Obstacle avoidance method for backpack carrying robot, storage medium and related device

Abstract

The present application relates to the technical field of carrying robot, and discloses an obstacle avoidance method for a carrying robot, a storage medium and related devices, wherein the obstacle avoidance method comprises: detecting a rack feature through a sensor; if the rack feature is detected, matching the rack feature in a rack model feature library to obtain a specific model of a rack carried by the carrying robot; querying accurate size information of the carried rack according to the specific model of the rack; calculating a relative position relationship between the carried rack and the carrying robot according to the rack feature, the accurate size information of the rack and a platform rotation angle of the carrying robot; synthesizing a safe collision boundary of the carrying robot according to the accurate size information, a size of the carrying robot and the relative position relationship; if not, synthesizing the safe collision boundary of the carrying robot according to a preset general protection boundary; and performing dynamic obstacle avoidance on obstacles according to the safe collision boundary. The method can automatically adapt to the carrying of racks of different sizes.

Description

Technical Field

[0001] This invention relates to the field of handling robot technology, and specifically to an obstacle avoidance method, storage medium, and related devices for a backpack-type handling robot. Background Technology

[0002] With the development of industrial automation technology, the application of automated guided vehicles (AGVs) in industrial logistics scenarios is becoming increasingly widespread. Existing AGV intelligent obstacle avoidance systems typically control obstacle avoidance by setting static collision boundaries based on the fixed frame size of the AGV. However, for backpack AGVs carrying movable racks or shelves, carrying racks of different sizes can cause changes in the boundaries of the "vehicle-cargo assembly." Furthermore, the rotation of the lifting platform during transport dynamically affects the outer contour of the "vehicle-cargo assembly." If static boundaries based on the AGV frame size are still used for obstacle avoidance control, collision risks may arise. While pre-setting different collision boundaries for each rack in each posture can reduce collision risks, it significantly increases maintenance costs. Currently, most obstacle avoidance systems for backpack-type AGVs rely on manually setting collision boundaries based on the largest rack size. While this approach can ensure safe obstacle avoidance when carrying racks of various sizes at a relatively low cost, excessively large collision boundaries can lead to frequent sudden stops when the AGV passes through narrow areas or docking stations due to virtual collisions between the collision boundaries and static facilities on the map. This reduces the AGV's throughput and can cause goods on the racks to shake, resulting in damage. To address the damage caused by sudden stops, patent application number 202511436192.5, entitled "A 2D LiDAR Layered Shielding and Steplessly Regulated Collaborative Obstacle Avoidance System," proposes an obstacle avoidance system that incorporates prior environmental information. Through a layered collaborative architecture, it ensures natural speed changes during obstacle avoidance, preventing sudden acceleration or deceleration from affecting operational stability and efficiency. However, this solution still has room for improvement in terms of AGV throughput when dealing with the unique and complex obstacle avoidance boundary changes caused by the varied handling patterns of backpack-type AGVs. Furthermore, when an AGV carries a material rack, the rack legs are prone to generating "trailing point clouds" under LiDAR scanning (because the laser divergence angle penetrates to the object behind after hitting the edge, creating a point cloud without a corresponding actual object). Existing AGV obstacle avoidance solutions lack a mechanism to shield against these trailing point clouds, easily triggering misjudgments and causing abnormal deceleration or stopping. Therefore, it is necessary to design a new obstacle avoidance enhancement extension suitable for carrying robots, based on existing solutions, to overcome the aforementioned technical problems. Summary of the Invention

[0003] In order to overcome the shortcomings of the prior art, the purpose of this invention is to provide an obstacle avoidance method for a backpack-type transport robot, which can automatically set appropriate safety collision boundaries according to the characteristics of the rack on which it is mounted, so as to ensure the safe transport of different racks by the AGV and ensure good passability during the transport process.

[0004] To solve the above problems, the technical solution adopted by the present invention is as follows: an obstacle avoidance method for a backpack-type transport robot, comprising: Detect the characteristics of the material rack using sensors; If a rack feature is detected, the rack feature is matched in a pre-established rack model feature library to obtain the specific model of the rack carried by the handling robot; the precise size information of the rack carried is queried based on the specific model of the rack; the relative positional relationship between the rack carried and the handling robot is calculated based on the rack feature, the precise size information of the rack, and the platform rotation angle of the handling robot; and the safe collision boundary of the handling robot is synthesized based on the precise size information, the size of the handling robot, and the relative positional relationship. If no rack features are detected, the safe collision boundary of the handling robot is synthesized based on the preset general protection boundary, the size of the handling robot, and the platform rotation angle of the handling robot. Dynamic obstacle avoidance is performed based on safe collision boundaries.

[0005] The obstacle avoidance method for the aforementioned backpack-type handling robot includes the step of detecting rack features via sensors: The surrounding point cloud is obtained through sensors; Preprocess the point cloud; Cluster the point cloud and filter out interfering clusters in the clustering results; If the clustering result after filtering out interference is empty, it is determined that no rack feature was detected; If the clustering result after filtering out interference is not empty, the clustering result is divided into four quadrants with the center of the handling robot as the origin. Only the cluster closest to the origin is retained in each quadrant. The clusters of three adjacent quadrants are combined into a group, the right-angle features in each group are identified, and the long-side cluster, center cluster, and short-side cluster corresponding to each right-angle feature are determined to generate a feature candidate list. If the feature candidate list is not empty, the group with the smallest deviation from the ideal right angle in the feature candidate list is selected as the output feature of the group. The visualized rack features are output based on the output features of each group.

[0006] The obstacle avoidance methods for the aforementioned backpack-type transport robots also include: If a rack feature is detected, trailing noise filtering is performed on the point cloud detected by the sensor based on the precise size information of the matched rack and the relative positional relationship of the sensors on the handling robot.

[0007] The obstacle avoidance method for the aforementioned backpack-type handling robot includes the step of filtering trailing noise from the point cloud detected by the sensors based on the precise size information of the matched rack and the relative positional relationship between the sensors on the handling robot: Based on the position of each support leg of the matched support, the point cloud near each support leg is clustered to obtain the support leg point cloud cluster for each support leg, and the center coordinates and maximum outer boundary of each support leg point cloud cluster are calculated. Calculate whether the distance from point A in any point cloud to the center coordinate of each material rack leg point cloud cluster is greater than or equal to the effective filtering radius. If so, point A is determined to be a non-tailing noise point; otherwise, continue to the next step. Let point D be the center coordinate of the cluster of material rack leg points closest to point A. If the number of sensors is less than or equal to 2, then the position of the sensor closest to point D is taken as point E. If the number of sensors is greater than 2, then the position of the sensor that is closer to point D is selected from the two sensors that are closest to line AD and taken as point E. If the distance from point D to line AE is less than the filtering threshold, then point A is determined to be a trailing noise point; if the distance from point D to line AE is greater than or equal to the filtering threshold, then point A is determined to be a non-trailing noise point. Repeatedly traverse all points in the point cloud and filter out points that are identified as trailing noise.

[0008] The obstacle avoidance method for the aforementioned backpack-type transport robot includes the following step: synthesizing the safe collision boundary of the transport robot based on a preset universal protection boundary, the size of the transport robot, and the platform rotation angle of the transport robot. The system determines whether the transport robot is located in a low-speed passage area based on point cloud data acquired by the sensors and prior map data. If it is located in a non-low-speed passage area, the safe collision boundary of the handling robot is directly synthesized based on the preset general protection boundary, the size of the handling robot and the platform rotation angle of the handling robot; If the area is in a low-speed passage zone, the safe collision boundary of the transport robot is first synthesized based on the preset general protection boundary, the size of the transport robot, and the platform rotation angle of the transport robot. Then, the boundary of the safe collision boundary is contracted based on the closest distance between the transport robot and the obstacle.

[0009] A storage medium storing a computer program, which, when executed by a processor, implements the obstacle avoidance method of the aforementioned backpack-type transport robot.

[0010] A control device for a transport robot includes a processor and a memory, which are electrically connected via a data bus. The processor implements the obstacle avoidance method of the backpack transport robot described above by calling and executing a computer program in the memory.

[0011] A control device for a transport robot includes: an acquisition module for acquiring point cloud data detected by sensors; a rack feature detection and judgment module for determining whether a rack feature is detected based on the point cloud data acquired by the acquisition module; a rack model matching module for matching rack features in a pre-established rack model feature library when rack features are detected by the rack feature detection and judgment module to obtain the specific model of the rack carried by the transport robot, and querying the precise size information of the rack carried based on the specific model; a safety collision boundary synthesis module for synthesizing a safety collision boundary of the transport robot based on the precise size information, the size of the transport robot, and the relative position relationship when rack features are detected by the rack feature detection and judgment module; and for synthesizing a safety collision boundary of the transport robot based on a preset general protection boundary, the size of the transport robot, and the platform rotation angle of the transport robot when rack features are not detected by the rack feature detection and judgment module; and a dynamic obstacle avoidance module for controlling the transport robot to dynamically avoid obstacles based on the safety collision boundary synthesized by the safety collision boundary synthesis module.

[0012] The control device for the aforementioned handling robot further includes: a trailing noise filtering module, used to filter the point cloud acquired by the acquisition module for trailing noise based on the precise size information of the matching rack and the relative position relationship of the sensors on the handling robot when the rack feature detection and judgment module detects rack features; and a dynamic boundary shrinkage module, used to determine whether the handling robot is located in a low-speed passage area based on the point cloud data acquired by the acquisition module and prior map data when the rack feature detection and judgment module does not detect rack features, and if it is located in a low-speed passage area, shrink the boundary of the safety collision boundary synthesis module based on the closest distance between the handling robot and the obstacle.

[0013] A handling robot includes the control device described above.

[0014] Compared to existing technologies, the advantages of this invention are as follows: The obstacle avoidance scheme of the backpack-type handling robot in this embodiment detects the characteristics of the racks using sensors such as LiDAR, and matches the detected rack characteristics with pre-stored rack models to obtain precise rack size information. Based on the precise rack size information, the relative positional relationship between the rack and the handling robot, and the size of the handling robot, a suitable safe collision boundary is automatically synthesized. This eliminates the need to manually set appropriate safe collision boundaries for each type of material, ensuring the safe handling of racks by the handling robot while avoiding the impact on the robot's passage efficiency caused by uniformly setting collision boundaries according to the largest rack size. For racks without rack legs and whose characteristics cannot be detected, although a safe collision boundary is synthesized according to a uniformly preset universal protection boundary, when the handling robot detects narrow areas or workstations requiring low-speed passage, the boundary of the safe collision boundary is appropriately contracted based on the distance to the nearest obstacle to improve the handling robot's passage efficiency in these narrow areas. Furthermore, the obstacle avoidance scheme of the backpack-type handling robot in this embodiment of the invention will also mark and shield the "trailing point cloud" near the legs of the matching rack based on the precise size information of the rack, thereby avoiding the "trailing point cloud" generated by the LiDAR scanning of the rack legs being misjudged as an obstacle, which would lead to abnormal retrieval or shutdown during the handling process, thus ensuring handling efficiency.

[0015] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0016] Figure 1 This is a flowchart of an obstacle avoidance method for a backpack-type transport robot according to an embodiment of the present invention.

[0017] Figure 2 This is a flowchart illustrating the process of identifying and outputting rack features according to an embodiment of the present invention.

[0018] Figure 3 This is a flowchart illustrating the process of filtering trailing noise in an embodiment of the present invention.

[0019] Figure 4 This is a flowchart illustrating the dynamic shrinkage process of the safety collision boundary in an embodiment of the present invention. Detailed Implementation

[0020] The embodiments of the present invention are described in detail below, with reference to... Figure 1 The present invention provides an obstacle avoidance method for a backpack-type transport robot, comprising the following steps: Detect the characteristics of the material rack using sensors; If a rack feature is detected, the rack feature is matched in a pre-established rack model feature library to obtain the specific model of the rack carried by the handling robot; the precise size information of the rack carried is queried based on the specific model of the rack; the relative positional relationship between the rack carried and the handling robot is calculated based on the rack feature, the precise size information of the rack, and the platform rotation angle of the handling robot; and the safe collision boundary of the handling robot is synthesized based on the precise size information, the size of the handling robot, and the relative positional relationship. If no rack features are detected, the safe collision boundary of the handling robot is synthesized based on the preset general protection boundary, the size of the handling robot, and the platform rotation angle of the handling robot. Dynamic obstacle avoidance is performed based on safe collision boundaries.

[0021] This obstacle avoidance method utilizes sensors such as LiDAR to acquire the positional relationships and dimensional features of structures like the rack legs. This allows it to match and identify the specific model of the rack carried by the AGV from a pre-recorded rack model feature library. Based on this model, the precise dimensions of the rack can be determined, enabling the AGV to combine and align the point cloud data with the rack's accurate dimensions. This allows the AGV to determine the rack's position relative to the AGV and fuse the rack's outer contour with the AGV's outer contour to obtain a more accurate outer contour of the "vehicle-cargo assembly" after the rack is carried. This allows the AGV to automatically set an appropriate safety collision boundary based on the assembly's outer contour. The obstacle avoidance method of this invention automatically identifies the model of different racks and automatically sets a suitable size of safety collision boundary. This avoids the collision accidents that are easily caused by static collision boundaries set according to the size of the AGV frame, and also avoids the impact on the passage efficiency of the AGV in narrow areas by setting the collision boundary according to the largest rack size. It can ensure high handling and passage efficiency while carrying racks of different models and sizes, and also eliminates the need for users to manually set collision boundaries for replacement racks, thereby improving the operation and maintenance efficiency of the automated logistics system.

[0022] In practice, for movable racks with roller-equipped legs, the rack features are extracted using point cloud data of the rack legs obtained by sensors such as LiDAR. Specifically, in this embodiment, referring to... Figure 2First, the point cloud data obtained by the sensor is filtered and preprocessed to remove obvious noise. During preprocessing, a filtering radius can be set based on the circumcircle radius of the largest shelf to filter out points outside the radius and eliminate interference from obstacles such as external walls. Then, the preprocessed point cloud data is clustered, and the clustering results can be filtered again by combining prior static map data to filter out interfering clustering items such as walls. If the clustering result after filtering out interference is empty, it means that the shelf carried by the AGV is a shelf with a flat bottom without shelf legs. In this case, a safety collision boundary is synthesized according to the general protection boundary set based on the largest shelf size. If the clustering result after filtering out interference is not empty, it means that the shelf carried is a shelf with shelf legs. The size and model of the shelf can be identified based on the shelf features such as the relative positional relationship between the shelf legs. Then, the shelf information is further extracted from the clustering results.

[0023] In practice, material racks are mostly rectangular frame structures. The rack legs are usually located at the four corners of the rack's bottom surface, and there are usually support beams between the rack legs and the main body of the rack. The support beams will form a 90-degree right angle structure in the top view. The right angle structure can be detected and identified by detecting clustering results, thereby quickly determining the approximate top view outline of the rack. Based on the features of the top view outline, existing rack models can be matched to obtain the specific model of the rack. Therefore, in this embodiment, a rectangular coordinate system is established with the center of the handling robot as the origin, and the clustering results are divided into four quadrants. Only the cluster closest to the origin is retained in each quadrant. Then, the clusters of three adjacent quadrants are grouped together, resulting in four groups: quadrants 1-2-3, 2-3-4, 3-4-1, and 4-1-2. Right-angle features in each group are identified, and the long-side cluster, center cluster, and short-side cluster corresponding to each right-angle feature are determined, generating a feature candidate list. If the feature candidate list is empty, meaning no right-angle features are identified, it is considered that there are no detected shelf features. If the feature list is not empty, the right-angle features identified in each group are compared with the ideal right angle, and the results are determined based on the comparison. For the results, only the right-angle feature with the smallest deviation from the ideal right angle is retained in each group, and its corresponding long-side cluster, short-side cluster, and center cluster are retained as the output features of that group. Finally, based on the output features of each group, the long-side clusters in each group are merged into the long side of the rack to be identified, the short-side clusters in each group are merged into the short side of the rack to be identified, and the center cluster is regarded as the position of the rack leg. This outputs a visual rack feature containing the rack leg position of the identified rack, as well as approximate values ​​of the rack's length and width. This allows the AGV to calculate the similarity between the rack feature and the bottom surface features of each rack stored in the rack model feature library, and identify the rack model with the highest similarity as the specific model of the rack carried by the AGV.

[0024] In practice, for racks with support legs, when sensors such as LiDAR acquire point cloud data of the surrounding area via laser scanning, the laser has a certain divergence angle. At certain angles, when the laser emitted by the radar illuminates the edge of the rack leg, part of the emitted laser light may illuminate the rack leg and part may illuminate objects behind it. This results in the radar receiving a superposition of two reflected light spots, causing the radar to judge the target as being between two surfaces, thus detecting a virtual point cloud without a physical entity—a "tailing point cloud." These non-physical point clouds may be mistaken for obstacles, leading to AGV misdetection and incorrect obstacle avoidance behavior, affecting AGV throughput and potentially causing sudden stops and cargo collapse. Therefore, in some embodiments, to eliminate the influence of tailing point clouds, for racks whose features can be detected, it is necessary to filter the tailing noise near the rack legs based on the specific dimensions of the matched rack. This reduces the risk of AGV misdetecting obstacles and improves the smoothness and safety of the handling process.

[0025] Specifically, refer to Figure 3 First, based on the position of each support leg of the matched rack, the point cloud data near the support legs in the detected point cloud data is clustered to obtain a support leg point cloud cluster for each support leg, and the center coordinates and maximum outer boundary (maximum circumcircle) of each support leg point cloud cluster are calculated. Then, each point in the point cloud is traversed, and an effective filtering radius and filtering threshold are set based on the precise dimensions of the specific rack to determine which points are trailing noise. Specifically, let the traversed point be point A. The distance from point A to the center coordinates of each support leg point cloud cluster is calculated to see if it is greater than or equal to the effective filtering radius. If so, point A is considered to be far from the support leg and not near support leg A, thus eliminating the possibility of point A being a trailing noise and directly classifying it as non-trailing noise. If not, subsequent steps are continued to further investigate the risk of point A being a trailing noise. It is understood that the effective filtering radius can be the maximum distance between the four support legs of the rack, or the radius of the support's circumcircle.

[0026] Since light travels in a straight line through the air, the line connecting the position of a point in the point cloud to the sensor position represents the round-trip path of the light emitted by the radar. When the round-trip path is close to the vicinity of the support leg, it indicates that the point may be a trailing noise point caused by the time difference between the reflected light from two light spots in front of and behind the support leg. Therefore, as long as the position of the sensor closest to point A and the support leg is determined, and the distance between the line connecting point A and the center of the support leg is sufficiently far away from the support leg, the risk of the point being a trailing noise point can be ruled out. Specifically, refer to... Figure 3In this embodiment, the center coordinates of the cluster of material rack legs closest to point A are set as point D. Point D can be identified as the center point of the material rack leg closest to A. If the number of sensors is less than or equal to 2, the sensor closest to point D is considered to be the sensor closest to both points A and D. If the number of sensors is greater than 2, the sensor closest to line AD is considered the closest sensor. The position of the closest sensor is set as point E. If the distance from point D to line AE is less than a preset filtering threshold, point A is considered a trailing noise point. If the distance from point D to line AE is greater than or equal to the filtering threshold, point A is considered a non-trailing noise point. It is understood that the value of the filtering threshold should be slightly larger than the circumcircle radius of the matching material rack leg. The value by which the filtering threshold needs to exceed the circumcircle radius of the material rack leg can be obtained experimentally.

[0027] Understandably, after the AGV matches the precise dimensions of the carried rack, it can combine the obtained point cloud data and compare the relative positions of the four rack legs in the precise dimensions with the positions of the rack leg point cloud clusters to obtain the precise relative position of the rack to the AGV. After obtaining this precise relative position, the AGV can superimpose the precise outer contour of the rack with the outer contour of the AGV. The union of the superimposed rack contour and the AGV's frame contour is used as the outer contour of the combined vehicle-cargo assembly. Finally, the outer contour of the vehicle-cargo assembly is expanded by a preset safety threshold to obtain the final safe collision boundary. Similarly, for racks whose features cannot be detected, the union of a preset general protection boundary and the AGV's outer contour is used as the final safe collision boundary. During the handling process, when the lifting platform used by the AGV to carry the material rack rotates, the safety collision boundary of the vehicle-cargo assembly needs to be adjusted accordingly. At this time, based on the rotation angle of the lifting platform, the outer contour or general protection boundary of the carried material rack is rotated in the same way with the rotation center of the lifting platform as the center. The rotated contour is then merged with the AGV frame contour to obtain the outer contour of the rotated vehicle-cargo assembly. Finally, a preset safety threshold is added to the outer contour of the adjusted vehicle-cargo assembly to obtain the adaptive safety collision boundary after the lifting platform rotates.

[0028] Understandably, for racks whose features cannot be detected, the safety collision boundary is set based on a pre-set, large, general protection boundary. When passing through narrow areas requiring low-speed passage or docking stations, the AGV may still frequently stop due to this large safety collision boundary, affecting its throughput. Therefore, in some embodiments, for racks whose features cannot be detected, when the AGV locates itself in a narrow area requiring low-speed passage based on point cloud data obtained from sensors and a pre-recorded prior static map, the boundary of the safety collision boundary will be appropriately narrowed to improve the AGV's passage efficiency in these narrow areas.

[0029] Reference Figure 4 In this embodiment, the AGV device specifically traverses each point in the acquired point cloud data. Using SLAM technology and a pre-recorded static map, the AGV's position is determined. During the traversal, if a traversed point overlaps with the static map, the point's position relative to the AGV is calculated, and it is determined whether the point is in front of, to the left of, or to the right of the AGV. After traversing all points in the point cloud, the distances to the nearest point to the AGV among the points overlapping with the static map in front, to the left, and to the right are recorded as the front nearest distance, the left nearest distance, and the right nearest distance, respectively. When both the left nearest distance and the right nearest distance are less than a preset narrow environment judgment threshold, the AGV is considered to be in a narrow passage. When all three distances are less than the preset narrow environment judgment threshold, the AGV is considered to be in a concave, narrow docking station. In both cases, the safety collision boundary needs to be narrowed. When the AGV is in a narrow passage, if the nearest distance on the left is less than the left obstacle avoidance boundary to the left of the safety collision boundary, the left obstacle avoidance boundary of the safety collision boundary will be contracted to the nearest left distance; if the nearest distance on the right is less than the right obstacle avoidance boundary to the right of the safety collision boundary, the right obstacle avoidance boundary of the safety collision boundary will be contracted to the nearest right distance. When the AGV is in a narrow docking station, if the nearest distance in front is less than the front obstacle avoidance boundary in front of the safety collision boundary, the front side of the safety collision boundary will be contracted to the nearest front distance. It is understandable that the safety collision boundary cannot be contracted to less than the preset minimum protection boundary to avoid excessive contraction of the safety collision boundary leading to a collision accident.

[0030] It is understandable that the specific method of dynamically avoiding obstacles based on the safety collision boundary and point cloud data obtained by sensors, combined with prior map data, can be found in the application No. 202511436192.5, Publication No. CN121209504A entitled "A 2D LiDAR Layered Shielding with Stepless Speed ​​Regulation Cooperative Obstacle Avoidance System," which will not be elaborated here. Based on the solutions disclosed in the aforementioned patents, the obstacle avoidance method of this invention systematically enhances and expands the unique and complex obstacle avoidance challenges faced by backpack AGVs when handling materials with varying shapes, caused by the changing outer contours of the vehicle and cargo. It can be deeply integrated with the basic obstacle avoidance framework proposed in the previous case to form a complete backpack AGV safety navigation solution suitable for highly flexible industrial logistics scenarios, further improving the smoothness of backpack AGV material handling, reducing material damage during handling, and ensuring handling efficiency.

[0031] Based on the same inventive concept, embodiments of the present invention also provide a storage medium storing a computer program, which, when executed by a processor, implements the obstacle avoidance method of the backpack-type transport robot described above.

[0032] In some possible implementations, various aspects of the obstacle avoidance method for the backpack-type transport robot provided by the present invention can also be implemented in the form of a program product, which includes program code. When the program product is run on a device, the program code is used to cause the control device to perform the steps in the obstacle avoidance method for the backpack-type transport robot according to the various exemplary embodiments of the present application described above.

[0033] By designing and programming the processor, the code corresponding to the obstacle avoidance method of the backpack-type transport robot described in the foregoing embodiments can be embedded into the chip, thereby enabling the chip to execute the steps of the obstacle avoidance method of the backpack-type transport robot shown in the embodiments of the present invention during operation. How to design and program the processor is a technique known to those skilled in the art, and will not be described in detail here.

[0034] Based on the same inventive concept, embodiments of the present invention also provide a control device for a backpack-type transport robot, including a processor and a memory. The processor is communicatively connected to the memory via a data bus, and the processor can implement the obstacle avoidance method of the backpack-type transport robot described above by calling and executing a computer program in the memory.

[0035] In one possible design, the processor may include one or more processing units. The processor and memory may be implemented on the same chip or on separate chips. The processor may be a general-purpose processor, such as a central processing unit (CPU), digital signal processor, application-specific integrated circuit, field-programmable gate array or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the obstacle avoidance method for the backpack-type transport robot disclosed in the embodiments of this application can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules in the processor.

[0036] Memory, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. Memory can include at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type memory, random access memory (RAM), static random access memory (SRAM), programmable read-only memory (PROM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic memory, magnetic disk, optical disk, etc. Memory is any other medium capable of carrying or storing desired program code in the form of instructions or data structures that can be accessed by a computer, but is not limited to this. The memory in the embodiments of this application can also be a circuit or any other device capable of implementing storage functions for storing program instructions and / or data.

[0037] In some embodiments, the control device may include an acquisition module, a rack model matching module, a safety collision boundary synthesis module, and a dynamic obstacle avoidance module. The acquisition module acquires point cloud data detected by sensors. The rack feature detection and judgment module determines whether a rack feature is detected based on the point cloud data acquired by the acquisition module. The rack model matching module matches the rack feature in a pre-established rack model feature library when the rack feature detection and judgment module detects a rack feature, obtains the specific model of the rack carried by the handling robot, and queries the precise size information of the carried rack based on the specific model. The safety collision boundary synthesis module synthesizes the safety collision boundary of the handling robot based on the precise size information, the size of the handling robot, and the relative positional relationship when the rack feature detection and judgment module detects a rack feature; it also synthesizes the safety collision boundary of the handling robot based on a preset general protection boundary, the size of the handling robot, and the platform rotation angle of the handling robot when the rack feature detection and judgment module does not detect a rack feature. The dynamic obstacle avoidance module controls the handling robot to dynamically avoid obstacles based on the safety collision boundary synthesized by the safety collision boundary synthesis module.

[0038] In some embodiments, the control device further includes a trailing noise filtering module and a dynamic boundary shrinking module. The trailing noise filtering module filters the point cloud acquired by the acquisition module for trailing noise when the rack feature detection and judgment module detects rack features, based on the precise size information of the matched rack and the relative positional relationship of the sensors on the handling robot. The dynamic boundary shrinking module determines whether the handling robot is located in a low-speed passage area based on the point cloud data acquired by the acquisition module and prior map data when the rack feature detection and judgment module does not detect rack features. If it is located in a low-speed passage area, the boundary of the safety collision boundary synthesis module is shrunk based on the closest distance between the handling robot and the obstacle.

[0039] Based on the same inventive concept, embodiments of the present invention also propose a handling robot, including the aforementioned control device, which can automatically identify the model of the rack it carries and obtain the precise size information of the rack under the control of the aforementioned control device, thereby automatically setting appropriate safety collision boundaries to ensure safe transportation of the rack, while ensuring handling efficiency and reducing the operation and maintenance costs of the automated logistics system composed of the handling robot.

[0040] It should be noted that in the description of this invention, "several" means one or more, "more than" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. If "first" or "second" is mentioned, it is only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0041] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0042] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0043] These computer program instructions can also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0044] In the description of this invention, unless otherwise explicitly defined, terms such as "set up," "install," and "connect" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this invention in conjunction with the specific content of the technical solution.

[0045] The above embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of protection of the present invention. Any non-substantial changes and substitutions made by those skilled in the art based on the present invention shall fall within the scope of protection claimed by the present invention.

Claims

1. An obstacle avoidance method for a backpack-type transport robot, characterized in that, include: Detect the characteristics of the material rack using sensors; If a rack feature is detected, the rack feature is matched in the pre-established rack model feature library to obtain the specific model of the rack carried by the handling robot; the precise size information of the rack carried is queried based on the specific model of the rack; the relative positional relationship between the rack carried and the handling robot is calculated based on the rack feature, the precise size information of the rack, and the platform rotation angle of the handling robot. The safe collision boundary of the handling robot is synthesized based on precise dimensional information, the size of the handling robot, and its relative positional relationship. If no rack features are detected, the safe collision boundary of the handling robot is synthesized based on the preset general protection boundary, the size of the handling robot, and the platform rotation angle of the handling robot. Dynamic obstacle avoidance is performed based on safe collision boundaries.

2. The obstacle avoidance method for the backpack-type transport robot according to claim 1, characterized in that, The step of detecting the characteristics of the material rack using sensors includes: The surrounding point cloud is obtained through sensors; Preprocess the point cloud; Cluster the point cloud and filter out interfering clusters in the clustering results; If the clustering result after filtering out interference is empty, it is determined that no rack feature was detected; If the clustering result after filtering out interference is not empty, the clustering result is divided into four quadrants with the center of the handling robot as the origin. Only the cluster closest to the origin is retained in each quadrant. The clusters of three adjacent quadrants are combined into a group, the right-angle features in each group are identified, and the long-side cluster, center cluster, and short-side cluster corresponding to each right-angle feature are determined to generate a feature candidate list. If the feature candidate list is not empty, the group with the smallest deviation from the ideal right angle in the feature candidate list is selected as the output feature of the group. The visualized rack features are output based on the output features of each group.

3. The obstacle avoidance method for the backpack-type transport robot according to claim 1, characterized in that, Also includes: If a rack feature is detected, trailing noise filtering is performed on the point cloud detected by the sensor based on the precise size information of the matched rack and the relative positional relationship of the sensors on the handling robot.

4. The obstacle avoidance method for the backpack-type transport robot according to claim 3, characterized in that, The step of filtering trailing noise in the point cloud detected by the sensors based on the precise size information of the matched rack and the relative positional relationship of the sensors on the handling robot includes: Based on the position of each support leg of the matched support, the point cloud near each support leg is clustered to obtain the support leg point cloud cluster for each support leg, and the center coordinates and maximum outer boundary of each support leg point cloud cluster are calculated. Calculate whether the distance from point A in any point cloud to the center coordinate of each material rack leg point cloud cluster is greater than or equal to the effective filtering radius. If so, point A is determined to be a non-tailing noise point; otherwise, continue to the next step. Let point D be the center coordinate of the cluster of material rack leg points closest to point A. If the number of sensors is less than or equal to 2, then the position of the sensor closest to point D is taken as point E. If the number of sensors is greater than 2, then the position of the sensor that is closer to point D is selected from the two sensors that are closest to line AD and taken as point E. If the distance from point D to line AE is less than the filtering threshold, then point A is determined to be a trailing noise point; if the distance from point D to line AE is greater than or equal to the filtering threshold, then point A is determined to be a non-trailing noise point. Repeatedly traverse all points in the point cloud and filter out points that are identified as trailing noise.

5. The obstacle avoidance method for a backpack-type transport robot according to claim 1, characterized in that, The step of synthesizing the safe collision boundary of the transport robot based on the preset general protection boundary, the size of the transport robot, and the platform rotation angle of the transport robot includes: The system determines whether the transport robot is located in a low-speed passage area based on point cloud data acquired by the sensors and prior map data. If it is located in a non-low-speed passage area, the safe collision boundary of the handling robot is directly synthesized based on the preset general protection boundary, the size of the handling robot and the platform rotation angle of the handling robot; If the area is in a low-speed passage zone, the safe collision boundary of the transport robot is first synthesized based on the preset general protection boundary, the size of the transport robot, and the platform rotation angle of the transport robot. Then, the boundary of the safe collision boundary is contracted based on the closest distance between the transport robot and the obstacle.

6. A storage medium storing a computer program, characterized in that, When the computer program is invoked and executed by the processor, it implements the obstacle avoidance method of the backpack transport robot according to any one of claims 1 to 5.

7. A control device for a transport robot, characterized in that, It includes a processor and a memory, which are electrically connected via a data bus. The processor implements the obstacle avoidance method of the backpack transport robot according to any one of claims 1 to 5 by calling and executing a computer program in the memory.

8. A control device for a transport robot, characterized in that, include: The acquisition module is used to acquire point cloud data detected by the sensors; The rack feature detection and judgment module is used to determine whether rack features are detected based on the point cloud data acquired by the acquisition module. The rack model matching module is used to match the rack features in a pre-established rack model feature library when the rack feature detection and judgment module detects rack features, to obtain the specific model of the rack carried by the handling robot, and to query the precise size information of the rack carried based on the specific model of the rack. The safety collision boundary synthesis module is used to synthesize the safety collision boundary of the handling robot based on the precise size information, the size of the handling robot, and the relative position relationship when the rack feature detection and judgment module detects rack features; when the rack feature detection and judgment module does not detect rack features, it synthesizes the safety collision boundary of the handling robot based on the preset general protection boundary, the size of the handling robot, and the platform rotation angle of the handling robot. The dynamic obstacle avoidance module is used to control the transport robot to dynamically avoid obstacles based on the safe collision boundary synthesized by the safe collision boundary synthesis module.

9. The control device for the handling robot according to claim 8, characterized in that, Also includes: The trailing noise filtering module is used to filter the trailing noise of the point cloud obtained by the acquisition module according to the precise size information of the matching material rack and the relative position relationship of the sensors on the handling robot when the material rack feature detection and judgment module detects the material rack feature. The dynamic boundary shrinkage module is used to determine whether the transport robot is located in a low-speed passage area based on the point cloud data and prior map data obtained by the acquisition module when the rack feature detection and judgment module does not detect rack features. If it is located in a low-speed passage area, the boundary of the safety collision boundary synthesis module is shrunk based on the closest distance between the transport robot and the obstacle.

10. A transport robot, characterized in that, Includes the control device according to any one of claims 7 to 9.

Images