Four-legged robot path planning method, device and storage medium based on large model

Abstract

The application discloses a four-legged robot path planning method and device based on a large model and a storage medium, relates to the technical field of four-legged robots, and comprises the following steps: acquiring motion contrast data based on a large model; acquiring an optimal foot point area, a warning foot point area and a dangerous foot point area based on the motion contrast data; and adjusting a to-be-planned path to obtain a standard motion path to control robot motion. The application is used for solving the problem that, in the existing robot path planning method, a targeted path planning method is lacked when a four-legged robot moves to various terrains, the four-legged robot has abnormal behaviors such as unstable center of gravity or toppling in unknown positions of different terrains after path planning, and thus the four-legged robot cannot effectively complete a motion target.

Description

Technical Field

[0001] This invention relates to the field of quadruped robot technology, specifically to a method, device, and storage medium for quadruped robot path planning based on a large model. Background Technology

[0002] Quadruped robots are biomimetic robots that move using four legs, mimicking the locomotion of quadrupedal animals. They are capable of adapting to complex terrain and are widely used in military, rescue, and industrial fields. Quadruped robot path planning refers to determining the optimal or feasible trajectory of a robot from its starting point to its target point in a complex environment through algorithms. Its core lies in balancing motion efficiency, stability, and terrain adaptability.

[0003] Existing methods for path planning in quadruped robots typically place the path to be planned and the heading angle on a 3D map to transform a 2D navigation task into a 3D spatial search task. By simplifying the quadruped robot's geometry, the goal is to improve path planning efficiency and obtain smoother paths. While this approach ensures the path planning results are not overly conservative and improves the efficiency of obtaining smooth paths, it lacks specific path planning methods for quadruped robots navigating various terrains. This leads to abnormal behaviors such as instability or tipping at unknown locations on different terrains after path planning, hindering the quadruped robot's ability to effectively complete its movement objectives. For example, patent application CN115167425A discloses a quadruped robot... The proposed method for map building and path planning in robots transforms the navigation task in a two-dimensional plane into a three-dimensional spatial search problem by representing the location and heading angle to be planned in a three-dimensional grid map, thus improving the efficiency of path planning. It also simplifies the geometry of the quadruped robot to construct unique kinematic constraints and plan a smooth path. Other methods for quadruped robot path planning typically focus on obstacle avoidance improvements and still fail to address the lack of targeted path planning methods for quadruped robots navigating diverse terrains. This leads to abnormal behaviors such as instability or tipping at unknown locations in different terrains after path planning, hindering the robot's ability to effectively achieve its movement goals. Therefore, it is necessary to improve existing quadruped robot path planning methods. Summary of the Invention

[0004] This invention aims to at least partially solve one of the technical problems in the prior art. By proposing a path planning method, device, and storage medium for quadruped robots based on a large model, it addresses the lack of a targeted path planning method for quadruped robots moving through various terrains. This results in abnormal behaviors such as instability or tipping over at unknown locations in different terrains after path planning, which affect the normal movement of the quadruped robot and prevent it from effectively completing its movement objectives.

[0005] To achieve the above objectives, in a first aspect, this application provides a path planning method for quadruped robots based on a large model, comprising the following steps:

[0006] Motion data of the quadruped robot is obtained based on a large model and is referred to as motion analysis data. The motion analysis data is then filtered, and multiple quadruped motion terrains and motion comparison data for each quadruped motion terrain are obtained based on the filtering results. The motion comparison data includes routine data and accident data.

[0007] The motion comparison data of each quadrupedal movement terrain were analyzed, and the preferred stepping point area, warning stepping point area, and dangerous stepping point area corresponding to each quadrupedal movement terrain were obtained based on the analysis results.

[0008] The path to be planned is obtained, and the path is segmented based on all quadrupedal movement terrain to obtain multiple segmented planned paths. Each segmented planned path is adjusted based on the preferred stepping point area, warning stepping point area, and dangerous stepping point area of ​​the quadrupedal movement terrain. The path restored by the adjusted segmented planned path is recorded as the standard motion path of the quadrupedal robot, and the robot's movement is controlled by the standard motion path.

[0009] Furthermore, motion data of the quadruped robot is obtained based on a large model and denoted as motion analysis data. The motion analysis data is then filtered, and based on the filtering results, multiple quadrupedal motion terrains and motion comparison data for each quadrupedal motion terrain are obtained, including:

[0010] Obtain the data corresponding to the motion analysis data from the sensor data of the quadruped robot and record it as motion sensing data; use a large model to parse the motion sensor data and obtain the data corresponding to the quadruped robot's uphill, downhill, straight-line, and turning motions from the motion sensing data, and record them as uphill sensing data, downhill sensing data, straight-line sensing data, and turning sensing data, respectively. The quadruped's motion terrain includes uphill terrain, downhill terrain, straight-line terrain, and turning terrain.

[0011] Furthermore, the motion analysis data is filtered, and based on the filtering results, multiple quadrupedal movement terrains and motion control data for each quadrupedal movement terrain are obtained from the motion analysis data, including:

[0012] For uphill sensing data: The data corresponding to the uphill sensing data in the motion analysis data is recorded as uphill analysis data of uphill terrain. The uphill analysis data contains multiple sets of data, and each set of data corresponds to one uphill movement of the quadruped robot.

[0013] Data with abnormal behavior warnings among multiple sets of uphill analysis data are recorded as accident data, while data without abnormal behavior warnings among multiple sets of uphill analysis data are recorded as regular data.

[0014] Furthermore, the motion control data for each quadrupedal movement terrain were analyzed, and based on the analysis results, the preferred stepping point areas, warning stepping point areas, and dangerous stepping point areas corresponding to each quadrupedal movement terrain were obtained, including:

[0015] For any set of accident data: Based on the uphill sensing data corresponding to the accident data, obtain the planar diagram of the path of the quadruped robot in the uphill motion corresponding to the accident data from the top view, and record it as the top view motion planar diagram; obtain the minimum bounding rectangle of the top view motion planar diagram, and record it as the landing point analysis rectangle; obtain all the landing points of the four mechanical legs of the quadruped robot in the uphill motion corresponding to the accident data, and mark them as the robot landing points within the landing point analysis rectangle;

[0016] Among all mechanical landing points, the mechanical landing point generated when abnormal behavior data in the accident data is uploaded is obtained and recorded as the accident landing point; mechanical landing points other than accident landing points are recorded as regular landing points; the maximum contour obtained by fitting all regular landing points is obtained and the area within the contour is recorded as the regular landing point area; the maximum contour obtained by fitting all accident landing points is obtained and the area within the contour other than the regular landing point area is recorded as the accident landing point area.

[0017] Furthermore, the motion comparison data for each quadrupedal movement terrain were analyzed, and based on the analysis results, the preferred stepping point area, warning stepping point area, and dangerous stepping point area corresponding to each quadrupedal movement terrain were obtained, including:

[0018] For any set of conventional data: Based on the analysis method of accident data, obtain the landing point analysis rectangle corresponding to the conventional data and all mechanical landing points within the landing point analysis rectangle. The area within the maximum contour obtained by fitting all mechanical landing points is recorded as the conventional landing point area.

[0019] Obtain the general landing area and accident landing area for all accident data, and the general landing area for all general data.

[0020] Furthermore, the motion comparison data for each quadrupedal movement terrain were analyzed, and based on the analysis results, the preferred stepping point area, warning stepping point area, and dangerous stepping point area corresponding to each quadrupedal movement terrain were obtained, including:

[0021] Randomly acquire a set of data α from all accident data and regular data; denote the landing point analysis rectangle of data α as the standard rectangle; for all accident data other than data α: scale the landing point analysis rectangle of the accident data, as well as the regular landing point area and accident landing point area within the landing point analysis rectangle, proportionally based on the standard rectangle, until the landing point analysis rectangle of the accident data is the same as the standard rectangle. During proportional scaling, the movement direction of the quadruped robot within the standard rectangle and the landing point analysis rectangle is kept consistent.

[0022] For all regular data except for data α: scale the landing point analysis rectangle of the regular data proportionally based on the standard rectangle until the landing point analysis rectangle of the regular data is the same as the standard rectangle;

[0023] All scaled-down landing point analysis rectangles are made to coincide with the standard rectangle, and the resulting standard rectangle is denoted as the landing point coincidence rectangle. During the coincidence, the movement direction of the quadruped robot within all landing point analysis rectangles is kept consistent.

[0024] Furthermore, the motion comparison data for each quadrupedal movement terrain were analyzed, and based on the analysis results, the preferred stepping point area, warning stepping point area, and dangerous stepping point area corresponding to each quadrupedal movement terrain were obtained, including:

[0025] The area within the overlapping rectangle containing only regular landing points is designated as the preferred landing point area for uphill motion; the area within the overlapping rectangle containing only accident landing points is designated as the dangerous landing point area for uphill motion; and the area within the overlapping rectangle where both regular and accident landing points coincide is designated as the warning landing point area for uphill motion.

[0026] Based on the analysis method of uphill sensing data, downhill sensing data, straight-line sensing data, and turning sensing data are analyzed separately, and the preferred stepping point areas, warning stepping point areas, and dangerous stepping point areas corresponding to downhill terrain, straight-line movement terrain, and turning terrain are obtained.

[0027] Further, the path to be planned is obtained, and the path is segmented based on all quadrupedal movement terrain to obtain multiple segmented planned paths. Each segmented planned path is adjusted based on the preferred stepping point area, warning stepping point area, and dangerous stepping point area of ​​the quadrupedal movement terrain. The path restored from the adjusted segmented planned paths is recorded as the standard motion path of the quadruped robot. Controlling the robot's movement using the standard motion path includes:

[0028] Before the quadruped robot performs its movements, the path it takes during movement is obtained and recorded as the path to be planned. A large model is used to segment the path to be planned. The segmentation process includes: using the large model to simulate the quadruped robot's movement in the path to be planned, and recording the obtained sensor data as simulated sensing data; based on the parsing method of the motion sensor data in the large model, the large model is used to obtain the uphill terrain, downhill terrain, straight-line movement terrain, and turning terrain in the path to be planned from the simulated sensing data, and these are marked as segmented planning paths.

[0029] For any segmentation planning path: Based on the simulated sensor data, obtain the planar view of the path from the top view when the simulated quadruped robot moves in the segmentation planning path, and denot it as the simulated top view planar view; denote the minimum bounding rectangle of the simulated top view planar view as the simulated top view rectangle;

[0030] When the segmentation planning path is an uphill terrain, the simulated top view rectangle is scaled proportionally based on the landing point overlap rectangle corresponding to the uphill movement until the movement direction of the quadruped robot in the landing point overlap rectangle is the same as that in the simulated top view rectangle. The simulated top view rectangle and the landing point overlap rectangle can completely overlap. In the simulated top view rectangle after proportional scaling, the preferred stepping point area, the dangerous stepping point area, and the warning stepping point area are marked with the same as the landing point overlap rectangle corresponding to the uphill movement.

[0031] The simulated top-down rectangle is restored to its state before being scaled proportionally. Based on the scaling ratio of the simulated top-down rectangle during restoration, the preferred stepping point area, dangerous stepping point area, and warning stepping point area within the simulated top-down rectangle are scaled proportionally.

[0032] Based on the analysis method when the segmented planning path is an uphill terrain, the same analysis method is performed when the segmented planning path is a downhill terrain, a straight-movement terrain, or a turning terrain, and the corresponding simulated top view rectangle and the preferred stepping point area, dangerous stepping point area, and warning stepping point area within the simulated top view rectangle are obtained.

[0033] Based on the position of each segmented planning path in the path to be planned, the path formed by the preferred stepping point areas of all segmented planning paths is recorded as the standard motion path of the quadruped robot. When the quadruped robot performs movement, it moves along the standard motion path. When the landing point of any mechanical foot of the quadruped robot is in the warning stepping point area of ​​any segmented planning path during actual movement, a path deviation warning is sent until the landing points of all mechanical feet of the quadruped robot are in the preferred stepping point area.

[0034] When any mechanical leg lands in a dangerous step area of ​​any segmented planned path during actual movement, the quadruped robot stops moving and reports to the staff.

[0035] Secondly, this application provides an electronic device including a processor and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the steps in the method described above are performed.

[0036] Thirdly, this application provides a storage medium on which a computer program is stored, which, when executed by a processor, performs the steps of the method described above.

[0037] The beneficial effects of this invention are as follows: First, this application acquires motion analysis data based on a large model; the motion analysis data is then filtered, and multiple quadrupedal movement terrains and motion comparison data for each quadrupedal movement terrain are obtained based on the filtering results; then, the motion comparison data for each quadrupedal movement terrain is analyzed, and the preferred stepping point area, warning stepping point area, and dangerous stepping point area corresponding to each quadrupedal movement terrain are obtained based on the analysis results. The advantage of this is that by acquiring motion comparison data for multiple quadrupedal movement terrains and obtaining the corresponding preferred stepping point area, warning stepping point area, and dangerous stepping point area, it is possible to obtain the stepping points of the quadrupedal robot's mechanical feet when it moves to each terrain, based on the existing motion data of the quadrupedal robot, when it can move normally, when there is a low probability of abnormal behavior, and when there is a high probability of abnormal behavior. This allows for effective adjustment of the path for different terrains in the planned path during subsequent path planning, based on the preferred stepping point area, warning stepping point area, and dangerous stepping point area. This reduces the probability of abnormal behavior when the quadrupedal robot moves to each terrain along the planned path, thereby ensuring that the quadrupedal robot can effectively complete the movement objective.

[0038] This application also obtains the path to be planned, and segments the path based on all quadrupedal movement terrain. Based on the preferred stepping point area, warning stepping point area, and dangerous stepping point area of ​​the quadrupedal movement terrain, each segmented planned path is adjusted, and the path restored by the adjusted segmented planned path is recorded as the standard movement path of the quadruped robot. The robot's movement is controlled by the standard movement path. The advantage of this is that by segmenting the path to be planned, different terrains in the path to be planned can be separated, thereby adjusting the path corresponding to each terrain in the path to be planned, so that the path restored after adjustment can ensure that the quadruped robot can effectively complete the movement target. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating the steps of the method of the present invention;

[0040] Figure 2 This is a schematic diagram illustrating the acquisition of the landing point analysis rectangle according to the present invention;

[0041] Figure 3This is a schematic diagram illustrating the acquisition of the conventional impact area and the accident impact area according to the present invention.

[0042] Figure 4 This is a schematic diagram of the path to be planned according to the present invention;

[0043] Figure 5 This is a schematic diagram of the segmented planning path corresponding to the path to be planned in this invention;

[0044] Figure 6 This is a schematic diagram of the electronic device of the present invention. Detailed Implementation

[0045] 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.

[0046] Example 1, please refer to Figure 1 As shown, this application provides a path planning method for quadruped robots based on a large model, including the following steps:

[0047] Step S1: Obtain motion data of the quadruped robot based on the large model, and record it as motion analysis data; filter the motion analysis data, and obtain multiple quadruped motion terrains and motion comparison data of each quadruped motion terrain based on the filtering results. The motion comparison data includes conventional data and accident data.

[0048] Step S1 includes: Step S101, acquiring the data corresponding to the motion analysis data from the sensor data of the quadruped robot and recording it as motion sensing data; using a large model to parse the motion sensor data, and acquiring the data corresponding to the quadruped robot's uphill, downhill, straight-line, and turning motions from the motion sensing data, respectively recording them as uphill sensing data, downhill sensing data, straight-line sensing data, and turning sensing data, wherein the quadruped motion terrain includes uphill terrain, downhill terrain, straight-line motion terrain, and turning terrain;

[0049] In the actual implementation process, if there are multiple different terrains besides uphill terrain, downhill terrain, straight-movement terrain and turning terrain, such as stair-shaped uphill and sloping uphill, the four-legged movement terrain is divided into more detailed categories to ensure that the four-legged movement terrain can include all the terrains in the planned path, so as to achieve more comprehensive path planning.

[0050] Step S102, for the uphill sensing data: the data corresponding to the uphill sensing data in the motion analysis data is recorded as the uphill analysis data of the uphill terrain. The uphill analysis data contains multiple sets of data, and each set of data corresponds to one uphill movement of the quadruped robot.

[0051] In the specific implementation process, for example, in the uphill sensing data, if the quadruped robot starts to rise in height relative to the horizontal plane at time t1 and continues until time t2 when the quadruped robot stops rising in height relative to the horizontal plane, then the data between time t1 and t2 in the motion analysis data and the uphill sensing data can be recorded as the uphill sensing data of the uphill terrain.

[0052] Step S103: Record the data with abnormal behavior warnings in the multiple sets of uphill analysis data as accident data, and record the data without abnormal behavior warnings in the multiple sets of uphill analysis data as normal data.

[0053] Step S2: Analyze the motion comparison data of each quadrupedal movement terrain, and obtain the preferred stepping point area, warning stepping point area, and dangerous stepping point area corresponding to each quadrupedal movement terrain based on the analysis results;

[0054] Step S2 includes: Step S201, for any set of accident data: based on the uphill sensing data corresponding to the accident data, obtain a planar diagram of the path of the quadruped robot in the uphill motion corresponding to the accident data from a top-down perspective, and record it as the top-down motion planar diagram; obtain the minimum bounding rectangle of the top-down motion planar diagram, and record it as the landing point analysis rectangle; obtain all the landing points of the four mechanical legs of the quadruped robot in the uphill motion corresponding to the accident data, and mark them as the robot landing points within the landing point analysis rectangle;

[0055] In specific implementation processes, such as during a data analysis, the path of a quadruped robot moving uphill is as follows: Figure 2 As shown by curve CC1 between points AA1 and BB1, the planar view of the path obtained from the top-down perspective is DD1. Since DD1 is a rectangle, the corresponding landing point analysis rectangle is also DD1. By obtaining the landing point analysis rectangle, it is possible to ensure that when there are uphill paths with various terrains, all uphill paths can be analyzed equally by uniformly obtaining the landing point analysis rectangle. This results in the preferred landing point area, warning landing point area, and dangerous landing point area corresponding to all uphill paths, ensuring more comprehensive data analysis.

[0056] Step S202: Obtain the mechanical landing point generated when abnormal behavior data in the accident data is uploaded from all mechanical landing points, and record it as the accident landing point; record the mechanical landing points other than the accident landing points as regular landing points; obtain the maximum contour obtained by fitting all regular landing points, and record the area within the contour as the regular landing point area; obtain the maximum contour obtained by fitting all accident landing points, and record the area within the contour other than the regular landing point area as the accident landing point area.

[0057] In the specific implementation process, for example, during a data analysis, all the landing points in the landing point analysis rectangle DD1 are as follows: Figure 3 As shown, the center of all △ is the accident landing point, and the center of all ○ is the normal landing point. Through analysis, it can be found that region EE1 is the normal landing point region, and regions FF1 and FF2 are the accident landing points.

[0058] Step S2 also includes: Step S203, for any set of conventional data: based on the analysis method of accident data, obtain the landing point analysis rectangle corresponding to the conventional data and all mechanical landing points within the landing point analysis rectangle, and record the area within the maximum contour obtained by fitting all mechanical landing points as the conventional landing point area.

[0059] Step S204: Obtain the regular landing area and accident landing area of ​​all accident data, and the regular landing area of ​​all regular data.

[0060] Step S2 further includes: Step S205, randomly acquiring a set of data α from all accident data and regular data; recording the landing point analysis rectangle of data α as the standard rectangle; for all accident data other than data α: scaling the landing point analysis rectangle of the accident data, as well as the regular landing point area and accident landing point area within the landing point analysis rectangle, proportionally based on the standard rectangle, until the landing point analysis rectangle of the accident data is the same as the standard rectangle, wherein, during proportional scaling, the movement direction of the quadruped robot within the standard rectangle and the landing point analysis rectangle is kept consistent;

[0061] Step S206: For all regular data except data α: scale the landing point analysis rectangle of the regular data proportionally based on the standard rectangle until the landing point analysis rectangle of the regular data is the same as the standard rectangle.

[0062] Step S207: Make all the scaled-down landing point analysis rectangles completely coincide with the standard rectangle, and record the resulting standard rectangle as the landing point coincidence rectangle. When coinciding, keep the movement direction of the quadruped robot in the same direction within all landing point analysis rectangles.

[0063] In the specific implementation process, by obtaining the overlapping rectangle of landing points, it is possible to integrate the regular landing point area and accident landing point area corresponding to the landing point in all paths when the quadruped robot performs uphill movement, so as to ensure that the obtained preferred landing point area, warning landing point area and dangerous landing point area are the result of combining all uphill movements of the quadruped robot, and are more in line with the quadruped robot's own movement habits.

[0064] Step S2 further includes: Step S208, where the area in the overlapping rectangle of landing points containing only the normal landing point area is recorded as the preferred landing point area for uphill movement, the area in the overlapping rectangle of landing points containing only the accident landing point area is recorded as the dangerous landing point area for uphill movement; and the area in the overlapping rectangle of landing points where the normal landing point area and the accident landing point area coincide is recorded as the warning landing point area for uphill movement.

[0065] Step S209: Based on the analysis method of uphill sensing data, analyze downhill sensing data, straight-line sensing data and turning sensing data respectively, and obtain the preferred step area, warning step area and dangerous step area corresponding to downhill terrain, straight-line movement terrain and turning terrain.

[0066] Step S3: Obtain the path to be planned, segment the path to be planned based on all quadrupedal movement terrain, and obtain multiple segmented planning paths; adjust each segmented planning path based on the preferred stepping point area, warning stepping point area, and dangerous stepping point area of ​​the quadrupedal movement terrain, and record the path restored by the adjusted segmented planning path as the standard motion path of the quadrupedal robot, and control the robot's movement with the standard motion path.

[0067] Step S3 includes: Step S301, before the quadruped robot performs movement, the path of the quadruped robot when performing movement is obtained and recorded as the path to be planned; the path to be planned is segmented using a large model, the segmentation process includes: using the large model to simulate the movement of the quadruped robot in the path to be planned, and recording the obtained sensor data as simulated sensing data; based on the parsing method of the motion sensor data when the large model is used to parse the motion sensor data, the uphill terrain, downhill terrain, straight-line movement terrain and turning terrain in the path to be planned are obtained from the simulated sensing data using the large model, and marked as segmented planning paths respectively;

[0068] In the specific implementation process, for example, during a data analysis, the path to be planned for the quadruped robot is as follows: Figure 4 As shown by curve CC2 between AA2 and BB2, by dividing the uphill terrain, downhill terrain, straight-line movement terrain, and turning terrain in the path to be planned, the paths corresponding to the uphill terrain, downhill terrain, straight-line movement terrain, and turning terrain can be obtained as follows: Figure 5As shown in the path from TT1 to TT4, all paths from TT1 to TT4 can be recorded as segmentation planning paths and subsequent analysis can be performed.

[0069] Step S302: For any segmentation planning path: Based on the simulated sensor data, obtain the planar view of the path from the top view when the simulated quadruped robot moves in the segmentation planning path, and record it as the simulated top view planar view; record the minimum bounding rectangle of the simulated top view planar view as the simulated top view rectangle.

[0070] Step S303: When the segmented planning path is an uphill terrain, the simulated top view rectangle is scaled proportionally based on the landing point overlap rectangle corresponding to the uphill movement until the movement direction of the quadruped robot in the landing point overlap rectangle is the same as that in the simulated top view rectangle. The simulated top view rectangle and the landing point overlap rectangle can completely overlap. In the simulated top view rectangle after proportional scaling, the preferred stepping point area, the dangerous stepping point area, and the warning stepping point area are marked with the same as the landing point overlap rectangle corresponding to the uphill movement.

[0071] Step S304: Restore the simulated top view rectangle to its state before scaling by equal proportion, and based on the scaling ratio of the simulated top view rectangle during restoration, scale the preferred step area, dangerous step area, and warning step area within the simulated top view rectangle by the same proportion.

[0072] In the specific implementation process, by placing the preferred stepping point area, dangerous stepping point area and warning stepping point area of ​​each path in all the segmented planning paths in equal proportion, it can be ensured that the obtained standard motion path is the motion trajectory with the lowest probability of abnormal behavior when the quadruped robot moves in the path to be planned. At the same time, the dangerous stepping point area and warning stepping point area can also monitor the real-time movement of the quadruped robot, so as to provide timely danger warning when the quadruped robot moves to the area where abnormal behavior is likely to occur.

[0073] Step S305: Based on the analysis method when the segmented planning path is uphill terrain, the same analysis method is performed when the segmented planning path is downhill terrain, straight-movement terrain, or turning terrain, and the corresponding simulated top view rectangle and the preferred step area, dangerous step area, and warning step area within the simulated top view rectangle are obtained.

[0074] Step S306: Based on the position of each segmented planning path in the path to be planned, the path formed by the preferred stepping point areas of all segmented planning paths is recorded as the standard motion path of the quadruped robot, and the quadruped robot performs motion along the standard motion path when it performs motion. When the landing point of any mechanical foot of the quadruped robot is in the warning stepping point area of ​​any segmented planning path during actual movement, a path deviation warning is sent until the landing points of all mechanical feet of the quadruped robot are in the preferred stepping point area.

[0075] Step S307: When any mechanical leg lands in a dangerous step area of ​​any segmented planned path during actual movement, stop the movement of the quadruped robot and report to the staff.

[0076] Example 2, please refer to Figure 6 As shown, Figure 6 A schematic diagram of an electronic device is provided, which may include a processor, a communication interface, a memory, and a communication bus. The processor, communication interface, and memory communicate with each other via the communication bus. The memory stores computer-readable instructions, and the processor can call these instructions. When the processor executes a computer-readable instruction, it performs steps similar to those in a large-model-based quadruped robot path planning method to achieve the following functions: First, it acquires motion analysis data based on a large model; it filters the motion analysis data, obtaining multiple quadrupedal motion terrains and motion reference data for each terrain based on the filtering results; then, it analyzes the motion reference data for each terrain, obtaining the preferred stepping point area, warning stepping point area, and danger stepping point area corresponding to each terrain; finally, it obtains the path to be planned, segmenting the path based on all quadrupedal motion terrains to obtain multiple segmented planning paths; it adjusts each segmented planning path based on the preferred stepping point area, warning stepping point area, and danger stepping point area of ​​the terrain, and records the path restored from the adjusted segmented planning paths as the standard motion path of the quadruped robot, using the standard motion path to control the robot's movement.

[0077] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0078] Example 3: This application also provides a computer program product, which includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by the computer, the computer can execute the quadruped robot path planning method based on a large model provided by the above methods. The method includes: first, acquiring motion analysis data based on a large model; filtering the motion analysis data, and acquiring multiple quadruped motion terrains and motion comparison data for each quadruped motion terrain based on the filtering results; then analyzing the motion comparison data for each quadruped motion terrain, and acquiring the preferred stepping point area, warning stepping point area, and dangerous stepping point area corresponding to each quadruped motion terrain based on the analysis results; finally, acquiring the path to be planned, segmenting the path to be planned based on all quadruped motion terrains, and obtaining multiple segmented planning paths; adjusting each segmented planning path based on the preferred stepping point area, warning stepping point area, and dangerous stepping point area of ​​the quadruped motion terrain, and recording the path restored by the adjusted segmented planning path as the standard motion path of the quadruped robot, and controlling the robot's movement with the standard motion path.

[0079] Example 4: This application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it performs the steps of the above-described quadruped robot path planning method based on a large model to achieve the following functions: First, it acquires motion analysis data based on a large model; it filters the motion analysis data and acquires multiple quadrupedal motion terrains and motion comparison data for each quadrupedal motion terrain based on the filtering results; then it analyzes the motion comparison data for each quadrupedal motion terrain and acquires the preferred stepping point area, warning stepping point area, and dangerous stepping point area corresponding to each quadrupedal motion terrain based on the analysis results; finally, it acquires the path to be planned, segments the path to be planned based on all quadrupedal motion terrains, and obtains multiple segmented planning paths; it adjusts each segmented planning path based on the preferred stepping point area, warning stepping point area, and dangerous stepping point area of ​​the quadrupedal motion terrain, and records the path restored by the adjusted segmented planning path as the standard motion path of the quadruped robot, using the standard motion path to control the robot's movement.

[0080] Based on the above description of the embodiments, the embodiments of the present invention can be provided as methods, systems, or computer program products. Based on this understanding, the above technical solutions, in essence or in terms of their contribution to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or certain parts of the embodiments.

[0081] In the embodiments provided in this application, it should be understood that the disclosed system or method can be implemented in other ways. The embodiments described above are merely illustrative. For example, the division of modules or units is only a logical functional division, and there may be other division methods in actual implementation. Furthermore, multiple modules or units may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interfaces. The indirect coupling or communication connection between systems, modules, and units may be electrical, mechanical, or other forms.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A path planning method for quadruped robots based on a large model, characterized in that, Includes the following steps: Motion data of the quadruped robot is obtained based on a large model and is referred to as motion analysis data. The motion analysis data is then filtered, and multiple quadruped motion terrains and motion comparison data for each quadruped motion terrain are obtained based on the filtering results. The motion comparison data includes routine data and accident data. The motion comparison data of each quadrupedal movement terrain were analyzed, and the preferred stepping point area, warning stepping point area, and dangerous stepping point area corresponding to each quadrupedal movement terrain were obtained based on the analysis results. The path to be planned is obtained, and the path is segmented based on all four-legged movement terrain to obtain multiple segmented planning paths. Each segmented planning path is adjusted based on the preferred stepping point area, warning stepping point area, and dangerous stepping point area of ​​the four-legged movement terrain. The path restored by the adjusted segmented planning path is recorded as the standard movement path of the four-legged robot, and the robot's movement is controlled by the standard movement path. The motion control data for each quadrupedal movement terrain were analyzed, and based on the analysis results, the preferred stepping point areas, warning stepping point areas, and dangerous stepping point areas for each quadrupedal movement terrain were obtained, including: For any set of accident data: Based on the uphill sensing data corresponding to the accident data, obtain the planar diagram of the path of the quadruped robot in the uphill motion corresponding to the accident data from the top view, and record it as the top view motion planar diagram; obtain the minimum bounding rectangle of the top view motion planar diagram, and record it as the landing point analysis rectangle; obtain all the landing points of the four mechanical legs of the quadruped robot in the uphill motion corresponding to the accident data, and mark them as mechanical landing points within the landing point analysis rectangle; Among all mechanical landing points, the mechanical landing point generated when abnormal behavior data in the accident data is uploaded is obtained and recorded as the accident landing point; mechanical landing points other than accident landing points are recorded as regular landing points; the maximum contour obtained by fitting all regular landing points is obtained and the area within the contour is recorded as the regular landing point area; the maximum contour obtained by fitting all accident landing points is obtained and the area within the contour other than the regular landing point area is recorded as the accident landing area. For any set of conventional data: Based on the analysis method of accident data, obtain the landing point analysis rectangle corresponding to the conventional data and all mechanical landing points within the landing point analysis rectangle. The area within the maximum contour obtained by fitting all mechanical landing points is recorded as the conventional landing point area. Obtain the general landing area and accident landing area of ​​all accident data, and the general landing area of ​​all general data; Randomly acquire a set of data α from all accident data and regular data; denote the landing point analysis rectangle of data α as the standard rectangle; for all accident data other than data α: scale the landing point analysis rectangle of the accident data, as well as the regular landing point area and accident landing point area within the landing point analysis rectangle, proportionally based on the standard rectangle, until the landing point analysis rectangle of the accident data is the same as the standard rectangle. During proportional scaling, the movement direction of the quadruped robot within the standard rectangle and the landing point analysis rectangle is kept consistent. For all regular data except for data α: scale the landing point analysis rectangle of the regular data proportionally based on the standard rectangle until the landing point analysis rectangle of the regular data is the same as the standard rectangle; All scaled-down landing point analysis rectangles are made to coincide with the standard rectangle, and the resulting standard rectangle is denoted as the landing point coincidence rectangle. During the coincidence, the movement direction of the quadruped robot within all landing point analysis rectangles is kept consistent.

2. The path planning method for a quadruped robot based on a large model according to claim 1, characterized in that, Motion data of the quadruped robot was obtained based on a large model and denoted as motion analysis data. The motion analysis data was then filtered, and based on the filtering results, multiple quadrupedal motion terrains and motion comparison data for each quadrupedal motion terrain were obtained, including: Obtain the data corresponding to the motion analysis data from the sensor data of the quadruped robot and record it as motion sensing data; use a large model to parse the motion sensor data and obtain the data corresponding to the quadruped robot's uphill, downhill, straight-line, and turning motions from the motion sensing data, and record them as uphill sensing data, downhill sensing data, straight-line sensing data, and turning sensing data, respectively. The quadruped's motion terrain includes uphill terrain, downhill terrain, straight-line terrain, and turning terrain.

3. The path planning method for a quadruped robot based on a large model according to claim 2, characterized in that, The motion analysis data was filtered, and based on the filtering results, multiple quadrupedal movement terrains and motion control data for each quadrupedal movement terrain were obtained from the motion analysis data, including: For uphill sensing data: The data corresponding to the uphill sensing data in the motion analysis data is recorded as uphill analysis data of uphill terrain. The uphill analysis data contains multiple sets of data, and each set of data corresponds to one uphill movement of the quadruped robot. Data with abnormal behavior warnings among multiple sets of uphill analysis data are recorded as accident data, while data without abnormal behavior warnings among multiple sets of uphill analysis data are recorded as regular data.

4. The path planning method for a quadruped robot based on a large model according to claim 3, characterized in that, The motion comparison data for each quadrupedal movement terrain were analyzed, and based on the analysis results, the preferred stepping point area, warning stepping point area, and dangerous stepping point area for each quadrupedal movement terrain were obtained, including: The area within the overlapping rectangle containing only regular landing points is designated as the preferred landing point area for uphill motion; the area within the overlapping rectangle containing only accident landing points is designated as the dangerous landing point area for uphill motion; and the area within the overlapping rectangle where both regular and accident landing points coincide is designated as the warning landing point area for uphill motion. Based on the analysis method of uphill sensing data, downhill sensing data, straight-line sensing data, and turning sensing data are analyzed separately, and the preferred stepping point areas, warning stepping point areas, and dangerous stepping point areas corresponding to downhill terrain, straight-line movement terrain, and turning terrain are obtained.

5. The path planning method for a quadruped robot based on a large model according to claim 4, characterized in that, The process involves obtaining the path to be planned, segmenting it based on all four-legged terrain features, and generating multiple segmented planned paths. Each segmented planned path is then adjusted based on the preferred stepping point areas, warning stepping point areas, and dangerous stepping point areas of the four-legged terrain. The path restored from the adjusted segmented planned paths is recorded as the standard motion path for the four-legged robot. Controlling the robot's movement using this standard motion path includes: Before the quadruped robot performs its movements, the path it takes during movement is obtained and recorded as the path to be planned. A large model is used to segment the path to be planned. The segmentation process includes: using the large model to simulate the quadruped robot's movement in the path to be planned, and recording the obtained sensor data as simulated sensing data; based on the parsing method of the motion sensor data in the large model, the large model is used to obtain the uphill terrain, downhill terrain, straight-line movement terrain, and turning terrain in the path to be planned from the simulated sensing data, and these are marked as segmented planning paths. For any segmentation planning path: Based on the simulated sensor data, obtain the planar view of the path from the top view when the simulated quadruped robot moves in the segmentation planning path, and denot it as the simulated top view planar view; denote the minimum bounding rectangle of the simulated top view planar view as the simulated top view rectangle; When the segmentation planning path is an uphill terrain, the simulated top view rectangle is scaled proportionally based on the landing point overlap rectangle corresponding to the uphill movement until the movement direction of the quadruped robot in the landing point overlap rectangle is the same as that in the simulated top view rectangle. The simulated top view rectangle and the landing point overlap rectangle can completely overlap. In the simulated top view rectangle after proportional scaling, the preferred stepping point area, the dangerous stepping point area, and the warning stepping point area are marked with the same as the landing point overlap rectangle corresponding to the uphill movement. The simulated top-down rectangle is restored to its state before being scaled proportionally. Based on the scaling ratio of the simulated top-down rectangle during restoration, the preferred stepping point area, dangerous stepping point area, and warning stepping point area within the simulated top-down rectangle are scaled proportionally. Based on the analysis method when the segmented planning path is an uphill terrain, the same analysis method is performed when the segmented planning path is a downhill terrain, a straight-movement terrain, or a turning terrain, and the corresponding simulated top view rectangle and the preferred stepping point area, dangerous stepping point area, and warning stepping point area within the simulated top view rectangle are obtained. Based on the position of each segmented planning path in the path to be planned, the path formed by the preferred stepping point areas of all segmented planning paths is recorded as the standard motion path of the quadruped robot. When the quadruped robot performs movement, it moves along the standard motion path. When the landing point of any mechanical foot of the quadruped robot is in the warning stepping point area of ​​any segmented planning path during actual movement, a path deviation warning is sent until the landing points of all mechanical feet of the quadruped robot are in the preferred stepping point area. When any mechanical leg lands in a dangerous step area of ​​any segmented planned path during actual movement, the quadruped robot stops moving and reports to the staff.

6. An electronic device, characterized in that, It includes a processor and a memory, the memory storing computer-readable instructions that, when executed by the processor, perform the steps of the method as described in any one of claims 1-5.

7. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it performs the steps of the method as described in any one of claims 1-5.

Images