Multi-pose humanoid robot environment navigation method based on binocular and 3D point cloud

Through binocular depth camera and three-dimensional point cloud technology, combined with machine learning, the problem of humanoid robot path generation in complex and private environments is solved, and efficient navigation without additional hardware is achieved in complex environments, and is suitable for multi-pose humanoid robots.

CN117146829BActive Publication Date: 2024-11-08JIANGSU YUNMU ZHIZAO TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202311418390.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-10-30
Publication Date
2024-11-08
Estimated Expiration
2043-10-30

AI Technical Summary

Technical Problem

Existing navigation systems are difficult to provide precise paths in complex and private environments, especially for multi-pose humanoid robots. Traditional systems cannot effectively handle the situations where humanoid robots bent, squat and walk on the ground, and require additional hardware and a lot of data training. .

Method used

Using binocular depth camera and three-dimensional point cloud technology, scene reconstruction, point cloud preprocessing, path calculation and navigation matrix mapping are carried out through machine learning, and a multi-pose humanoid robot environment navigation method based on binocular and three-dimensional point cloud is designed to generate A feasible path for humanoid robots in complex environments.

Benefits of technology

It realizes the rapid generation of multi-pose humanoid robot paths without additional hardware in complex and private environments, which has good development and expansion, is suitable for a variety of scenarios, and improves navigation accuracy and efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117146829B_ABST
    Figure CN117146829B_ABST
Patent Text Reader

Abstract

The multi-posture humanoid robot environment navigation method based on binocular and three-dimensional point cloud includes the following steps: S1: input scene to reconstruct point cloud; S2: perform point cloud preprocessing, input the information of the humanoid robot itself after point cloud preprocessing, and calculate the area that the humanoid robot can pass through in combination with the information of the humanoid robot itself; S3: extract the point cloud of the ground, and sparse the point cloud, and map the point cloud into a navigation matrix; S4: realize the multi-posture humanoid robot complex environment navigation within the navigation matrix. The multi-posture humanoid robot environment navigation method based on binocular and three-dimensional point cloud of the present invention is reasonably designed. When the humanoid robot is only equipped with a binocular depth camera, machine learning and three-dimensional point cloud are used to solve the path generation problem of the multi-posture humanoid robot in a complex and private environment, and it has good development and expansibility.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of robot vision, and in particular relates to a multi-posture humanoid robot environment navigation method based on binoculars and three-dimensional point clouds. Background Art

[0002] In recent years, with the rapid development of artificial intelligence and the improvement of robot control systems through the collaboration of software and hardware, the research on humanoid robots has gradually emerged in academia and industry. Among them, navigation is a key issue in the research of humanoid robots.

[0003] Normally, we use the global positioning system (GPS) for navigation. This method is suitable for use in situations where accuracy is not high, and people are needed as decision makers to eliminate the errors caused by the navigation system. In more complex situations, GPS cannot provide accurate navigation, such as home environments, office environments, and other environments with high privacy. Although the inertial navigation system can determine the relative position in the current environment, its error will accumulate over time, and both GPS and inertial navigation systems require the installation of additional hardware, such as an inertial measurement unit or a locator.

[0004] Navigation systems can also be implemented using artificial intelligence and machine learning, but they need to be trained with a large amount of data to derive the corresponding path. This method can achieve good results in some scenarios, such as predicting traffic congestion on the road and helping users get the best path. Therefore, this method has been widely used in car navigation. However, in some small scenarios, there is not such a large amount of data to train the navigation system, and the navigation system currently used does not take into account the structural variability of humanoid robots. Humanoid robots can not only walk upright, but also bend over, squat, and even walk on the ground in the foreseeable future. In traditional navigation systems, paths that only people can pass, such as squatting or even lying on their stomachs, are usually not displayed.

[0005] Therefore, the present invention has developed a multi-posture humanoid robot environment navigation method based on binoculars and three-dimensional point clouds, which accurately describes a three-dimensional scene through three-dimensional point clouds, and uses machine learning and three-dimensional point clouds to provide a feasible method for solving the path generation problem of multi-posture humanoid robots. Summary of the invention

[0006] Purpose of the invention: In order to overcome the above shortcomings, the purpose of the present invention is to provide a multi-posture humanoid robot environment navigation method based on binocular and three-dimensional point cloud. The method has reasonable design and adopts machine learning and three-dimensional point cloud to solve the path generation problem of the multi-posture humanoid robot in a complex and private environment when the humanoid robot is only equipped with a binocular depth camera. The method has good development and expansibility and broad application prospects.

[0007] The objective of the present invention is achieved through the following technical solutions:

[0008] A multi-posture humanoid robot environment navigation method based on binoculars and three-dimensional point cloud comprises the following steps:

[0009] S1: input scene reconstruction point cloud;

[0010] S2: performing point cloud preprocessing. After the point cloud preprocessing, the information of the humanoid robot itself is input, and the area through which the humanoid robot can pass is calculated in combination with the information of the humanoid robot itself;

[0011] S3: extract the ground point cloud, sparse the point cloud, and map the point cloud into a navigation matrix;

[0012] S4: Implementing complex environment navigation of the multi-posture humanoid robot within the navigation matrix.

[0013] Furthermore, in the above-mentioned multi-posture humanoid robot environment navigation method based on binoculars and three-dimensional point clouds, the step S1 specifically includes the following contents:

[0014] S11: Input the 3D point cloud information of the entire scene;

[0015] S12: The entire scene is reconstructed in three dimensions. After obtaining the reconstructed point cloud of the scene, the point cloud coordinate system is corrected. The ground of the point cloud coordinate system is laid flat on the x- and y-axis planes, and the entire scene is located in the positive direction of the z-axis of the point cloud coordinate system. The ground of the real space coordinate system is also laid flat on the x- and y-axis planes, and the entire scene is located in the positive direction of the z-axis of the real space coordinate system. The coordinate origins of the point cloud coordinate system and the real space coordinate system are located at the geometric center of the ground. The maximum coordinate value of the x-axis of the point cloud coordinate system is plus or minus 1, and the other two coordinate axes are converted according to the scene ratio.

[0016] S13: Set the dimensions of the scene corresponding to the point cloud coordinate system in the x, y, and z directions respectively. , , , the coordinates of the point cloud coordinate system and the coordinates of the real space coordinate system have the following conversion relationship:

[0017] ;

[0018] ;

[0019] ;

[0020] in, , , are the x, y, and z coordinate points in the point cloud coordinate system, , , They are the x, y, and z coordinate points in the real space coordinate system.

[0021] Furthermore, in the above-mentioned multi-posture humanoid robot environment navigation method based on binocular and three-dimensional point cloud, the point cloud preprocessing of step S2 specifically includes the following contents: denoising and segmenting the point cloud, extracting the boundary of the segmented point cloud, the space outside the boundary and the object inside the boundary, inserting a number of points inside the object boundary to ensure that the distance between any two points inside the object is less than the distance that the humanoid robot can pass through the space; wherein, the distance between any two points inside the object is set to , The minimum distance a humanoid robot can pass through space. .

[0022] Furthermore, in the above-mentioned multi-posture humanoid robot environment navigation method based on binocular and three-dimensional point cloud, the step S2 of calculating the area through which the humanoid robot can pass includes the following contents: on the plane, the height that the humanoid robot can pass through is determined according to the different postures of the humanoid robot, and according to the different postures of the humanoid robot, the heights that the humanoid robot can pass through in a standing posture and a squatting posture are set to be , ,in, is the height of the humanoid robot itself, , are the linear coefficients of the height through which the humanoid robot can pass in standing posture and squatting posture and the height of the humanoid robot itself; the width through which the humanoid robot can pass in standing posture and squatting posture are set to be , .

[0023] Furthermore, in the above-mentioned multi-posture humanoid robot environment navigation method based on binocular and three-dimensional point cloud, the step S2 of calculating the area through which the humanoid robot can pass also includes the following content: when the humanoid robot is in a standing posture, at this time, the distance between the ground and the obstacle is , the distance between obstacles on the ground in the plane formed by the x and y axes Should meet: , ; When the humanoid robot is in a squatting position, the distance between the ground and the obstacle is , the distance between obstacles on the ground in the plane formed by the x and y axes Should meet: , .

[0024] Furthermore, in the above-mentioned multi-posture humanoid robot environment navigation method based on binocular and three-dimensional point cloud, the step S2 of calculating the area that the humanoid robot can pass through also includes the following content: when the humanoid robot is in a standing posture, at this time, the distance between the ground and the obstacle is the z-axis coordinate of the coordinate of the point cloud formed by the obstacle closest to the ground, the distance between the obstacles is the horizontal distance between the two nearest coordinate points of the point cloud of the two objects, and the boundary point cloud point set of the object is set to , , then the distance between the ground and the obstacle The calculation method is: ; At this time, set the point set of the boundary point cloud of another object to , , introducing intermediate variables , is the distance between all points in the point set of the two boundary point clouds. If there are n points in each point set of the two boundary point clouds, the calculated have The following formula is used to calculate all the points in the point sets of the two boundary point clouds:

[0025] ;

[0026] When the humanoid robot is in a squatting posture, the calculation method is the same as that when the humanoid robot is in a standing posture.

[0027] Furthermore, in the above-mentioned multi-posture humanoid robot environment navigation method based on binocular and three-dimensional point cloud, the step S2 of calculating the area through which the humanoid robot can pass also includes the following content: after screening out all paths that meet the above conditions, that is, paths through which the humanoid robot can pass in multiple postures, the paths are spliced ​​and sorted by priority to calculate the area through which the humanoid robot can pass.

[0028] Furthermore, in the above-mentioned multi-posture humanoid robot environment navigation method based on binocular and three-dimensional point clouds, step S3 includes the following content: the point cloud is processed again, at this time, only the point cloud of the plane determined by the x and y axes of the point cloud coordinate system is needed, that is, the point cloud corresponding to the ground, the coordinate corresponding to the z axis of these point clouds is 0, and the remaining ground point clouds are only the point clouds corresponding to the ground that the humanoid robot can pass through, and these point clouds are sparsely processed and all of them are mapped into a navigation matrix.

[0029] Furthermore, in the above-mentioned multi-posture humanoid robot environment navigation method based on binocular and three-dimensional point cloud, the step S3 also includes the following contents: determining the size of the navigation matrix according to the number of point clouds and the size of the maximum coordinate values ​​of the x and y axes, the area where the navigation matrix is ​​0 corresponds to no points in the point cloud coordinate system, and the area where the navigation matrix is ​​not 0 indicates that there are points, that is, the path that the humanoid robot can pass through, and the values ​​in the navigation matrix can be set to different values ​​according to the different path priorities; a new coordinate system is introduced here, the matrix coordinate system, the center of the navigation matrix is ​​set to the coordinate origin, and the x and y axes are the horizontal and vertical directions of the navigation matrix respectively; wherein, is the conversion factor between the point cloud coordinate system and the matrix coordinate system, , is the coordinate of the point cloud coordinate system, is the matrix coordinate system coordinate, is a rounding function. At this time, the following formula is used to complete the conversion from the point cloud coordinate system to the matrix coordinate system:

[0030] .

[0031] Compared with the prior art, the present invention has the following beneficial effects: the multi-posture humanoid robot environment navigation method based on binocular and three-dimensional point cloud described in the present invention is reasonably designed and directly processes the point cloud. Based on the continuous updating of the point cloud processing method in academia, the method has good development and expansibility; when the humanoid robot is only equipped with a binocular depth camera, no other hardware is required. After obtaining the current scene and the robot's own parameters, machine learning and three-dimensional point cloud are used to solve the path generation problem of the multi-posture humanoid robot in a complex and private environment; the algorithm performs specific processing on the point set, which is faster and saves performance, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a flow chart of the multi-posture humanoid robot environment navigation method based on binocular and three-dimensional point cloud according to the present invention;

[0033] Figure 2 A schematic diagram of taking the point with the shortest distance in calculating the distance between two obstacles in Embodiment 1 of the present invention;

[0034] Figure 3 A schematic diagram of separately calculating obstacle distances of an object in two directions according to Embodiment 1 of the present invention;

[0035] Figure 4 It is a schematic diagram of a navigation matrix according to Embodiment 1 of the present invention. DETAILED DESCRIPTION

[0036] The following will be attached Figure 1-4 , Embodiment 1 clearly and completely describes the technical solution of the present invention. Obviously, the described embodiment is only a part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the protection scope of the present invention.

[0037] The following embodiment 1 provides a multi-posture humanoid robot environment navigation method based on binoculars and three-dimensional point clouds.

[0038] Example 1

[0039] The multi-posture humanoid robot environment navigation method based on binocular and three-dimensional point cloud of Example 1 is intended to solve the problem of path generation of the multi-posture humanoid robot in a complex and private environment when the humanoid robot is only equipped with a binocular depth camera and does not require other hardware.

[0040] like Figure 1 As shown, the method comprises the following steps:

[0041] S1: Input scene to reconstruct point cloud, as follows:

[0042] S11: First input the 3D point cloud information of the entire scene.

[0043] S12: After the entire scene is reconstructed in three dimensions and the reconstructed point cloud of the scene is obtained, the point cloud coordinate system is corrected. The point cloud coordinate system ground plane lays the ground on the x- and y-axis planes, and the entire scene is located in the positive direction of the z-axis of the point cloud coordinate system. The real space coordinate system ground plane also lays the ground on the x- and y-axis planes, and the entire scene is located in the positive direction of the z-axis of the real space coordinate system. To facilitate subsequent calculations, the coordinate origins of the point cloud coordinate system and the real space coordinate system are located at the geometric center of the ground, the maximum coordinate value of the x-axis of the point cloud coordinate system is plus or minus 1, and the other two coordinate axes are converted according to the scene ratio.

[0044] S13: Set the dimensions of the scene corresponding to the point cloud coordinate system in the x, y, and z directions respectively. , , , then the coordinates of the point cloud coordinate system and the coordinates of the real space coordinate system have the following conversion relationship:

[0045] ;

[0046] ;

[0047] ;

[0048] in, , , are the x, y, and z coordinate points in the point cloud coordinate system, , , They are the x, y, and z coordinate points in the real space coordinate system.

[0049] S2: Perform point cloud preprocessing. After point cloud preprocessing, input the information of the humanoid robot itself, and calculate the area through which the humanoid robot can pass in combination with the information of the humanoid robot itself. The details are as follows (all the following formulas are in point cloud coordinate system unless otherwise specified):

[0050] S21: First, denoise and segment the point cloud. In order to simplify the subsequent calculations, the point cloud needs to be processed. The boundary of the segmented point cloud is extracted. The space is outside the boundary, and the object itself is inside the boundary. In many cases, if the object is too large, it is difficult to distinguish the object itself from the space. Therefore, a small number of points are inserted inside the object boundary. As long as the distance between any two points inside the object is less than the distance that the humanoid robot can pass through the space, such as Figure 2 shown.

[0051] Set the distance between any two points inside the object to , in The minimum distance a humanoid robot can pass through space. .

[0052] The reason why it is set to , because the insertion point is randomly inserted. If the value is too small, the inserted points cannot cover all areas of the object. If a more accurate insertion point algorithm is used, the algorithm itself will take too much time.

[0053] S22: Input the information of the humanoid robot. On the plane, the height that the humanoid robot can pass through is determined according to the different postures of the humanoid robot. According to the different postures of the humanoid robot, the heights that the humanoid robot can pass through in standing posture and squatting posture are set respectively. , ,in, is the height of the humanoid robot itself, , are the linear coefficients of the height through which the humanoid robot can pass in standing posture and squatting posture and the height of the humanoid robot itself; the width through which the humanoid robot can pass in standing posture and squatting posture are set to be , .

[0054] like Figure 3 As shown, using a chair as an example, the obstacle distances in two directions are calculated separately: after the width calculation is completed, the height is calculated within a specific width.

[0055] When the humanoid robot is in a standing position, considering the obstacles in the Z-axis direction, the distance between the ground and the obstacles is , the distance between obstacles on the ground in the plane formed by the x and y axes Should meet: , .

[0056] When the humanoid robot is in a squatting position, considering the obstacles in the Z-axis direction, the distance between the ground and the obstacles is , the distance between obstacles on the ground in the plane formed by the x and y axes Should meet: , .

[0057] When the humanoid robot is in a standing position, the distance between the ground and the obstacle is the z-axis coordinate of the point cloud formed by the obstacle that is closest to the ground. The distance between obstacles is the horizontal distance between the two nearest coordinate points of the point cloud of two objects. The boundary point cloud point set of the object is set as , , then the distance between the ground and the obstacle The calculation method is: .

[0058] At this time, set the point set of the boundary point cloud of another object to , , introducing intermediate variables , is the distance between all points in the point set of the two boundary point clouds. If there are n points in each point set of the two boundary point clouds, the calculated have The following formula is used to calculate all the points in the point sets of the two boundary point clouds:

[0059] .

[0060] , The calculations are performed in sequence, and the subsequent calculations will be performed only when the former is satisfied.

[0061] When the humanoid robot is in a squatting posture, the calculation method is the same as that when the humanoid robot is in a standing posture.

[0062] After calculating the paths that the humanoid robot can pass under multiple postures, the paths are spliced ​​and sorted by priority. For example, the priority of standing is greater than that of squatting. This sorting method can be adjusted according to the situation. At this time, navigation at different positions is achieved according to the coordinates of different positions in the point cloud coordinate system.

[0063] S3: Extract the ground point cloud, sparse the point cloud, and map the point cloud into a navigation matrix as follows:

[0064] S31: After all the paths that meet the above requirements are screened out, the point cloud is processed again. At this time, only the point cloud of the plane determined by the x and y axes of the point cloud coordinate system is needed, that is, the point cloud corresponding to the ground. The coordinates corresponding to the z axis of these point clouds are 0, and the remaining ground point clouds are only the point clouds corresponding to the ground that the humanoid robot can pass through. Because the z-axis coordinates of the remaining point clouds are all 0 at this time, all these point clouds can be mapped into a matrix. In order to prevent the point cloud data from being too large, these point clouds need to be sparsely processed, and then all these point clouds are mapped into a navigation matrix.

[0065] S32: Determine the size of the navigation matrix according to the number of point clouds and the maximum coordinate values ​​of the x and y axes. The area where the navigation matrix is ​​0 corresponds to no point in the point cloud coordinate system. The area where the navigation matrix is ​​not 0 indicates that there are points, that is, the path that the humanoid robot can pass. The values ​​in the navigation matrix can be set to different values ​​according to the different path priorities. Figure 4 As shown, an example of a navigation matrix: The following matrix, the bold part, is the x-axis horizontally and the y-axis vertically. 0 means it is not passable, 2 means the robot can pass by squatting, and 3 means it can pass by standing. Navigation can be performed within this matrix. Of course, the matrix and complexity in the real environment are much greater than this example.

[0066] Here, a new coordinate system, the matrix coordinate system, is introduced. The center of the navigation matrix is ​​set as the coordinate origin, and the x and y axes are the horizontal and vertical directions of the navigation matrix respectively. is the conversion factor between the point cloud coordinate system and the matrix coordinate system, , is the coordinate of the point cloud coordinate system, is the matrix coordinate system coordinate, is a rounding function. At this time, the following formula is used to complete the conversion from the point cloud coordinate system to the matrix coordinate system:

[0067] .

[0068] S4: Implementing complex environment navigation of the multi-posture humanoid robot within the navigation matrix.

[0069] The present invention has many specific application paths, and the above is only a preferred embodiment of the present invention. It should be pointed out that the above embodiments are only used to illustrate the present invention, and are not used to limit the protection scope of the present invention. For those of ordinary skill in the art, several improvements can be made without departing from the principle of the present invention, and these improvements should also be regarded as the protection scope of the present invention.

Claims

1. A multi-pose humanoid robot environment navigation method based on binocular and three-dimensional point cloud, It is characterized in that The steps include: S1: input scene reconstruction point cloud; S2: perform point cloud preprocessing. After point cloud preprocessing, input the information of the humanoid robot itself, and calculate the area that the humanoid robot can pass through in combination with the information of the humanoid robot itself; after calculating the paths that the humanoid robot can pass through in multiple postures, splice the paths and sort them by priority; S3: extract the ground point cloud, sparse the point cloud, and map the point cloud into a navigation matrix; S31: After all paths satisfying the above are screened out, the point cloud is processed again. At this time, only the point cloud of the plane determined by the x and y axes of the point cloud coordinate system is needed, that is, the point cloud corresponding to the ground. The coordinates corresponding to the z axis of these point clouds are 0, and the remaining ground point clouds are only the point clouds corresponding to the ground that the humanoid robot can pass through. The z axis coordinates of the remaining point clouds are all 0 at this time. All these point clouds are mapped into a matrix. In order to prevent the point cloud data from being too large, a sparse processing is performed on these point clouds, and then all these point clouds are mapped into a navigation matrix; S32: Determine the size of the navigation matrix according to the number of point clouds and the maximum coordinate values ​​of the x and y axes. The area where the navigation matrix is ​​0 corresponds to no point in the point cloud coordinate system. The area where the navigation matrix is ​​not 0 indicates that there are points, that is, the path that the humanoid robot can pass through. The values ​​in the navigation matrix can be set to different values ​​according to the priority of the path. Here, a new coordinate system, the matrix coordinate system, is introduced. The center of the navigation matrix is ​​set as the coordinate origin, and the x and y axes are the horizontal and vertical directions of the navigation matrix respectively. Among them, is the conversion factor between the point cloud coordinate system and the matrix coordinate system, , is the coordinate of the point cloud coordinate system, is the matrix coordinate system coordinate, is a rounding function. At this time, the following formula is used to complete the conversion from the point cloud coordinate system to the matrix coordinate system: ; S4: Implementing navigation of the multi-posture humanoid robot in a complex environment within the navigation matrix.

2. According to claim 1, the multi-posture humanoid robot environment navigation method based on binoculars and three-dimensional point cloud, It is characterized in that The step S1 specifically includes the following contents: S11: Input the 3D point cloud information of the entire scene; S12: The entire scene is reconstructed in three dimensions. After obtaining the reconstructed point cloud of the scene, the point cloud coordinate system is corrected. The ground of the point cloud coordinate system is laid flat on the x- and y-axis planes, and the entire scene is located in the positive direction of the z-axis of the point cloud coordinate system. The ground of the real space coordinate system is also laid flat on the x- and y-axis planes, and the entire scene is located in the positive direction of the z-axis of the real space coordinate system. The coordinate origins of the point cloud coordinate system and the real space coordinate system are located at the geometric center of the ground. The maximum coordinate value of the x-axis of the point cloud coordinate system is plus or minus 1, and the other two coordinate axes are converted according to the scene ratio. S13: Set the dimensions of the scene corresponding to the point cloud coordinate system in the x, y, and z directions respectively. , , , the coordinates of the point cloud coordinate system and the coordinates of the real space coordinate system have the following conversion relationship: ; ; ; in, , , are the x, y, and z coordinate points in the point cloud coordinate system, , , They are the x, y, and z coordinate points in the real space coordinate system.

3. According to claim 1, the multi-posture humanoid robot environment navigation method based on binoculars and three-dimensional point cloud, It is characterized in that The point cloud preprocessing of step S2 specifically includes the following contents: denoising and segmenting the point cloud, extracting the boundary of the segmented point cloud, where the space is outside the boundary and the object itself is inside the boundary, inserting a number of points inside the boundary of the object to ensure that the distance between any two points inside the object is less than the distance that the humanoid robot can pass through the space; wherein, the distance between any two points inside the object is set to , The minimum distance a humanoid robot can pass through space. .

4. The multi-posture humanoid robot environment navigation method based on binoculars and three-dimensional point cloud according to claim 3, It is characterized in that The step S2 of calculating the area through which the humanoid robot can pass includes the following contents: on the plane, the height at which the humanoid robot can pass is determined according to the different postures of the humanoid robot. According to the different postures of the humanoid robot, the heights at which the humanoid robot can pass in a standing posture and a squatting posture are set respectively. , ,in, is the height of the humanoid robot itself, , are the linear coefficients of the height through which the humanoid robot can pass in standing posture and squatting posture and the height of the humanoid robot itself; the width through which the humanoid robot can pass in standing posture and squatting posture are set to be , .

5. The multi-posture humanoid robot environment navigation method based on binoculars and three-dimensional point cloud according to claim 4, It is characterized in that The step S2 of calculating the area through which the humanoid robot can pass also includes the following contents: when the humanoid robot is in a standing position, at this time, the distance between the ground and the obstacle is , the distance between obstacles on the ground in the plane formed by the x and y axes Should meet: , ; When the humanoid robot is in a squatting position, the distance between the ground and the obstacle is , the distance between obstacles on the ground in the plane formed by the x and y axes Should meet: , .

6. The multi-posture humanoid robot environment navigation method based on binoculars and three-dimensional point cloud according to claim 5, It is characterized in that The step S2 of calculating the area that the humanoid robot can pass through also includes the following contents: when the humanoid robot is in a standing posture, at this time, the distance between the ground and the obstacle is the z-axis coordinate of the coordinate of the point cloud formed by the obstacle closest to the ground, the distance between the obstacles is the horizontal distance between the two nearest coordinate points of the point clouds of the two objects, and the boundary point cloud point set of the object is set to , , then the distance between the ground and the obstacle The calculation method is: ; At this time, set the point set of the boundary point cloud of another object to , , introducing intermediate variables , is the distance between all points in the point set of the two boundary point clouds. If there are n points in each point set of the two boundary point clouds, the calculated have The following formula is used to calculate all the points in the point sets of the two boundary point clouds: ; When the humanoid robot is in a squatting posture, the calculation method is the same as that when the humanoid robot is in a standing posture.

7. The multi-posture humanoid robot environment navigation method based on binoculars and three-dimensional point cloud according to claim 6, It is characterized in that The step S2 of calculating the area through which the humanoid robot can pass also includes the following contents: after screening out all paths that meet the above conditions, that is, paths through which the humanoid robot can pass in multiple postures, the paths are spliced ​​and sorted by priority to calculate the area through which the humanoid robot can pass.