Multi-lidar quadruped robot and point cloud data processing method

Abstract

The application discloses a multi-laser-radar quadruped robot and a point cloud data processing method, and comprises a body, two front leg members installed on both sides of the front part of the body, and two rear leg members installed on both sides of the rear part of the body; only a first laser radar is arranged forwardly on the front end mounting surface of the body, and the first laser radar has a first scanning area facing the front of the body; only a second laser radar is arranged rearwardly on the rear end mounting surface of the body, and the second laser radar has a second scanning area facing the rear of the body; the first scanning area and the second scanning area have an intersection line, which is located below the lowest part of the bottom of the body and above the highest part of the preset passable obstacle on the ground. Only two laser radars are needed to sense the front ground, the ground below the body and the rear ground area where the robot travels, the omnidirectional vision of the robot and the vision of the landing area are ensured through fewer sensors, the sensor cost is reduced, and the sensor layout space is reduced.

Description

Technical Field

[0001] This invention relates to the field of quadruped robot technology, and more particularly to a multi-laser radar quadruped robot and a point cloud data processing method. Background Technology

[0002] LiDAR, as a 3D sensor, has been widely used in robotics, especially in structured scenarios where a single low-cost LiDAR can meet many application requirements. The main purpose of developing quadruped robots is to enable their biomimetic leg structure to be used in both structured and unstructured scenarios. Complex scenarios place higher demands on the performance, layout, and usage of LiDAR sensors, where a single LiDAR often cannot meet the application requirements. Robot mapping and localization tasks require LiDAR to capture as much data as possible about the surrounding environment, while navigation tasks focus on detecting obstacles at close range. Perception-motion-control fusion focuses more on information about the landing points around the robot's legs. For example, existing technology discloses an intelligent quadruped robot (Chinese Patent Publication No. CN113184078A) that uses a combination of a depth camera mounted on the front of the robot and a LiDAR mounted on the back. However, this LiDAR mounted on top of the robot is difficult to use for terrain perception of the quadruped's legs, while the depth camera on the front is limited by accuracy and lighting conditions, failing to achieve ideal environmental perception. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a quadruped robot with multiple lidar sensors, comprising a body, two front legs mounted on both sides of the front of the body, and two hind legs mounted on both sides of the rear of the body.

[0004] The first laser radar is arranged only forward on the front mounting surface of the body, and the first laser radar has a first scanning area facing the front of the body.

[0005] The second lidar is arranged only rearward on the rear mounting surface of the body, and the second lidar has a second scanning area facing the rear of the body.

[0006] The first scanning area and the second scanning area have an intersecting line, which is located below the lowest point of the bottom of the machine body and above the highest point of a pre-set obstacle on the ground.

[0007] Preferably, the first lidar is mounted facing forward on the front mounting surface of the machine body and has a first mounting angle relative to the horizontal direction, the first mounting angle being -70 to 20 degrees.

[0008] Preferably, the second lidar located on the rear mounting surface of the machine body is installed at a downward angle, and the foremost edge of the area where the second scanning area intersects with the ground exceeds the foremost edge of the vertical projection of the machine body onto the ground.

[0009] Preferably, the second lidar is installed at a downward tilt and has a second installation angle with the horizontal direction, the second installation angle being -90 to -20 degrees.

[0010] Preferably, the multi-laser quadruped robot also has a third laser radar. The body includes a head, a body, and a tail connected in sequence. The third laser radar is installed on the upper side of the head, body, or tail and has a third scanning area facing upwards. The first scanning area, the second scanning area, and the third scanning area intersect each other and have a common field of view.

[0011] Preferably, the third lidar is mounted upwards on the upper mounting surface of the body and has a third mounting angle relative to the vertical direction, the third mounting angle being -20 to 20 degrees.

[0012] Preferably, the third scanning area does not contact the highest point of the top of the machine body, and the first scanning area, the second scanning area and the third scanning area are connected and surround the front, rear, top and bottom areas of the machine body.

[0013] This invention also discloses a point cloud data processing method, based on a multi-LiDAR quadruped robot as described in any of the foregoing descriptions, comprising the following steps:

[0014] The point cloud data generated by each lidar installed on the aircraft is collected, and the point cloud data is stitched together and converted into the aircraft coordinate system to form the first point cloud dataset.

[0015] The second point cloud dataset is formed by filtering out the body point cloud in the first point cloud dataset using a preset cuboid filtering algorithm.

[0016] The joint angle information of the two front legs and two hind legs of the current quadruped robot is obtained. After filtering out the body point cloud of the four legs below the knees from the second point cloud dataset using the leg point cloud filtering algorithm, the final environmental point cloud dataset is formed.

[0017] Preferably, the front leg and rear leg components adopt a three-joint foot structure, having a hip joint motor, a thigh joint motor, and a lower leg joint motor.

[0018] Preferably, the joint angle information of the two front legs and two hind legs of the current quadruped robot is obtained. After filtering out the body point clouds of the four legs below the knees from the second point cloud dataset using a leg point cloud filtering algorithm, the final environment point cloud dataset is formed, which specifically includes:

[0019] Obtain the rotation angles of each joint in the two front legs and two rear legs of the current quadruped robot, and calculate the knee control point q of each leg. k and ankle control point q a The coordinates in the body coordinate system are:

[0020]

[0021] Where q0 is the origin of the body coordinate system, This is the static variation matrix of the motor from the body to the hip joint. This represents the static variation matrix from the hip joint motor to the thigh joint motor. This represents the static transformation matrix from the thigh joint motor to the lower leg joint motor. This is the static variation matrix from the lower leg joint motor to the knee control point. The static change matrix from the lower leg joint motor to the ankle control point; This is the rotational variation matrix generated by the rotation of the hip joint motor. This is the rotational change matrix generated by the rotation of the thigh joint motor. The rotational change matrix generated by the rotation of the lower leg joint motor;

[0022]

[0023] Where a h a is the joint rotation angle of the hip joint motor. t a is the joint rotation angle of the thigh joint motor. c This refers to the joint rotation angle of the lower leg joint motor.

[0024] Based on the coordinates of the knee and ankle control points of each leg component, four approximate geometric bodies of the lower leg are constructed, and their geometric body region coordinates in the body coordinate system are obtained. After filtering out the point clouds located in the four geometric body regions from the second point cloud dataset based on the geometric body region coordinates, the final environmental point cloud dataset is formed.

[0025] Compared with the prior art, the beneficial effects of the present invention are:

[0026] This invention discloses a multi-LiDAR quadruped robot that can achieve complete perception of the ground in front of the robot, below the robot, and behind the robot using only a first LiDAR mounted forward on the front end and a second LiDAR mounted backward on the rear end, thus avoiding blind spots in areas such as below the robot. By using fewer sensors, the robot maintains a wide field of view in all directions and on its landing area, achieving a better balance between sensor cost and environmental perception. The dual-LiDAR ground detection scheme reduces the number of LiDARs while ensuring sufficient detection range and enabling the robot's mapping, localization, navigation, and terrain perception functions. This reduces sensor cost, space requirements, and overall robot weight.

[0027] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] To gain a more complete understanding of this application and its beneficial effects, the following description will be provided in conjunction with the accompanying drawings. In the following description, the same reference numerals denote the same parts.

[0030] Figure 1 This is a schematic diagram of the structure of a multi-laser radar quadruped robot disclosed in an embodiment of the present invention.

[0031] Figure 2 This is a schematic diagram of the LiDAR scanning area disclosed in an embodiment of the present invention.

[0032] Figure 3 This is a schematic diagram of the structure of a multi-laser radar quadruped robot disclosed in another embodiment of the present invention.

[0033] Figure 4 This is a schematic diagram of the lidar scanning area disclosed in another embodiment of the present invention.

[0034] Figure 5 This is a schematic diagram of the installation angle of a lidar according to another embodiment of the present invention.

[0035] Figure 6 This is a schematic diagram illustrating the specific process of a point cloud data processing method disclosed in another embodiment of the present invention.

[0036] Figure 7 This is a schematic diagram illustrating the specific process of leg coordinates disclosed in another embodiment of the present invention. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0038] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0039] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0040] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains. The terms “first,” “second,” and similar terms used in the specification and claims of this patent application do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms “an” or “a” and similar terms do not indicate a limitation of quantity, but rather indicate the presence of at least one.

[0041] As attached Figure 1 and 2As shown, this embodiment discloses a multi-LiDAR quadruped robot, including a body 1, two front legs 2 mounted on both sides of the front of the body, and two rear legs 3 mounted on both sides of the rear of the body. Only one first LiDAR 4 is arranged forward on the front mounting surface of the body 1, and this first LiDAR 4 has a first scanning area 41 facing forward of the body. Only one second LiDAR 5 is arranged rearward on the rear mounting surface of the body 1, and the second LiDAR 5 has a second scanning area 51 facing rearward of the body. The first scanning area 41 and the second scanning area 51 intersect at a line 61, which is located below the lowest point of the bottom of the body and above the highest point of a pre-set obstacle on the ground. Figure 2 As shown, the field of view of the head and tail lidars intersect at a line 61, which is located above the ground. To ensure that taller obstacles are also within the field of view, its height is set slightly higher than that of typical obstacles. The position and height of the intersection point are adjusted by adjusting the installation height and angle of the head and tail lidars. However, it is necessary to ensure a certain ground-level common field of view 13 for both the head and tail lidars to avoid blind spots in ground scanning. In addition, since the rear lidar is placed downwards and backwards, there is a certain risk of it touching the ground and causing sensor damage. Therefore, the lowest installation height of the rear lidar must ensure that the bottom of the lidar is higher than the bottom of the quadruped robot's torso.

[0042] In this embodiment, the body 1 can also be further divided into a fuselage and a head 7 and a tail 8 that are respectively connected to the front and rear ends of the fuselage, according to the front and rear positions. In this case, the first laser radar 4 is installed at the front end of the head and the second laser radar 5 is installed at the rear end of the tail.

[0043] In this embodiment, in the horizontal direction, upward angle is positive and downward angle is negative. In the vertical direction, rightward angle is positive and leftward angle is negative.

[0044] In this embodiment, the first lidar 4 is mounted forward on the front mounting surface of the body 1, and has a first mounting angle relative to the horizontal direction, which is -70 to 20 degrees. A preferred angle range is -20 to 20 degrees. Alternatively, a 0-degree angle is preferred, ensuring a small angle between the first lidar fixed to the head and the horizontal line, thus guaranteeing the quadruped robot's surrounding field of vision for mapping, localization, and navigation functions. Simultaneously, it provides a ground visibility range 11 for the head radar in the quadruped robot's landing area, which can be used for terrain perception and motion control functions.

[0045] Preferably, the second lidar 5 is installed at a downward tilt and has a second mounting angle with the horizontal direction, the second mounting angle being -90 to -20 degrees. A preferred angle is -70 degrees, which results in a larger angle between the second lidar fixed at the tail and the horizontal line, ensuring sufficient visibility of the quadruped robot's landing area and providing a ground visibility range 12 for the tail lidar in the quadruped robot's landing area.

[0046] In this embodiment, the second LiDAR 5, located on the rear mounting surface of the body 1, is installed at a downward angle, and the foremost point of the area where the second scanning area intersects with the ground exceeds the foremost point of the body's vertical projection on the ground. Specifically, the ground visibility range 12 formed by the second scanning area of ​​the tail LiDAR extends beyond the head of the quadruped robot, ensuring that even if the head LiDAR is damaged in extreme cases, the tail LiDAR can still maintain a relatively complete field of vision of the landing area, thus satisfying the quadruped robot's terrain perception, movement control, and other functions. At the same time, the tail LiDAR also provides a certain degree of surrounding vision, such as the area behind the tail, which can be used to satisfy the quadruped robot's mapping, localization, navigation, and other functions.

[0047] Furthermore, the area where the second scanning area 51 of the second lidar 5 intersects with the ground, i.e., the ground visibility range 12 and 13 of the tail lidar, covers the pre-landing range of the two front legs taking one step forward. Preferably, the area where the second scanning area intersects with the ground, i.e., the ground visibility range of the tail lidar, should cover the pre-landing range of the robot's front legs taking one step forward at the set maximum stride. This satisfies the quadruped robot's landing area vision and ensures that the tail lidar can also maintain the area's visibility even if the head lidar is damaged.

[0048] In this embodiment, the overlapping area of ​​the first scanning area 41 and the second scanning area 51 is located in front of the front leg, thereby ensuring sufficient field of view coverage. The first lidar 4 and the second lidar 5 have a hemispherical field of view of 360 degrees horizontally and 90 degrees vertically.

[0049] This embodiment discloses a multi-LiDAR quadruped robot that, by deploying only two LiDARs on its body—a first LiDAR facing forward on the front mounting surface and a second LiDAR facing backward on the rear mounting surface—achieves complete perception of the ground in front of the robot, below the robot, and behind it, avoiding blind spots such as those below the robot. Fewer sensors ensure comprehensive vision of the robot's surroundings and landing area, achieving a better balance between sensor cost and environmental perception. This dual-LiDAR ground detection scheme reduces the number of LiDARs while maintaining detection range and enabling the robot's mapping, localization, navigation, and terrain perception functions. This lowers sensor cost, reduces sensor layout space, and decreases the robot's weight.

[0050] In another embodiment, as shown in the appendix Figure 3 and 4 As shown, another type of multi-LiDAR quadruped robot is also disclosed. This quadruped robot includes the same structure as the previous embodiment, namely, a body 1, two front legs 2 mounted on both sides of the front of the body, and two rear legs 3 mounted on both sides of the rear of the body. Only one first LiDAR 4 is arranged forward on the front mounting surface of the body 1, and this first LiDAR 4 has a first scanning area 41 facing forward of the body. Only one second LiDAR 5 is arranged rearward on the rear mounting surface of the body 1, and this second LiDAR 5 has a second scanning area 51 facing rearward of the body. The first scanning area 41 and the second scanning area 51 intersect at a line 61, which is located below the lowest point of the bottom of the body and above the highest point of a pre-set passable obstacle on the ground. In addition, the multi-LiDAR quadruped robot disclosed in this embodiment also has a third LiDAR 9, which is mounted on the upper side of the body, i.e., the back of the body. Specifically, the third lidar 9 can be installed on the upper side of the head 7, the upper side of the body, or the upper side of the tail 8, without any specific restrictions, because its placement will not significantly affect the overall field of view of the quadruped robot. The installed third lidar 9 has a third scanning area 91 facing upwards on the robot body, and the first scanning area 41, the second scanning area 51, and the third scanning area 91 intersect each other in pairs to have a common field of view.

[0051] In this embodiment, in the horizontal direction, upward angle is positive and downward angle is negative. In the vertical direction, rightward angle is positive and leftward angle is negative.

[0052] In this embodiment, the third lidar is mounted upwards on the upper mounting surface of the robot body, and has a third mounting angle relative to the vertical direction. This third mounting angle is between -20 and 20 degrees, with a preferred angle of 0 degrees. That is, the angle between the third lidar and the vertical line is controlled within a range of ±20 degrees. This upper lidar contributes less to the field of vision of the landing area and is mainly used to ensure the quadruped robot's surrounding field of vision. Based on the quadruped robot of Embodiment 1, it can significantly increase the scanning of the area above the robot, thereby meeting the needs of the quadruped robot for mapping, localization, and navigation.

[0053] In this embodiment, the third scanning area does not contact the highest point of the robot's top, thereby reducing interference from the various devices mounted on the back of the robot to the third lidar scan. Furthermore, the first, second, and third scanning areas are connected and surround the front, rear, top, and bottom areas of the robot, effectively reducing blind spots in the scanning observation around the robot's perimeter.

[0054] In this embodiment, the overlapping area of ​​the first scanning area 41 and the second scanning area 51 forms a visible range on the ground between the front and rear legs, extending beyond the foremost point of the robot's vertical projection on the ground. Preferably, the first lidar 4 is mounted forward and downward on the front mounting surface of the robot 1, with a first mounting angle relative to the horizontal direction, preferably -50 degrees, to minimize scanning blind spots at extremely close proximity to the robot's front end. The second lidar 5 is mounted backward and downward on the front mounting surface of the robot 1, with a second mounting angle relative to the horizontal direction, preferably -50 degrees. The first lidar 4 and the second lidar 5 have a hemispherical field of view of 360 degrees horizontally and 90 degrees vertically.

[0055] For details, see attached. Figure 5 As shown, the intersection line s formed by the intersection points at the same height, if the height of the intersection line and the installation angle of a certain lidar at the head or tail are determined, the installation angle of another lidar can be determined by the field of view of the lidar at the intersection point. If the installation angle α of the head lidar is known, the height h at which the head lidar reaches the intersection line is... f The height h of the intersection line reached by the tail laser radar r If the distance between the head and tail lidars is d, then the installation angles α and β of the head and tail lidars satisfy the following relationship:

[0056]

[0057] The lidar is installed at a positive angle when facing forward and upward or backward and upward, and at a negative angle when facing forward and downward or backward and downward. The specific installation angles α and β of the lidar do not need to be exactly equal to these theoretical values. Appropriate adjustments can be made to meet the requirements for observing the ground and obstacles within the field of view of the landing area.

[0058] This embodiment focuses on describing the differences from Embodiment 1. Other identical content can be found in the relevant content description in Embodiment 1.

[0059] This embodiment discloses a quadruped robot with a three-LiDAR system mounted on its back. The combined configuration of three LiDARs achieves maximum spectral coverage around the robot, ensuring ideal four-way and landing area visibility. This provides excellent hardware support for critical tasks such as mapping, localization, navigation, perception, and motion control. It enhances the all-weather, all-scene adaptability of the quadruped robot system. Simultaneously, it boasts a wide visible field of view and limits blind spots to areas with minimal impact on the robot's application, enabling it to perform mapping, localization, navigation, perception, and motion control tasks without replacing or adding new sensors. Furthermore, by using only LiDAR, without using depth cameras or other auxiliary sensors, and utilizing a single type of sensor to cover the quadruped robot's perception field of view, it simplifies system complexity and post-processing difficulty, reducing reliance on computing power while maintaining the quadruped robot's perception capabilities.

[0060] In another embodiment, a point cloud data processing method is also disclosed for use in the multi-LiDAR quadruped robot disclosed in the foregoing embodiments. This method filters out the body point cloud from the multi-LiDAR point cloud while retaining the environmental point cloud, thereby reducing the impact of non-environmental point clouds on subsequent applications of the LiDAR point cloud, as shown in the attached figure. Figure 6 As shown, the point cloud data processing method may specifically include the following:

[0061] Step S1: Collect point cloud data generated by each lidar installed on the aircraft body, stitch together the point cloud data and convert it into the aircraft body coordinate system to form the first point cloud dataset.

[0062] In this embodiment, step S1 further includes a calibration step for the lidar. This calibration step unifies the multiple lidars to the robot's coordinate system, facilitating subsequent algorithm calls and processing of the stitched lidar data. For this multi-lidar quadruped robot, the shared field of view of the multiple lidars is very small, making conventional lidar calibration methods difficult to handle. The specific calibration steps are as follows.

[0063] Simultaneously, point clouds from multiple lidar sensors and information from the lidar's built-in inertial measurement unit (IMU) are collected. In a scene with relatively rich planar features, the quadruped robot is controlled to walk a certain distance and collect a certain amount of environmental information in that scene.

[0064] For each LiDAR, a LiDAR IMU mapping algorithm is run based on its point cloud and the IMU information it contains. Specifically, algorithms such as Fast_LIO or other mapping algorithms with similar performance can be used to obtain point cloud maps for multiple LiDARs.

[0065] Select any lidar as a reference and calculate the extrinsic parameters of other lidars relative to that lidar. This process is achieved through two steps: coarse registration and fine registration.

[0066] Features are extracted from point cloud maps corresponding to multiple LiDAR sensors, and a coarse registration algorithm is used to calculate the transformation matrix between the point cloud maps based on these features. Specifically, algorithms such as VPFBR-L, or coarse registration algorithms with similar performance, can be used. This transformation matrix serves as the initial extrinsic parameter matrix and is used as the initial input to the fine registration algorithm.

[0067] The transformation matrix between point cloud maps corresponding to multiple lidars is further calculated using a fine registration algorithm, such as ICP or NDT. This transformation matrix serves as the extrinsic parameter matrix for other lidars and a reference lidar.

[0068] By stitching together the point clouds of each lidar sensor and unifying them to a selected lidar coordinate system using the extrinsic parameter matrix, and then calculating the extrinsic parameter matrix from the lidar sensor to the robot's body coordinate system, the stitched lidar point cloud can be transformed into the quadruped robot's body coordinate system for subsequent use. Using the point cloud unified to the selected lidar coordinate system and the IMU mounted on the robot as input, a joint lidar-IMU calibration algorithm can be used. Algorithms such as LiDAR_IMU_Init, or similar calibration algorithms, can also be selected. Obtaining the extrinsic parameter matrix from the lidar coordinate system to the robot's body coordinate system completes the multi-lidar sensor calibration.

[0069] Step S2: The body point cloud in the first point cloud dataset is filtered out using a preset cuboid filtering algorithm to form the second point cloud dataset.

[0070] Specifically, the length from the front of the robot's head to the rearmost part of its tail, the width from the leftmost to the rightmost part of the robot's body, and the height from the knees to the top of the robot's body are obtained. Based on these lengths, widths, and heights, a cuboid dimension for the filtering algorithm is generated. The cuboid filtering algorithm is used to initially filter out the robot's point cloud, that is, to filter out the point cloud within the range of the robot's length from the front to the rearmost part, the left and right width range, and the height range from the knees to the top of the body, because the point cloud within this range is unlikely to come into contact with the environment, so it can be directly filtered out.

[0071] Step S3: Obtain the joint angle information of the two front legs and two hind legs of the current quadruped robot. After filtering out the body point cloud of the four legs below the knees from the second point cloud dataset using the leg point cloud filtering algorithm, the final environmental point cloud dataset is formed.

[0072] Specifically, the data of the preliminary body point cloud and the joint angle information reported by the robot are used as input. The body point cloud below the knee is filtered out according to the leg point filtering algorithm. The reason for filtering this part of the point cloud separately is that it is in direct contact with the ground, and it is impossible to simply determine whether it is a laser point that hits the leg or an environmental object on the ground. Whether it is deleted directly or completely retained, it may have an adverse effect on terrain perception, motion control and navigation.

[0073] In this embodiment, the quadruped robot uses a three-joint leg structure for both its front and hind legs, which includes a hip joint motor, a thigh joint motor, and a lower leg joint motor. Step S3 may specifically include the following:

[0074] Step S31: Obtain the rotation angles of each joint in the two front legs and two rear legs of the current quadruped robot, and calculate the knee control point q of each leg. k and ankle control point q a The coordinates in the body coordinate system are:

[0075]

[0076] Where q0 is the origin of the body coordinate system, This is the static variation matrix of the motor from the body to the hip joint. This represents the static variation matrix from the hip joint motor to the thigh joint motor. This represents the static transformation matrix from the thigh joint motor to the lower leg joint motor. This is the static variation matrix from the lower leg joint motor to the knee control point. This is the static variation matrix from the lower leg joint motor to the ankle control point. The static variation matrix is ​​derived from the quadruped robot's structural engineering drawing and depends on the robot's geometry and configuration. For a specific robot, it is a constant value, generally involving only positional changes, and can be expressed as:

[0077]

[0078] in Measured from the structural engineering drawings of the quadruped robot. For example, the static variation matrix from the body to the hip joint motors. Corresponding to specific x, y, z positional differences The above formula can be used to obtain... Other static transformation matrices can be obtained in the same way.

[0079] This is the rotational variation matrix generated by the rotation of the hip joint motor. This is the rotational change matrix generated by the rotation of the thigh joint motor. The rotational change matrix generated by the rotation of the lower leg joint motor:

[0080]

[0081] Where a h a is the joint rotation angle of the hip joint motor. t a is the joint rotation angle of the thigh joint motor. c This refers to the joint rotation angle of the lower leg joint motor.

[0082] Step S32: Based on the coordinates of the knee control point and ankle control point of each leg piece, construct four approximate geometric bodies of the lower leg and obtain their geometric body region coordinates in the body coordinate system. After filtering out the point clouds located in the four geometric body regions from the second point cloud dataset based on the geometric body region coordinates, the final environmental point cloud dataset is formed.

[0083] Specifically, after obtaining the knee control points q for each leg... k and ankle control point q a After determining the coordinates in the body coordinate system, the relationship between each point cloud in the stitched and transformed second point cloud dataset and these two control points can be calculated to distinguish between body point clouds and environment point clouds. (See attached image) Figure 7 As shown, it is possible to use two control points q on the same leg. k and q a A geometric shape is constructed; in this embodiment, a cylinder can be used as an example for analysis and judgment. Points falling inside the cylinder are identified as body point clouds, and points falling outside the cylinder are identified as environment point clouds. Specifically, this can be achieved by calculating the distance d between each point and the line connecting the two control points belonging to the same leg component.i Is it less than the threshold τ? d And whether the angle between the lines connecting the two control points is less than the threshold τ. γ If the point exists, the distance d from the line connecting the two control points to which one of the legs belongs is... i Less than or equal to the threshold τ d And the angle between the lines connecting the two control points is greater than or equal to the threshold τ. γ If the condition is met, the point is identified as an organism point cloud; otherwise, it is identified as an environment point cloud.

[0084] Finally, after the cuboid filtering algorithm and the leg point filtering algorithm, the LiDAR point cloud has reduced the body point cloud as much as possible, and the point cloud data after removing the body point cloud can be released to the user.

[0085] The point cloud data processing method disclosed in the above embodiments filters the fuselage point cloud by using a preset cuboid filtering algorithm on the stitched point cloud obtained from multiple lidars, and then filters the point cloud data again using a leg point cloud filtering algorithm based on the size of each leg component and the current joint angle information of the shutdown motor, so as to minimize the fuselage point cloud and eliminate the interference of the fuselage point cloud on lidar environmental observation.

[0086] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural transformations made using the contents of the present invention's specification and drawings under the inventive concept of the present invention, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A quadruped robot with multiple lidar sensors, characterized in that: It includes a body, two front leg pieces installed on both sides of the front of the body, and two rear leg pieces installed on both sides of the rear of the body; The first laser radar is arranged only forward on the front mounting surface of the body, and the first laser radar has a first scanning area facing the front of the body. The second lidar is arranged only rearward on the rear mounting surface of the aircraft body. The second lidar has a second scanning area facing the rear of the aircraft body. The second lidar is installed at a downward tilt, and the foremost point of the area where the second scanning area intersects with the ground exceeds the foremost point of the aircraft body's vertical projection on the ground. The first scanning area and the second scanning area have an intersecting line, which is located below the lowest point of the bottom of the body and above the highest point of the ground that can be passed through the obstacle; the second scanning area of ​​the second laser radar intersects with the ground, that is, the ground visibility range of the tail laser radar covers the pre-landing range of the two front legs taking one step forward.

2. The quadruped robot with multiple lidar sensors according to claim 1, characterized in that: The first lidar is mounted facing forward on the front mounting surface of the machine body and has a first mounting angle relative to the horizontal direction, the first mounting angle being -70 to 20 degrees.

3. The quadruped robot with multiple lidar sensors according to claim 2, characterized in that: The second lidar is installed at a downward tilt and has a second installation angle with the horizontal direction, the second installation angle being -90 to -20 degrees.

4. The quadruped robot with multiple lidar sensors according to claim 1, characterized in that: It also has a third lidar. The aircraft body includes a head, a fuselage and a tail connected in sequence. The third lidar is installed on the upper side of the head, fuselage or tail and has a third scanning area facing the top of the aircraft body. The first scanning area, the second scanning area and the third scanning area intersect each other and have a common field of view.

5. The quadruped robot with multiple lidar sensors according to claim 4, characterized in that: The third lidar is mounted upwards on the upper mounting surface of the body and has a third mounting angle relative to the vertical direction, the third mounting angle being -20 to 20 degrees.

6. The quadruped robot with multiple lidar sensors according to claim 5, characterized in that: The third scanning area does not contact the highest point of the top of the machine body. The first scanning area, the second scanning area, and the third scanning area are connected and surround the front, rear, top, and bottom areas of the machine body.

7. A point cloud data processing method, based on the multi-LiDAR quadruped robot as described in any one of claims 1-6, characterized in that, The steps include the following: The point cloud data generated by each lidar installed on the aircraft is collected, and the point cloud data is stitched together and converted into the aircraft coordinate system to form the first point cloud dataset. The second point cloud dataset is formed by filtering out the body point cloud in the first point cloud dataset using a preset cuboid filtering algorithm. The joint angle information of the two front legs and two hind legs of the current quadruped robot is obtained. After filtering out the body point cloud of the four legs below the knees from the second point cloud dataset using the leg point cloud filtering algorithm, the final environmental point cloud dataset is formed.

8. The point cloud data processing method according to claim 7, characterized in that: The front and rear leg components adopt a three-joint foot structure, which includes a hip joint motor, a thigh joint motor, and a lower leg joint motor.

9. The point cloud data processing method according to claim 8, characterized in that, The joint angle information of the two front legs and two hind legs of the current quadruped robot is obtained. After filtering out the body point cloud of the four legs below the knees from the second point cloud dataset using a leg point cloud filtering algorithm, the final environment point cloud dataset is formed, which specifically includes: Obtain the rotation angles of each joint in the two front legs and two rear legs of the current quadruped robot, and calculate the knee control points of each leg. and ankle control points The coordinates in the body coordinate system are: in Let be the origin of the body coordinate system, where This is the static variation matrix of the motor from the body to the hip joint. This represents the static variation matrix from the hip joint motor to the thigh joint motor. This represents the static transformation matrix from the thigh joint motor to the lower leg joint motor. This is the static variation matrix from the lower leg joint motor to the knee control point. The static change matrix from the lower leg joint motor to the ankle control point; This is the rotational variation matrix generated by the rotation of the hip joint motor. This is the rotational change matrix generated by the rotation of the thigh joint motor. The rotational change matrix generated by the rotation of the lower leg joint motor; in This refers to the joint rotation angle of the hip joint motor. This refers to the joint rotation angle of the thigh joint motor. This refers to the joint rotation angle of the lower leg joint motor. Based on the coordinates of the knee and ankle control points of each leg component, four approximate geometric bodies of the lower leg are constructed, and their geometric body region coordinates in the body coordinate system are obtained. After filtering out the point clouds located in the four geometric body regions from the second point cloud dataset based on the geometric body region coordinates, the final environmental point cloud dataset is formed.

Images