] 4. This method is a supplement to the obstacle detection function;

[0048] 5. This method has a drop protection function. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a schematic diagram of the lidar layout;

[0051] Figure 2 A flowchart for solving the plane equation;

[0052] Figure 3 This is a flowchart for solving the equation of a straight line. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] 1. First, the installation of the robot's 3D and 2D LiDARs needs to meet certain design requirements. For example, the radar installation... Figure 1 As shown, the 3D radar's x-axis forms a certain angle θ with the horizontal ground, and the scattered point cloud on the ground is used for ground detection. The 2D lidar's x-axis forms a certain angle with the horizontal plane. In unobstructed conditions, point clouds within a certain field of view (FOV) ahead will fall onto the ground. These conditions are prerequisites for effective road condition assessment.

[0055] 2. Extracting the plane equation

[0056] The plane equation α is obtained by using a 3D lidar to illuminate the ground, and the calculation process follows... Figure 2 The algorithm flow described.

[0057] 2.1) Record the iteration number M = 0 at this point. Randomly select m points from the n radar point cloud data, and use these m points to calculate and fit a plane α. M If a fit cannot be achieved, a new point cloud should be selected.

[0058] 2.2) Traverse the remaining point cloud x∈(nm), and determine whether x lies in plane α. M If the condition is not met, the remaining point cloud is re-traversed; if the condition is met, the point is added to the planar intrapoint list.

[0059] 2.3) Determine if all point clouds have been traversed. If not, return to step 2.2). If so, record the plane equation α. M and the corresponding number of interior points x M To a list L a middle;

[0060] 2.4) Record the iteration count M = M + 1, and determine whether M > K is satisfied. If satisfied, then in L... a Find the plane equation α with the most interior points; if it is not satisfied, return to step 2.2.

[0061] The selection of the number of iterations is based on the following: each calculation uses n points in the model. When the number of iterations k is sufficiently large, (1-ω n ) k The probability becomes very small. The higher the confidence level, the more likely it is that at least one set of correct model parameters can be obtained in k iterations. The probability of at least one successful sampling in k iterations is 1-(1-ω). n ) k Then the value of k that satisfies a certain success rate z is the iteration termination condition we need. The loop exits when the actual number of iterations exceeds k.

[0062] 3 Extracting the equation of the straight line

[0063] Similar to the method for extracting plane equations, according to Figure 3 The described algorithm process obtains the equation of the line b.

[0064] 3.1) Record the iteration number M = 0 at this point. Randomly select m points from the n radar point cloud data and use these m points to calculate and fit a straight line b. M If a fit cannot be achieved, a new point cloud should be selected.

[0065] 3.2) Traverse the remaining point cloud x∈(nm), and determine whether x lies on the line b. M If the condition is not met, the remaining point cloud is re-traversed; if the condition is met, the point is added to the list of points within the line.

[0066] 3.3) Determine if all point clouds have been traversed. If so, record the plane equation b. M and the corresponding number of inner palaces y M To a list L b middle;

[0067] 3.4) Record the iteration number M = M + 1, and determine whether M > K is satisfied. If it is satisfied, record the plane equation b M and the corresponding number of inner halls y M into a list L b If not, return to step 3.2).

[0068] 4. The following is to realize the evaluation of the road condition through the following logical judgments:

[0069]

[0070] When

[0071] The road condition ahead is poor and impassable.

[0072] When Use α * as the plane equation to judge the positional relationship between the straight line b and the plane α * :

[0073] When |d| < T, the straight line is on the plane, and it is determined that the road condition is good and normal passage is allowed;

[0074] When |d| > T and d < 0, the straight line is below the plane, and it is determined that there is a risk of road surface drop and passage is prohibited;

[0075] When |d| > T and d > 0, the straight line is above the plane, and it is determined that there is a protrusion on the road surface and passage is prohibited / slowed down.

[0076] When Use b * as the straight line equation to judge the positional relationship between the straight line b * and the plane α:

[0077] When |d| < T, the straight line is on the plane, and it is determined that the road condition is good and normal passage is allowed;

[0078] When |d| > T && d < 0, the straight line is below the plane, and it is determined that there is a risk of road surface drop and passage is prohibited;

[0079] When |d| > T && d > 0, the straight line is above the plane, and it is determined that there is a protrusion on the road surface and passage is prohibited / slowed down.

[0080] When Judge the positional relationship between the straight line b and the plane α:

[0081] When |d| < T, b is on a. At this time, it is necessary to further judge the relationship between α and α * :

[0082] When the distance |d| between α and α * < T, it is determined that the road condition is good and normal passage is allowed

[0083] When α and α * If distance |d|>T&&d<0, it is determined that there is a risk of falling off the road surface, and passage is prohibited;

[0084] When α and α * If distance |d|>T&&d>0, the road condition is determined to be good; if there is a bump in the road surface, passage is prohibited / reduced.

[0085] If |d|>T and d<0, and the straight line is below the plane, it is determined that there is a risk of the road surface falling, and passage is prohibited.

[0086] If |d|>T and d>0, and the straight line is above the plane, it is determined that there is a protrusion on the road surface, and passage is prohibited / reduced.

[0087] Finally, the evaluation results are fed back to the application layer, which then processes them in combination with the perception and road surface evaluation results.

[0088] In the above formula:

[0089]

[0090]

[0091] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. In particular, for the device embodiments, the above descriptions are merely preferred embodiments of the present invention. Since they are fundamentally similar to the method embodiments, the descriptions are relatively simple, and relevant parts can be referred to the descriptions of the method embodiments. The above descriptions are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention, without departing from the principle of the present invention, should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for evaluating outdoor robot road surfaces, characterized in that... Includes the following steps: 1) Install lidar, including 3D lidar. The axis forms a certain angle with the horizontal ground. Some point clouds are scattered on the ground; 2D lidar The axis forms a certain angle with the horizontal plane. In the absence of obstruction, point clouds within a portion of the field of view in front will fall onto the ground; 2) Obtain a plane by using points projected onto the ground by 3D lidar. The ideal plane obtained from the geometric relationship between 3D LiDAR and the robot. ; 3) Obtain straight lines using the point cloud data of 2D LiDAR on the ground. A straight line in an ideal condition, obtained from the geometric relationship between 2D LiDAR and the robot. ; 4) The road condition status is evaluated through the following logical judgments: ； when && Jump to step a, the road conditions ahead are poor and impassable; when && Jump to step b, and determine the straight line. and plane Positional relationship: When T < T, the straight line lies on the plane, indicating good road conditions and normal passage is possible; If T and d < 0, and the straight line is below the plane, it is determined that there is a risk of the road surface falling, and passage is prohibited; If T and d > 0, and the straight line is above the plane, it is determined that there is a protrusion on the road surface, and passage is prohibited / reduced. when && Jump to step c, determine the straight line. and plane Positional relationship: When T < T, the straight line lies on the plane, indicating good road conditions and normal passage is possible; If T && d < 0, the straight line is below the plane, indicating a risk of the road surface falling, and passage is prohibited; If T && d > 0, and the straight line is above the plane, it is determined that there is a protrusion on the road surface, and passage is prohibited / reduced. when && Jump to step d and determine the straight line. and plane Positional relationship: < T time, At point a, further judgment is needed. and Relationship: when and Distance | d * |< T * The road conditions were determined to be good, and passage was permitted. when and Distance | d * |> T * && d * If the value is less than 0, it indicates a risk of falling off the road surface, and passage is prohibited. when and Distance | d * |> T * && d * > 0, indicating a bump in the road surface, prohibit / reduce speed when proceeding; If T and d < 0, and the straight line is below the plane, it is determined that there is a risk of the road surface falling, and passage is prohibited; If T and d > 0, and the straight line is above the plane, it is determined that there is a protrusion on the road surface, and passage is prohibited / reduced. In the above formula, T is the threshold distance between the line and the plane, and d is the distance between the line and the plane; 5) Feedback the evaluation results to the application layer, which then combines the perception and road surface evaluation results for comprehensive processing.

2. The outdoor robot road surface evaluation method according to claim 1, characterized in that: Step 2) The plane Obtained through the following steps: 2.1) Record that the iteration number M=0 at this point. Randomly select m points from the n radar point cloud data, and use these m points to calculate and fit a plane. If a fit cannot be achieved, a new point cloud should be selected. 2.2) Traverse the remaining point cloud Determine whether x lies in the plane. If the condition is not met, the remaining point cloud is re-traversed; if the condition is met, the point is added to the planar intrapoint list. 2.3) Determine if all point clouds have been traversed. If not, return to step 2.2). If so, record the plane. and the corresponding number of interior points To a list middle; 2.4) Record the current iteration number M = M + 1, select the iteration number k for the termination condition, and determine whether M > k is satisfied. If satisfied, then... Find the plane with the most interior points If the condition is not met, return to step 2.

2.

3. The outdoor robot road surface evaluation method according to claim 1, characterized in that: Step 3) The straight line b is obtained through the following steps: 3.1) Record that the iteration number M=0 at this point. Randomly select m points from the n radar point cloud data, and use these m points to calculate and fit a straight line. If a fit cannot be achieved, a new point cloud should be selected. 3.2) Traverse the remaining point cloud Determine if x lies on the line If the condition is not met, the remaining point cloud is re-traversed; if the condition is met, the point is added to the list of points within the line. 3.3) Determine if all point clouds have been traversed. If not, return to step 3.2). If so, record the line. and the corresponding number of interior points To a list middle; 3.4) Record the current iteration number M = M + 1, select the iteration number k for the termination condition, and determine whether M > k is satisfied. If satisfied, then... Find the line b with the most interior points; if this condition is not met, return to step 3.

2.

4. The outdoor robot road surface evaluation method according to claim 2 or 3, characterized in that: The selection of the iteration number k is based on the following: each calculation model uses n points, and when the iteration number... When the sampled n points are large enough, the probability of obtaining the correct model can be calculated. It will become very small; the higher the confidence level, the more iterations... At least one set of correct model parameters can be obtained in k samplings, and the probability of success in at least one of the k samplings is: Then a certain success rate is satisfied. of The value is the iteration termination condition we need. When the actual number of iterations is greater than When that happens, exit the loop.

Images