Robotic clamping system

Abstract

The application relates to a robot clamping system, comprising: a transmission assembly for driving a clamping assembly to move; the clamping assembly is used for clamping an object to be clamped; an environment sensing module is used for collecting environment images and point cloud data of an environment where the clamping assembly is located; a body sensing module is used for detecting force conditions and motion states of the clamping assembly; a control module is used for determining a target position of the object to be clamped according to the environment images and the point cloud data; the control module is further used for controlling the transmission assembly to drive the clamping assembly to move to a preset range of the target position; the control module is further used for controlling the clamping assembly to perform a clamping action on the object to be clamped; the control module is further used for determining a clamping completion condition of the clamping assembly on the object to be clamped according to the force conditions and the motion states returned by the body sensing module in the execution process of the clamping action; and the control module is further used for controlling the transmission assembly and the clamping assembly according to the clamping completion condition. The system can improve the reliability of transporting garbage cans.

Description

Technical Field

[0001] This application relates to the field of robotics, and in particular to a robotic gripping system. Background Technology

[0002] With the deepening of smart city construction and the continuous rise in labor costs, the automation and intelligentization of sanitation operations have become a trend.

[0003] As a core piece of automated sanitation equipment, the ability of outdoor cleaning robots to autonomously collect garbage is crucial, and a key aspect of this is the automatic clamping and emptying of garbage bins scattered throughout the area. The garbage bin clamping technology involved in related technologies is mainly applied to large garbage collection vehicles. When adapted or simplified for small, mobile outdoor cleaning robots, there are core requirements for meeting the demands of lightweight design, high reliability, high adaptability, and full autonomy for the garbage bin clamping operation of outdoor intelligent cleaning robots. Summary of the Invention

[0004] Therefore, it is necessary to provide a robot gripping system that can improve working accuracy and stability in response to the above-mentioned technical problems.

[0005] In a first aspect, this application provides a robot gripping system, which includes: a transmission component, a gripping component, an environmental perception module, a body perception module, and a control module; wherein...

[0006] The transmission assembly is slidably connected to the clamping assembly; the control module is communicatively connected to the environmental sensing module, the body sensing module, the transmission assembly, and the clamping assembly, respectively.

[0007] The transmission component is used to drive the clamping component to move;

[0008] The clamping assembly is used to clamp the object to be clamped;

[0009] The environmental perception module is used to collect environmental images and point cloud data of the environment in which the clamping component is located, and send the environmental images and point cloud data to the control module;

[0010] The body sensing module is used to detect the force and motion state of the clamping assembly, and send the force and motion state to the control module.

[0011] The control module is used to determine the target position of the object to be clamped based on the environmental image and the point cloud data.

[0012] The control module is also used to control the transmission component to drive the clamping component to move to a preset range of the target position;

[0013] The control module is also used to control the clamping component to perform a clamping action on the object to be clamped;

[0014] The control module is also used to determine the clamping completion status of the clamping component on the object to be clamped based on the force and motion status returned by the body sensing module during the execution of the clamping action.

[0015] The control module is also used to control the transmission assembly and the clamp assembly according to the completion status of the clamp.

[0016] The aforementioned robot gripping system overcomes the problem of unstable recognition by a single sensor in complex outdoor environments by simultaneously collecting environmental images and point cloud data for fusion recognition and positioning. By monitoring the force and motion status of the gripping components during the gripping process and making redundant judgments, it overcomes the shortcomings of traditional solutions that cannot accurately judge the gripping completion status and cannot handle abnormal working conditions. Thus, it can significantly improve the robot's autonomous operation capability and reliability in outdoor environments. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this embodiment or related technologies, the accompanying drawings used in the description of this embodiment or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a structural block diagram of a robot gripping system in one embodiment;

[0019] Figure 2 This is a schematic diagram of the robot gripping system in another embodiment;

[0020] Figure 3 This is a schematic diagram of the robot gripping system in another embodiment. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0022] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0023] In some embodiments, such as Figure 1 As shown, a robot gripping system 200 is provided, including the following modules:

[0024] Transmission assembly 202, clamping assembly 204, environmental sensing module 206, body sensing module 208, and control module 210; wherein,

[0025] The transmission assembly 202 is slidably connected to the clamping assembly 204; the control module 210 is communicatively connected to the environmental sensing module 206, the body sensing module 208, the transmission assembly 202, and the clamping assembly 204, respectively.

[0026] The transmission component 202 is used to drive the clamping component 204 to move;

[0027] The clamping assembly 204 is used to clamp the object to be clamped;

[0028] The environmental perception module 206 is used to collect environmental images and point cloud data of the environment where the clamping component 204 is located, and send the environmental images and point cloud data to the control module 210.

[0029] The body sensing module 208 is used to detect the force and motion state of the clamping assembly 204, and send the force and motion state to the control module 210.

[0030] The control module 210 is used to determine the target position of the object to be clamped based on the environmental image and the point cloud data.

[0031] The control module 210 is also used to control the transmission component 202 to drive the clamping component 204 to move within a preset range of the target position;

[0032] The control module 210 is also used to control the clamping component 204 to perform a clamping action on the object to be clamped;

[0033] The control module 210 is also used to determine the clamping completion status of the clamping component 204 on the object to be clamped based on the force and motion status returned by the body sensing module 208 during the execution of the clamping action.

[0034] The control module 210 is also used to control the transmission assembly 202 and the clamp assembly 204 according to the clamp completion status.

[0035] The robotic gripping system can be used for outdoor cleaning, such as performing garbage sweeping and collection tasks in outdoor environments like parks and streets. It is understood that this embodiment is not limited to cleaning robots, but can also be applied to other mobile robot platforms that require automatic target gripping.

[0036] The transmission assembly 202 may include a movable base and a driving component. The movable base provides autonomous movement capability and may include components such as a frame, drive wheels, casters, a drive motor, a reducer, and a battery. It can perform forward, backward, turning, and rotating movements according to the instructions of the control module 210. The transmission assembly 202 adopts a four-wheel independent drive or differential drive mode, which has good maneuverability and stability, and can travel stably on complex outdoor road surfaces (such as asphalt roads, masonry roads, and slight slopes). Correspondingly, the sliding component is used to drive the clamping assembly 204 to clamp objects of different heights vertically, thereby ensuring that the clamping center height matches the clamping part height of the object to be clamped. Specifically, the driving component includes a vertical slide rail, a horizontal slide rail, and a vertical driving component, forming a complete vertical lifting system.

[0037] The clamping assembly 204 refers to a mechanical actuator used for directly contacting, clamping, or securing an object to be clamped. The clamping assembly 204 can be designed according to the shape, size, and weight of the object. For example, the clamping assembly 204 can be a pair of symmetrically arranged clamping arms, the ends of which have arc-shaped clamping plates adapted to the outer wall of the object to be clamped. The clamping assembly 204 is driven by a drive motor to achieve opening and closing actions, thereby completing the clamping and releasing of the object to be clamped. Figure 2 As shown, the clamping assembly holds the object to be clamped, and the transmission assembly drives the clamping assembly to move.

[0038] The control module 210 refers to the core computing unit on the robot used to process sensor data, execute control algorithms, and issue control commands. The control module 210 may include components such as a processor (e.g., CPU, GPU, ARM chip), memory (for storing control programs and temporary data), and communication interfaces (for interacting with other modules). Specifically, the control module 210 can be an industrial control computer or an embedded system running a real-time operating system (e.g., ROS system) and integrating the robot control method described in the foregoing embodiments. The control module 210 is electrically connected to the recognition module (receiving environmental images and point cloud data), the transmission component 202 (controlling lifting and lowering actions), the gripping component 204 (controlling opening and closing actions), and the robot's mobile chassis (controlling driving actions), forming a complete "perception-decision-execution" closed loop.

[0039] The environmental perception module 206 can be a vision sensor or a LiDAR. The environmental image can be two-dimensional or three-dimensional image data acquired by a vision sensor such as an RGB camera, monocular camera, binocular camera, or depth camera. The environmental image provides visual feature information such as the color, texture, and shape of the object to be grasped, thus serving as an important basis for the robot to identify the target type and rough location. Point cloud data refers to a set of three-dimensional spatial points acquired by a LiDAR such as a single-line LiDAR, multi-line LiDAR, or solid-state LiDAR. Each point in the point cloud data contains its coordinate information (X, Y, Z) in three-dimensional space. Point cloud data can directly provide three-dimensional geometric information such as the distance, height, and orientation of the object to be grasped, and is unaffected by changes in ambient lighting.

[0040] Specifically, during the clamping task, the control module 210 controls the vision sensor and LiDAR to operate continuously at a preset frequency. The vision sensor acquires environmental images of the area in front of or around the robot and outputs two-dimensional image data; the LiDAR scans the environment and outputs three-dimensional point cloud data. The data acquired by the two sensors are synchronized in the time dimension (i.e., the environmental image and point cloud data acquired at the same time correspond to the same spatial area) and are transmitted to the robot together. Optionally, after receiving the environmental images and point cloud data acquired by the environmental perception module 206, the control module 210 can store them in a cache for later use in subsequent work processes. In practical applications, the acquisition frequency can be configured according to the moving speed of the transmission component 202 and the task requirements.

[0041] The object to be clamped refers to the target object that needs to be clamped. The specific type of the object to be clamped depends on the application scenario of the robot clamping system. Specifically, the object to be clamped can be a trash can, such as a 120L or 240L trash can. It is understood that this embodiment can also be applied to other objects that need to be automatically clamped and have clampable structural features, such as cargo boxes, packages, tools, etc., and this embodiment does not impose any limitations on this.

[0042] Considering the inherent limitations of single sensors in complex outdoor environments—for example, visual sensors are susceptible to interference from lighting, weather, and shadows, making it difficult to provide stable and accurate distance information, while lidar, although providing accurate three-dimensional coordinates, lacks color and texture information, making it difficult to distinguish different types of targets (such as trash cans and bushes)—this embodiment utilizes an environmental sensing module to simultaneously acquire multi-dimensional data such as environmental images and point cloud data. By leveraging the complementary characteristics of multi-dimensional data, all-weather, highly robust target recognition and localization can be achieved.

[0043] Specifically, the control module 210 can perform fusion processing on the environmental image and point cloud data to identify the object to be clamped (e.g., a trash can) and calculate the precise position and orientation of the object in three-dimensional space as the target position. The purpose of the fusion processing is to utilize the visual features (such as color, texture, and shape) provided by the environmental image and the three-dimensional geometric information (such as distance, height, and plane equations) provided by the point cloud data to achieve a more robust recognition and localization effect than a single sensor. For example, the control module 210 can perform target detection on the environmental image to determine the approximate area of ​​the object to be clamped in the image. Then, using sensor calibration parameters, the point cloud data is projected onto the image area to obtain a three-dimensional point cloud belonging to the object to be clamped. Finally, by performing geometric calculations on these point clouds (such as plane fitting and center point calculation), the three-dimensional spatial pose of the object to be clamped relative to the robot (including X-coordinate, Y-coordinate, Z-coordinate, and orientation angle) is obtained as the target position of the object to be clamped.

[0044] The control module 210 then converts the target position into a navigation command, and controls the transmission component 202 to move within a preset range of the target position according to the navigation command. The preset range refers to the ideal distance interval for the clamping component 204 to perform subsequent clamping actions. This range can be preset based on the working distance of the clamping component 204, the braking performance of the transmission component 202, and the safety requirements of the working environment. For example, the preset range can be set to 0.5 meters to 1.2 meters from the object to be clamped, ensuring that the clamping component 204 is roughly facing the object. Once the transmission component 202 moves within this preset range, it is considered that the clamping component 204 has entered a preparatory area suitable for performing the clamping action.

[0045] The clamping assembly 204 refers to a mechanical actuator used to contact, clamp, or fix an object to be clamped. The specific form of the clamping assembly 204 can be designed according to the shape, size, and weight of the object to be clamped. Specifically, the clamping assembly 204 can be a pair of symmetrically arranged clamping arms, the ends of which have arc-shaped clamping plates adapted to the outer wall of the object to be clamped.

[0046] The force condition detected by the body sensing module 208 refers to the characteristic parameters of the interaction force between the clamping component 204 and the object to be clamped during the clamping action. In this embodiment, the force condition may include the current value and torque value of the motor driving the clamping component 204 obtained by the body sensing module 208, or the readings of force sensors (such as pressure sensors or strain gauges) installed on the clamping component 204. Considering that the motor current is proportional to the output torque, and the output torque is converted into the clamping force of the clamping component 204 on the target object through the transmission mechanism, the force condition can be indirectly characterized by detecting the current value of the driving motor by the body sensing module 208. In this embodiment, by setting the body sensing module 208 to detect the current instead of a dedicated force sensor, no additional hardware cost is required, the system structure is simpler, the reliability is higher, and the problem of sensor damage in harsh outdoor environments can be avoided.

[0047] The motion state refers to the position or stage of motion of the clamping component 204 during the clamping action. The motion state can include whether the clamping component 204 has reached the preset clamping position, the current opening / closing angle of the clamping component 204, and the movement speed or acceleration of the clamping component 204. Specifically, the body sensing module 208 can include a limit switch installed on the clamping component 204, which is triggered when the clamping component 204 moves to the preset clamping position. Considering that the force applied alone cannot determine whether the clamping component 204 has reached the correct position—for example, the clamping component 204 may be prematurely obstructed due to clamping a foreign object, at which point the force has reached the required level but the target object has not been correctly clamped—this embodiment uses the body sensing module 208 to simultaneously monitor the force applied to the clamping component 204 and its motion state, forming a dual-redundant judgment, thereby improving the accuracy and safety of robot control.

[0048] The gripping completion status refers to whether the robot's gripping action on the object is successfully completed and the degree of completion. The gripping completion status can include at least one of the following states: "gripping successful" (gripping component 204 has correctly gripped the target object and the gripping force is sufficient), "gripping abnormal" (an abnormality occurs during the gripping process, such as empty gripping, gripping a foreign object, deformation of the barrel causing failure to reach the target position, etc.), and "gripping failed" (unable to successfully grip after multiple attempts).

[0049] Specifically, when the transmission component 202 moves to a preset range within the target position, the control module 210 initiates the clamping program, controlling the clamping component 204 to perform closing and clamping actions to clamp the object to be clamped. The specific control method for the clamping action can be configured in the control module 210 according to the type of clamping component 204. For example, for a motor-driven clamping component 204, the control module 210 can control the rotation direction and angle of the motor to achieve the closing and opening of the clamping component 204.

[0050] During the clamping action, the force condition of the clamping component 204, continuously acquired by the body sensing module 208, reflects the contact pressure between the clamping component 204 and the object to be clamped, serving as the basis for determining whether it has been "clamped". Correspondingly, the motion state of the clamping component 204 reflects its current motion stage or position, serving as the basis for determining whether it has "moved to the correct position". For example, the force condition can be obtained by the body sensing module 208 monitoring the current value of the motor driving the clamping component 204. When the clamping component 204 is running under no-load, the motor current remains at a low base value. However, when the clamping component 204 contacts the object to be clamped and begins to squeeze, the load increases, and the motor current rises accordingly. The motion state can be obtained through the aforementioned limit switch. When the clamping component 204 moves to the preset clamping position, the limit switch is triggered, outputting an electrical signal.

[0051] Unlike related technologies that rely on a single criterion (such as using only a limit switch to determine if the clamping position is reached, or using only a pressure sensor to determine if contact is made), this embodiment employs a dual-redundant judgment mechanism of force and motion to more accurately identify various clamping conditions. Specifically, the body sensing module 208 monitors the force situation (such as motor current value) in real time and transmits the force situation back to the control module 210. When the control module 210 determines that the force situation has reached a preset threshold (indicating that the clamping force is sufficient), it checks the motion status (such as whether the limit switch has been triggered) transmitted back by the body sensing module 208. If the control module 210 determines that the motion status also indicates that the clamping component 204 has reached the preset clamping position, then the clamping component 204 is determined to have successfully clamped the object to be clamped. Correspondingly, if the control module 210 determines that the force has reached the standard but the motion status shows that the clamping position is not reached, then the clamping component 204 is determined to have an abnormal clamping condition (for example, the clamping component 204 may have clamped a foreign object, or the object to be clamped may have undergone severe deformation, making it impossible to push into place). Therefore, the clamping completion status can be defined as a specific state value, such as "clamping successful", "clamping abnormal", or "clamping failed". This state value will serve as the basis for subsequent control decisions.

[0052] For example, during the closing and feeding process of the clamping assembly 204, the current value I of the drive motor driving the clamping arm, transmitted back by the body sensing module 208, is compared in real time with a preset clamping force threshold I_hold (e.g., 1.0A). If the current value I rises and reaches I_hold, the current triggering state of the limit switch is read. If the limit switch has been triggered (e.g., a high level is detected), it is determined that "clamping is successful"; if the limit switch has not been triggered (e.g., it is still at a low level), it is determined that "clamping is abnormal". This dual redundancy judgment mechanism can effectively identify various working conditions such as normal clamping, empty clamping, clamping foreign objects, and barrel deformation, avoiding the uncertainty caused by relying solely on position or force judgment in related technologies.

[0053] When the clamping completion status is "clamping successful," the control module 210 controls the clamping assembly 204 to maintain the current clamping force, ensuring that the object to be clamped will not loosen during subsequent movement. It also controls the transmission assembly 202 to lift the clamping assembly 204, thereby raising the object to be clamped to a safe height (e.g., more than 15 cm off the ground) to avoid collisions during the movement of the object. After reaching the target position, the control module 210 controls the transmission assembly 202 to lower the clamping assembly 204 and controls the clamping assembly 204 to release the object to be clamped, resetting to the ready-to-work state, completing a full clamping cycle.

[0054] Correspondingly, when the clamping completion status is "clamping abnormal," the control module 210 executes a preset fault handling strategy. For example, the control module 210 controls the clamping assembly 204 to stop feeding to avoid damage to the equipment due to continued force; or the control module 210 sends an alarm message to the management personnel or the backend system via sound and light or wireless communication; or the control module 210 controls the clamping assembly 204 to reverse a small distance to release any possible stress, and then tries clamping again. If clamping still fails after multiple retries, the control module 210 determines that the current task has failed, records the abnormal information (such as the current position of the robot, the state of the object to be clamped, etc.), skips the current target and continues to execute other tasks, and reports the abnormal situation to the preset management (such as the backend server or the administrator's terminal device) for manual handling.

[0055] The robot clamping system provided in this embodiment overcomes the problem of unstable identification by a single sensor in complex outdoor environments by simultaneously acquiring environmental images and point cloud data for fusion recognition and positioning. By simultaneously monitoring the force and motion state of the clamping component 204 during the clamping process and making redundant judgments, it overcomes the shortcomings of traditional solutions that cannot accurately judge the completion of clamping and cannot handle abnormal working conditions, thereby significantly improving the autonomous operation capability and reliability in outdoor environments.

[0056] In some embodiments, the control module 210 is further configured to:

[0057] The environmental image is identified by a pre-trained deep learning object detection model to obtain the two-dimensional bounding box corresponding to the object to be grabbed in the environmental image.

[0058] The point cloud data is projected onto the environmental image according to the calibration extrinsic parameters corresponding to the environmental perception module 206 to obtain the point cloud projection map corresponding to the clamping component 204, and the corresponding candidate point cloud clusters are extracted from the point cloud projection map according to the two-dimensional bounding box.

[0059] Outlier filtering is performed on the candidate point cloud clusters, and the three-dimensional spatial pose of the object to be clamped relative to the clamping component 204 is calculated based on the filtered point cloud as the target position.

[0060] The target location is converted into a navigation target point in the global map coordinate system, and the transmission component 202 is controlled to navigate to the preset range of the navigation target point;

[0061] The position of the clamping component 204 is adjusted so that the angle between the clamping component 204 and the object to be clamped is less than a preset angle threshold.

[0062] Deep learning object detection models refer to computational models built on deep neural networks used to identify and locate specific target objects in images. Unlike traditional color- or template-matching-based recognition methods, deep learning object detection models can automatically learn multi-level features of targets (such as edges, corners, textures, shapes, and even semantic information), thus exhibiting stronger robustness to changes in lighting, background interference, and target pose variations. For example, deep learning object detection models can be YOLO series models (such as YOLOv5s and YOLOv8), which are characterized by fast detection speed and high accuracy, making them suitable for deployment on edge computing platforms with high real-time requirements, such as robotics. The training dataset for this model can include trash can images under different times (daytime, nighttime), different weather conditions (sunny, cloudy, rainy), different angles, different backgrounds, and different colors (green, black, blue, gray, etc.). Data augmentation techniques such as flipping, scaling, brightness adjustment, and adding noise can be used to improve the model's generalization ability.

[0063] A two-dimensional bounding box (BWT) refers to the region in a two-dimensional image where a target object is marked with a rectangle. A BWT can be represented by the coordinates of the top-left and bottom-right corners (or the center point coordinates plus width and height) of the rectangle. While a BWT can tell the control module 210 the approximate location of the object to be clamped in the image, it may not provide information about the object's distance and orientation in three-dimensional space.

[0064] In this embodiment, a deep learning object detection model is introduced to perform forward inference calculations on the image, outputting the detection results of all target objects in the image belonging to a preset category (such as "trash can"). Each detection result includes at least: the category confidence of the target object, and the two-dimensional bounding box of the target object in the image (e.g., the coordinates of the upper left corner of the bounding box are (x_min, y_min), and the coordinates of the lower right corner are (x_max, y_max)). Unlike traditional recognition methods based on color or template matching, the deep learning model can effectively overcome problems such as changes in lighting, shadows, and background interference in outdoor environments (e.g., green objects in a green belt are similar in color to green trash cans), providing more robust recognition results, thereby solving the problems of "whether there is an object to be grabbed in the field of view" and "which region of the image is the object to be grabbed in".

[0065] The calibration extrinsic parameters include the sensor parameters corresponding to the environmental perception module 206. These sensor parameters include coordinate transformation parameters between different sensors (such as between a camera and a LiDAR), specifically rotation matrices and translation vectors. Using these sensor extrinsic parameters, point coordinates in one sensor coordinate system can be transformed to another. The calibration of these sensor extrinsic parameters can be completed before the robot leaves the factory or during system deployment. For example, it can be obtained by collecting multiple sets of corresponding points at different locations on a calibration board and solving them using algorithms such as PnP (Perspective-n-Point) or ICP (Iterative Closest Point). The accuracy of the calibrated extrinsic parameters directly affects the accuracy of subsequent point cloud projection and fusion localization.

[0066] A point cloud projection map is a mapping generated by projecting each point in a 3D point cloud onto a 2D environment image using sensor extrinsic parameters. In the point cloud projection map, each pixel position (x, y) can be associated with the corresponding point cloud depth value z. Using the point cloud projection map, image recognition results (2D bounding boxes) can be spatially aligned with 3D point cloud data, thereby enabling rapid cropping of the target point cloud.

[0067] Candidate point cloud clusters refer to the collection of 3D point clouds that are initially considered to belong to the object to be clamped, extracted from the point cloud projection map based on the 2D bounding box. Candidate point cloud clusters may contain the point cloud of the object to be clamped, or they may contain a small number of background points (such as ground points, points of objects behind, etc.), which need to be purified in subsequent processing.

[0068] Specifically, in this embodiment, the control module 210 can use pre-calibrated sensor extrinsic parameters (such as the rotation matrix R and translation vector T of the lidar to the camera) to project each point (X_lidar, Y_lidar, Z_lidar) in the 3D point cloud data collected by the environmental perception module 206 (such as lidar) onto the environmental image plane captured by the camera, thereby obtaining the pixel coordinates (u, v) of the point in the image. After performing the above projection operation on all points in the point cloud, a point cloud projection map is obtained, in which each pixel position (u, v) is associated with the corresponding point cloud depth information. Then, based on the 2D bounding box, all point clouds within the area covered by the bounding box are cropped from the point cloud projection map to form one or more candidate point cloud clusters. Since the 2D bounding box can closely fit the outer contour of the target object, the cropped candidate point cloud clusters mainly contain point clouds belonging to the object to be clamped, but may contain a small number of background points (such as ground points and points of objects behind).

[0069] Outliers are points in point cloud data that are spatially isolated from the main point cloud. Outliers can be caused by factors such as measurement noise from lidar, small floating objects in the environment (e.g., dust, leaves), and specular reflections from the surface of the target object. If outliers are not filtered out, they can negatively impact the accuracy of subsequent planar fitting and pose calculations.

[0070] Three-dimensional spatial pose refers to the position and orientation of a target object in three-dimensional space. Position can be represented by three-dimensional coordinates (X, Y, Z), and orientation can be represented by an orientation angle (such as the yaw angle around the Z-axis). In this application, the three-dimensional spatial pose includes at least the coordinates of the center point of the object to be gripped and the orientation of the barrel. This information is the key basis for the robot to plan the gripping path and adjust its own orientation.

[0071] Considering that the candidate point cloud cluster may contain outliers (such as isolated noise points, ground points, background object points, etc.), these points can affect the accuracy of subsequent geometric calculations. Therefore, in this embodiment, the control module 210 performs outlier filtering on the candidate point cloud cluster. Specifically, the outlier filtering can employ a statistical filtering method: the control module 210 calculates the average distance between each point in the candidate point cloud cluster and its k nearest neighbors. Assuming that this average distance follows a Gaussian distribution, points whose average distance exceeds a certain standard deviation of the mean (such as 1 or 2 standard deviations) are identified as outliers and removed. Optionally, a radius filtering method can be used: for each point, if the number of neighboring points within a specified radius (such as 0.2 meters) is less than a preset threshold, the control module 210 identifies it as an outlier and removes it.

[0072] After filtering out outliers, the remaining point cloud mainly represents the object to be clamped. Geometric features can be extracted and calculated from these point clouds to obtain the 3D spatial pose of the object relative to the robot. Specifically, since the sides of a standard trash can (such as a 120L or 240L trash can) can be relatively flat curved surfaces or planes, the control module 210 can use the Random Sample Consensus (RANSAC) algorithm to fit the dominant plane (which can correspond to the front or side of the trash can) from the filtered point cloud. By calculating the mathematical equation of this plane, we can obtain: the plane center point: the geometric center of all points on this plane, serving as the 3D coordinates (X_s, Y_s, Z_s) of a reference point on the trash can surface; and the plane normal vector: describing the orientation of the plane, i.e., the orientation vector of the front of the trash can. Combined with the standard geometric dimensions of the trash can (such as diameter and height), the precise coordinates (X_g, Y_g, Z_g) of the actual clamping center point of the trash can (which can be the center of the can body or the handle position) relative to the robot coordinate system, as well as the orientation angle θ, can be calculated as the target position.

[0073] The navigation target point refers to the target coordinates obtained after transforming the target position to the global map coordinate system, which are used for navigation by the transmission component 202. The global map coordinate system can be maintained by a SLAM (Simultaneous Localization and Mapping) system, which is a reference coordinate system fixed to the world (such as a park or street). After transforming the target position to the global map coordinate system, the transmission component 202 can perform global path planning on the constructed environment map to achieve autonomous navigation from the current position to the target position. The calculated target position (i.e., the three-dimensional spatial pose of the object to be gripped relative to the robot) is combined with the robot's current global pose (obtained through localization technologies such as SLAM), and the coordinates of the gripping center point are transformed to the global map coordinate system to obtain the navigation target point that can be directly used by the navigation system of the transmission component 202.

[0074] After obtaining the navigation target point, the control module 210 sends it to the navigation system of the transmission component 202. This allows the navigation system to plan a smooth path from the current location of the transmission component 202 to the navigation target point on the global map, and then controls the transmission component 202 to autonomously travel along this path until it moves the gripping component 204 to a preset range from the navigation target point. The preset range refers to the ideal distance interval for the gripping component 204 to perform subsequent gripping actions, and can be set according to the working distance of the gripping component 204, the robot's braking performance, and environmental safety requirements. For example, in one embodiment, the preset range can be set to 0.5 meters to 1.2 meters from the target location, with the robot's gripping component 204 approximately facing the target location.

[0075] The preset angle threshold refers to the maximum allowable angle between the clamping assembly 204 and the object to be clamped when the transmission assembly 202 and the clamping assembly 204 are adjusting their poses. This threshold can be set according to the working range of the clamping assembly 204, the shape of the object to be clamped, and the required operational accuracy. For example, it can be set to 5° or 10°. When the angle between the clamping assembly 204 and the object to be clamped is less than this threshold, it is considered that the robot is facing the object to be clamped and has the pose conditions to perform the clamping action.

[0076] Considering that after navigating to the preset range of the navigation target point, the clamping component 204 may not yet be fully aligned with the object to be clamped, for example, due to navigation errors or environmental factors, there may be a small angle between the orientation of the clamping component 204 and the orientation of the object to be clamped. Therefore, to ensure that the subsequent clamping action can be performed smoothly, the pose of the clamping component 204 needs to be fine-tuned. Specifically, based on the orientation angle of the object to be clamped, the difference between the current orientation of the clamping component 204 and that orientation is calculated. If the difference is greater than a preset angle threshold (e.g., 5°), the control module 210 sends a rotation command to the transmission component 202, causing it to rotate around its vertical axis until the angle between the clamping component 204 and the object to be clamped is less than the preset angle threshold. If the difference is already less than the preset angle threshold, no adjustment is needed, and the clamping component 204 is directly controlled to enter the subsequent clamping step, thereby enabling the clamping component 204 to enter the clamping preparation state with a better pose (e.g., aligned with the object to be clamped, at a suitable distance), creating favorable conditions for the successful execution of the subsequent clamping action.

[0077] This embodiment achieves robust 2D target detection through a deep learning model, and then achieves accurate 3D pose calculation through point cloud projection and geometric calculation. The two complement each other and can overcome problems such as lighting changes and background interference in complex outdoor environments, providing the robot with centimeter-level accurate target pose, thus laying a precise physical foundation for subsequent gripping actions.

[0078] In some embodiments, the control module 210 is further configured to: input an environmental image (e.g., an RGB image with a resolution of 640×480) acquired by the environmental perception module 206 into a pre-trained deep learning object detection model. This model may employ a lightweight YOLOv5s architecture, comprising a feature extraction network (Backbone), a feature fusion network (Neck), and a prediction network (Head) connected in sequence.

[0079] In the feature extraction stage, the environmental image is first fed into a feature extraction network. This network comprises multiple sequentially connected convolutional and pooling layers for downsampling and feature extraction of the input image, ultimately outputting three feature maps at different scales: 80×80, 40×40, and 20×20. These multi-scale feature maps correspond to different receptive fields in the original image; the 80×80 feature map is responsible for detecting small targets, and the 20×20 feature map is responsible for detecting large targets. The specific structural parameters of this feature extraction network are as follows: The Focus module slices the input image and then outputs a 32-channel feature map via convolution; subsequently, it passes through a first CBL module (composed of convolutional layers, batch normalization layers, and LeakyReLU activation function) with a kernel size of 3×3 and a stride of 2 to output a 64-channel feature map; the second CBL module outputs a 128-channel feature map, which is then connected to a first CSP module (composed of two convolutional layers and residual blocks) to output a 128-channel feature map; the third CBL module outputs a 256-channel feature map, which is then connected to a second CSP module to output a 256-channel feature map; the fourth CBL module outputs a 512-channel feature map, which is then connected to a third CSP module to output a 512-channel feature map, forming the 20×20 scale feature map. The fifth CBL module then outputs a 1024-channel feature map, and the SPP module (Spatial Pyramid Pooling) fuses features from different receptive fields, finally outputting the 40×40 and 80×80 scale feature maps.

[0080] Then, in the target detection stage, the three multi-scale feature maps mentioned above are input into the feature fusion network and the prediction network. The region generation network in the prediction network (in this embodiment, the detection head of YOLOv5s) predefines multiple candidate anchor boxes with different aspect ratios on each grid cell of each feature map. Specifically, for an 80×80 feature map, the anchor box sizes are set to (10,13), (16,30), and (33,23); for a 40×40 feature map, the anchor box sizes are set to (30,61), (62,45), and (59,119); and for a 20×20 feature map, the anchor box sizes are set to (116,90), (156,198), and (373,326). The region generation network predicts whether the target object exists within each anchor box and outputs the target confidence score and bounding box offset corresponding to each anchor box.

[0081] Next, based on a preset confidence threshold (e.g., 0.5), candidate anchor boxes with confidence scores higher than this threshold are selected from all anchor boxes as target anchor boxes. Then, the non-maximum suppression (NMS) algorithm is used to deduplicate the selected target anchor boxes. The IOU threshold of NMS is set to 0.4, that is, when the overlap between two target anchor boxes is greater than 0.4, the anchor box with the higher confidence score is retained and the anchor box with the lower score is discarded.

[0082] Finally, the retained position and size information of the target anchor frame is input into the model's bounding box regression layer. This regression layer fine-tunes the center point coordinates (x, y), width w, and height h of the target anchor frame based on the offset predicted by the network, outputting the final accurate two-dimensional bounding box, i.e., a rectangle defined by four coordinate values ​​(x_min, y_min, x_max, y_max). This rectangle represents the precise location of the object to be clamped (e.g., a 120L standard trash can) in the environmental image. This two-dimensional bounding box will then be used to guide subsequent point cloud data extraction and 3D pose calculation.

[0083] This application's embodiments employ a lightweight deep learning model (such as the YOLOv5s architecture) combined with multi-scale feature maps (80×80, 40×40, 20×20) to simultaneously achieve both accuracy and speed in target detection. Small-scale feature maps detect distant or small-sized objects to be grabbed, while large-scale feature maps detect nearby or large-sized objects. This allows the robot to stably and in real-time identify targets as it approaches them from a distance, avoiding detection failures due to changes in target size and improving the robustness of target recognition in outdoor environments. By specifically limiting the size of candidate anchor boxes (e.g., setting different sizes for different feature maps) and the confidence threshold (e.g., 0.5), false detections caused by a large amount of background noise (such as road surfaces, green belts, and other debris) can be effectively filtered out, ensuring high reliability of the 2D bounding boxes output by the model. These highly reliable 2D bounding boxes provide precise pixel-level region constraints for subsequent point cloud projection and 3D pose calculation, thereby reducing the point cloud search range and improving the overall system's computational efficiency and positioning accuracy.

[0084] In some embodiments, the control module 210 is further configured to:

[0085] Control the movement of the transmission component 202 and pause it at a preparatory position at a distance threshold from the object to be clamped;

[0086] The environment perception module 206 is controlled to acquire the environmental image and point cloud data at the preparatory position, and the target position is updated according to the environmental image and point cloud data corresponding to the preparatory position.

[0087] The clamping component 204 is pose-adjusted according to the updated target position, so that the clamping component 204 reaches a clamping preparation posture with a preset relative posture to the object to be clamped; in the clamping preparation posture, the angle between the clamping component 204 and the object to be clamped is less than the preset angle threshold.

[0088] The distance threshold refers to the distance boundary at which the transmission component 202 switches from long-distance navigation to short-distance precision positioning under the control of the control module 210. This distance threshold can be set according to the effective working distance of the environmental perception module 206, positioning accuracy requirements, and environmental safety factors. When the transmission component 202 is far from the object to be clamped (e.g., 10 to 15 meters), the target position calculated based on environmental images and point cloud data has a certain error (e.g., at the decimeter level). Directly using this target position as the clamping reference for controlling the clamping component 204 may lead to clamping failure. Therefore, in this embodiment, the control module 210 first controls the transmission component 202 to move to a relatively close preparatory position (i.e., within the distance threshold range), and at this position, the environmental perception module 206 is controlled to re-identify and locate the target to be clamped with higher precision.

[0089] Specifically, in this embodiment, based on the navigation target point, the control module 210 controls the transmission component 202 to autonomously travel along the planned path. During the movement of the transmission component 202, the environmental perception module 206 monitors in real time the distance between the current position of the clamping component 204 and the object to be clamped. When this distance decreases to a preset distance threshold (e.g., 0.8 meters), the control module 210 controls the transmission component 202 to stop moving and pauses it at its current position, which is the preparatory position.

[0090] It should be noted that the selection of the preparatory position should meet two conditions: first, it should be close enough to the object to be clamped so that the measurement accuracy of the environmental sensing module can meet the centimeter-level positioning requirements; second, it should be far enough away from the object to be clamped to avoid collision between the clamping component 204 and the object before pausing its movement. For example, for a 120L or 240L standard trash can, the preparatory position can be set at approximately 0.8 meters from the front of the trash can. At this distance, the point cloud density of the LiDAR is sufficient to support high-precision planar fitting, and the robot's clamping component 204 has not yet entered the clamping working area, leaving ample safety margin.

[0091] In this embodiment, the identification and localization algorithm is re-executed at the preparatory position to obtain more accurate target pose information. Unlike the first identification and localization (long distance, decimeter-level accuracy), this identification and localization is performed at close range (0.8 meters). This results in the target object occupying a larger pixel area in the environmental image, leading to higher confidence in the deep learning target detection model. Furthermore, the point cloud collected by the LiDAR on the surface of the target object is denser, resulting in higher accuracy in plane fitting and achieving centimeter-level or even millimeter-level positioning accuracy.

[0092] Specifically, after the robot pauses at the preparatory position, the following identification and localization process is executed again: First, the control module 210 uses a deep learning target detection model to identify targets in the environmental image collected by the environmental perception module 206 at the preparatory position, obtaining a two-dimensional bounding box of the object to be clamped; second, the control module 210 uses sensor extrinsic parameters to project the point cloud data collected at the preparatory position onto the environmental image, obtaining a point cloud projection map, and extracts candidate point cloud clusters based on the two-dimensional bounding box; then, the control module 210 performs outlier filtering on the candidate point cloud clusters and uses the RANSAC algorithm to fit the dominant plane; finally, the control module 210, combined with the standard geometric dimensions of the trash can, recalculates the three-dimensional spatial pose of the object to be clamped relative to the clamping component 204, as the updated target position.

[0093] The gripping preparation posture refers to the ideal relative pose between the gripping component 204 and the object to be gripped before executing the gripping action. In this posture, the gripping component 204 (such as a gripping arm) is facing the object to be gripped (i.e., the angle between them is less than a preset angle threshold), and the distance between the robot and the object to be gripped is within the working range of the gripping component 204. After reaching the gripping preparation posture, the gripping component 204 can directly initiate the gripping action without any further pose adjustments.

[0094] Specifically, the robot's pose is fine-tuned based on the updated target position (including the coordinates of the clamping center point and orientation angle of the object to be clamped). The pose fine-tuning may include: the control module 210 calculating the distance and direction between the current position of the clamping assembly 204 and the updated clamping center point, sending motion commands to the transmission assembly 202, causing the clamping assembly 204 to make small translations in the horizontal plane until the clamping assembly 204 reaches the optimal distance for performing the clamping action (for example, the distance between the clamping assembly 204 and the object to be clamped is 0.3 meters to 0.5 meters).

[0095] In some embodiments, the clamping assembly 204 includes a drive motor and a clamping arm; the drive motor is used to drive the clamping arm to move.

[0096] The body sensing module 208 includes a current detector and a limit switch; the current detector is used to detect the drive current output by the drive motor; the limit switch is used to detect the target event of the clamping arm moving to a preset clamping position.

[0097] The control module 210 is further configured to: control the transmission component 202 to drive the clamping component 204 to move until the clamping center height of the clamping component 204 matches the height of the clamping part on the object to be clamped, based on the height information of the object to be clamped;

[0098] The drive motor is controlled to drive the clamping arm to close in a direction closer to the object to be clamped;

[0099] The current detector is controlled to obtain the current value of the drive motor during the closing process of the clamping arm as the force condition;

[0100] The trigger state of the target event output by the limit switch during the closing process of the clamping arm is obtained as the motion state.

[0101] The height information refers to the vertical dimensions of the object to be clamped, specifically including the total height of the object and the height of the clamping part (such as the clamping ring or the middle of the trash can) above the ground. The height of the clamping part can be known in advance based on the trash can model, or it can be calculated in real time using the Z-coordinate information from the point cloud data.

[0102] The clamping center height refers to the optimal clamping position of the clamping assembly 204 in the vertical direction. For the clamping clamping assembly 204, the clamping center height can be set as the height of the geometric center of the arc-shaped clamping plate from the ground. In order for the clamping assembly 204 to stably clamp the object to be clamped, the clamping center height should match the height of the preset clamping part on the object to be clamped. For example, for a trash can, the optimal clamping part can be the reinforcing rib position or the "clamping ring" position in the upper part of the can body. This position has high structural strength and can ensure that the trash can remains balanced after being lifted.

[0103] Specifically, the control module 210 controls the transmission component 202 to move, causing the clamping component 204 to rise or fall as a whole. During the movement, the body sensing module 208 monitors the current height of the clamping component 204 in real time until the difference between the clamping center height of the clamping component 204 and the clamping part of the object to be clamped is less than a preset height error threshold (e.g., ±1 cm). At this point, the control module 210 determines that the height adjustment is complete and the clamping component 204 is aligned with the clamping part of the object to be clamped.

[0104] A drive motor is an actuator that provides the power for the clamp to close. The type of drive motor can be a servo motor, stepper motor, or brushless DC motor, depending on the requirements for control precision, response speed, and cost. For example, a 400W servo motor can be used, coupled with a ball screw mechanism to convert the motor's rotational motion into the linear motion of the clamp assembly 204, thereby achieving synchronous closing of the left and right clamp arms. Specifically, after the height adjustment is completed, the control module 210 sends a start command to the drive motor, causing it to start rotating. Power is then transmitted to the clamp assembly 204 via the transmission component 202 (such as a dual-sided output screw mechanism or a synchronous belt drive mechanism), for example, to the left and right clamp arms of the clamp assembly 204, driving them to move towards each other along the horizontal guide rail (i.e., closing towards the center). During the closing process, the control module 210 controls the drive motor to operate in a constant speed mode, allowing the clamp assembly to approach the object to be clamped at a smooth speed.

[0105] The current value refers to the amount of current consumed by the drive motor during operation. Considering that the current value of the drive motor is proportional to the output torque of the motor, and that the output torque is converted into the clamping force of the clamping assembly 204 on the target object through the transmission mechanism, the current value can be used as an indirect characterization of the force condition. When the clamping assembly 204 is running under no-load, the current value of the drive motor remains at a low base value (e.g., 0.3A); when the clamping assembly 204 contacts the object to be clamped and begins to squeeze, the load torque increases, and the motor needs to output a larger torque to maintain the closing motion, and the current value rises accordingly (e.g., from 0.3A to 1.0A). The magnitude of the change in the current value directly reflects the magnitude of the clamping force.

[0106] A limit switch is a sensor element installed on the clamping assembly 204 to detect whether the clamping assembly 204 has reached the preset clamping position. The specific type of limit switch can be a mechanical microswitch, a photoelectric switch, or a Hall effect switch. It should be noted that the preset clamping position refers to the position where the clamping arm pushes the trash can to the standard clamping depth. This position is designed considering the standard size of the trash can and the requirements for clamping stability. Clamping too shallowly may cause the trash can to slip during lifting, while clamping too deeply may damage the trash can or cause unnecessary energy consumption.

[0107] Specifically, during the process of the drive motor starting and driving the clamping assembly 204 to close, the control module 210 can control the current detector to collect the current value of the drive motor in real time at a preset sampling frequency (e.g., 100Hz to 1000Hz, corresponding to sampling once every 10 milliseconds to 1 millisecond), and store the collected current value in a buffer as raw data of the force condition. At the same time, the control module 210 monitors the output signal of the limit switch in real time. When the limit switch is not triggered, its output signal is in the first state (e.g., low level); when the clamping assembly 204 closes to the preset clamping position, the limit switch is triggered, and its output signal jumps to the second state (e.g., high level). Compared with related technologies that rely on only a single sensor (only current or only switch), the dual monitoring mechanism of this embodiment can identify a wider range of working conditions. For example, if the current value meets the standard but the limit switch is not triggered, it may mean that a foreign object has been clamped or the barrel is severely deformed; if the current value does not meet the standard but the limit switch has been triggered, it may mean that the clamping force is insufficient but the position is in place.

[0108] In some embodiments, the control module 210 is further configured to:

[0109] In response to the current value rising from a preset no-load current value, the current value of the drive motor is compared with a preset clamping current threshold.

[0110] When the current value reaches the clamping current threshold, the current trigger state of the limit switch is obtained;

[0111] If the current triggering state indicates that the limit switch has been triggered, it is determined that the clamping arm has successfully clamped the object to be clamped. If the current triggering state indicates that the limit switch has not been triggered, it is determined that the clamping arm has failed to clamp the object to be clamped.

[0112] The no-load current value refers to the current required to drive the motor when the gripping assembly 204 is not in contact with any object and is only overcoming its own friction and inertia. The no-load current value can be obtained through calibration before the robot leaves the factory, or it can be calibrated through a short period of no-load operation before each gripping task begins. For example, the no-load current value can be 0.3 amperes (A). When the current value is significantly higher than the no-load current value, it indicates that the gripping assembly 204 is under external load.

[0113] The rising current state refers to the trend of the current value changing from a lower value to a higher value. Unlike related technologies that only focus on whether the absolute value of the current exceeds a certain threshold, this embodiment first detects whether the current value is in a "rising state" to confirm that the clamping assembly 204 is indeed in contact and squeezing with the object to be clamped, rather than an abnormal current caused by other reasons (such as motor failure or voltage fluctuations). The detection of the rising current state can be achieved by calculating the first derivative of the current value (i.e., the rate of change of current): when the rate of change of current is continuously positive and exceeds a preset rate of change threshold, the current value is determined to be in a rising state.

[0114] The clamping current threshold (or clamping force threshold) refers to the current value required to determine that the clamping assembly 204 has applied sufficient clamping force to the object to be clamped. Since the motor current is proportional to the output torque, and the output torque is converted into clamping force through the transmission mechanism, the clamping force threshold can be indirectly set by setting the current threshold, and this threshold corresponds to the desired clamping force.

[0115] During the closing process of the clamping assembly 204, this embodiment continuously acquires the current value I of the drive motor at a preset sampling frequency (e.g., 100Hz). This embodiment first determines whether the current value I shows an upward trend starting from the no-load current value I_idle (e.g., 0.3A). If the change in current value over multiple consecutive sampling points (e.g., 10 consecutive sampling points) can be calculated, and if this change is consistently positive and the cumulative change exceeds a preset upward determination threshold (e.g., 0.05A), then the current value is determined to be in an upward state, indicating that the clamping assembly 204 has made contact with the object to be clamped and has begun to compress. In response to the determination that the current value is in an upward state, the current value I is compared with a preset clamping current threshold I_hold (e.g., 1.0A).

[0116] The current trigger state of the limit switch refers to the signal level output by the limit switch at the moment when the current value reaches the current threshold for clamping. There are two possible trigger states for the limit switch: triggered (e.g., outputting a high level, indicating that the clamping assembly 204 has moved to the preset clamping position) or not triggered (e.g., outputting a low level, indicating that the clamping assembly 204 has not yet reached the preset clamping position).

[0117] It should be noted that, considering that the current value reaching the threshold indicates that the clamping force is sufficient, if the limit switch is also triggered at this time, it means that the clamping assembly 204 has applied sufficient force and reached the correct position. The two corroborate each other, and the judgment result is reliable. If the limit switch is not triggered, it means that although the force is sufficient, the position is not in place, and an abnormal working condition may have occurred (such as clamping a foreign object or severe deformation of the barrel).

[0118] Specifically, limit switches can be installed on the left and right clamping arms of the clamping assembly 204 to obtain the triggering status of the two limit switches respectively. When both limit switches are triggered, the control module 210 determines that the movement state of the clamping assembly 204 meets the requirements; if either limit switch is not triggered, the control module 210 determines that the movement state of the clamping assembly 204 is abnormal. This dual-sided independent detection design can identify unilateral abnormal conditions such as when a trash can is not placed correctly, resulting in only one side being in position.

[0119] A successful clamping operation indicates that the clamping component 204 has correctly clamped the object to be clamped, with sufficient clamping force and proper positioning, allowing for safe subsequent lifting and transfer operations. A clamping malfunction indicates an unexpected situation occurring during the clamping process, making it impossible to confirm that the clamping component 204 has correctly clamped the object. Possible causes of clamping malfunctions include, but are not limited to: the object to be clamped not being in the expected position (empty clamp), foreign objects (such as stones or branches) being trapped between the clamping component 204 and the object, severe deformation of the object preventing it from being pushed into position, and limit switch malfunction. A clamping malfunction requires triggering a fault handling procedure, rather than continuing the lifting and transfer operations.

[0120] Specifically, if the limit switch is triggered (e.g., a high level is detected), it indicates that the clamping assembly 204 has moved to the preset clamping position. Simultaneously, since the current value has reached the clamping current threshold, it indicates that the clamping force is sufficient. Both conditions are met simultaneously, forming a double confirmation of "force meeting the standard and position in place." Based on this, this embodiment determines that the clamping is successful. For the design with dual limit switches, both the left and right limit switches must be triggered for the clamping to be considered successful.

[0121] Correspondingly, if the limit switch is not triggered (e.g., it remains at a low level), it indicates that although the current value has reached the threshold (the clamping force is sufficient), the clamping assembly 204 has not yet moved to the preset clamping position. This contradictory state of "force meeting the standard but position not in place" can mean that an abnormal working condition has occurred. For example, the clamping assembly 204 may have caught a foreign object (such as a stone) during the closing process, and the foreign object's obstruction prevents the clamping assembly 204 from continuing to close, even though the clamping force has reached the threshold; or the object to be clamped has undergone severe deformation (such as a dent in the wall of a trash can), causing the clamping assembly 204 to be unable to push the can to the standard clamping depth even if a sufficiently large clamping force is applied, so the control module 210 determines that the clamping is abnormal.

[0122] In some embodiments, if it is determined that the robot's gripping of the object to be gripped is abnormal, the control module 210 is further configured to: execute a preset fault handling strategy; the fault handling strategy includes at least one of controlling the drive motor to stop driving, sending a preset alarm message, and controlling the gripping component 204 to retreat a preset distance and then re-grip;

[0123] If the clamping component 204 is still deemed to have a clamping abnormality after the number of times it re-clamps reaches a preset threshold, the control module 210 is further configured to: record the current position of the clamping component 204 and report the current clamping abnormality to a preset management entity.

[0124] The fault handling strategy refers to a series of predefined operations performed by the robot to respond to the anomaly, protect equipment safety, and attempt to resume operation after a gripping anomaly occurs. The specific content of the fault handling strategy can be configured according to the type and severity of the anomaly and the safety requirements of the application scenario. This embodiment provides several combinable fault handling operations, such as stopping the drive, sending an alarm message, and re-gripping after rollback. These operations can be executed individually or in combination. For example, when a gripping anomaly occurs, stopping the drive and sending an alarm message can be performed simultaneously, followed by a rollback and retry.

[0125] Stopping the drive means immediately cutting off the power output of the drive motor, causing the clamping assembly 204 to stop its current closing feed motion. When a clamping abnormality occurs (such as clamping a foreign object or severe deformation of the barrel), continuing to drive the clamping assembly 204 to close may cause the clamping force to increase continuously, resulting in equipment damage (such as motor overload or transmission mechanism jamming) or damage to the object to be clamped (such as the barrel being cracked). Therefore, the control module 210 controls the drive motor to stop driving when it detects a clamping abnormality, thereby effectively reducing the risk of loss.

[0126] Preset alarm messages are used to notify administrators or the backend system of any anomalies. Alarm messages can take the form of: audible and visual alarms (such as a buzzer or flashing indicator light), wireless communication messages (such as messages sent to the backend server via 4G / 5G, Wi-Fi, or LoRa), or log entries (such as writing the anomaly to local storage). The content of the alarm message can include: the type of anomaly (such as "left gripper positioning anomaly"), the time of occurrence, the robot's current position, and the identification information of the object to be gripped. The technical significance of sending alarm messages is that it allows administrators to promptly understand the robot's operating status and intervene in serious anomalies in a timely manner, preventing the escalation of the fault.

[0127] The preset retraction distance refers to the distance by which the control module 210 controls the drive motor to reverse, causing the clamping assembly 204 to move a preset distance in the opening direction. The retraction distance can be set according to the type of clamping assembly 204 and the characteristics of the object to be clamped. If the retraction distance is too small, it may not be able to effectively release stuck foreign objects or relieve excessive stress; if the retraction distance is too large, the clamping assembly 204 may fully open, requiring it to close again when clamping again, reducing efficiency. For example, the preset retraction distance can be set to 5 mm, which is sufficient to release small foreign objects (such as stones or branches) and also relieve uneven stress caused by barrel deformation.

[0128] The number of re-grip attempts refers to the number of times the robot executes the "retreat-re-close" retry cycle after a gripping anomaly occurs. Since gripping anomalies may be caused by temporary factors (such as gripping a small, detachable foreign object, or the trash can adjusting its posture due to vibration), one or more retries may successfully complete the gripping. However, to avoid infinite retries that could cause the task to freeze or damage the equipment, an upper limit for the number of retries can be set in the control module 210, i.e., a preset threshold.

[0129] The preset threshold refers to the maximum number of retries allowed. Setting the preset threshold requires a trade-off between success rate and efficiency. If the threshold is too small, a potentially successful grab may be abandoned; if the threshold is too large, time may be wasted when a successful grab is not possible. Specifically, the preset threshold can be set to 2 retries: if the first grab fails, retry once; if the second attempt also fails, retry again; if the third attempt (after two retries) also fails, the task is considered a failure. This balances success rate and efficiency in outdoor scenarios. The preset management entity refers to the external system or personnel that receives reports of abnormal situations from the robot. Preset management entities can include: a backend server (for recording abnormal logs and triggering maintenance work orders), the administrator's terminal devices (such as mobile phones and computers for receiving alarm notifications in real time), and a remote monitoring center (for remote manual intervention), etc.

[0130] In one embodiment, the transmission assembly 202 includes: a vertical slide rail disposed on the chassis;

[0131] A horizontal slide rail is slidably engaged with the vertical slide rail, and the clamping assembly 204 is disposed on the horizontal slide rail;

[0132] A vertical drive component is connected to the horizontal slide rail for driving the horizontal slide rail to move along the vertical slide rail.

[0133] Among them, such as Figure 3As shown, a vertical slide rail refers to a guide element extending in the vertical direction (i.e., the vertical direction) to constrain the movement trajectory of the horizontal slide rail, ensuring that the clamping assembly 204 maintains linear movement during lifting and lowering, without lateral deviation or forward / backward tilting. The cross-sectional shape of the vertical slide rail can be circular, rectangular, or dovetail-shaped, depending on the load requirements and manufacturing cost. For example, a precision linear guide rail pair can be used, which features high rigidity, low friction coefficient, and smooth movement, and can withstand the overturning moment generated by the clamping assembly 204 during clamping and transport. Vertical slide rails can be one or more arranged in parallel. A single guide rail has a simple structure but weak torsional resistance; a double guide rail arrangement in parallel provides better torsional resistance and is suitable for applications with high loads. The fixed end of the vertical slide rail is mounted on the robot's mobile chassis, and its end (upper or lower end, depending on the installation direction) can be equipped with a mechanical hard limit to physically block the horizontal slide rail when it reaches its limit position, preventing it from exceeding its travel range.

[0134] like Figure 3 As shown, the horizontal slide rail refers to the moving component that slides in conjunction with the vertical slide rail, used to support the clamp assembly 204 and drive the vertical drive component. The horizontal slide rail may include a slider body and a mounting interface. The slider body mates with the vertical slide rail to form a low-friction kinematic pair; the mounting interface is used to fix the clamp assembly 204, ensuring a rigid connection between the clamp assembly 204 and the horizontal slide rail. Specifically, the horizontal slide rail is made of high-strength aluminum alloy or steel, minimizing weight while ensuring rigidity to reduce the load on the vertical drive component. Position detection elements (such as magnetic scales, optical scales, or Hall sensors) can also be installed on the horizontal slide rail for real-time feedback of the current position of the horizontal slide rail, thereby achieving closed-loop control.

[0135] The horizontal slide rail includes a base and a slider assembly. The base is a flat or frame-like structure used to support the clamping assembly 204, which is fixed to the base by bolts or other detachable connections. The slider assembly is mounted on the back of the base (i.e., the side in contact with the vertical slide rail) and forms a sliding engagement with the vertical slide rail. Rolling elements (such as balls or rollers) may be provided inside the slider assembly to reduce sliding friction and improve transmission efficiency. A position detection element may also be installed on the horizontal slide rail. For example, the position detection element can be a magnetic scale reader, which works in conjunction with a magnetic scale strip installed next to the vertical slide rail to read the absolute position of the horizontal slide rail in real time. Optionally, the position detection element can be a Hall sensor, which works in conjunction with magnets installed at different heights on the vertical slide rail to detect whether the horizontal slide rail has reached a preset key position (such as the highest position, lowest position, clamping position, etc.). The position detection element sends the detected position signal to the control module 210. The control module 210 determines whether the horizontal slide rail has reached the target position and sends a stop command to the vertical drive component, thus forming a closed-loop control.

[0136] A vertical drive component is an actuator that provides lifting power. It is connected to a horizontal slide rail and drives the horizontal slide rail to move along a vertical slide rail. The selection of a vertical drive component requires comprehensive consideration of factors such as load capacity, movement speed, control accuracy, and cost. The specific type of vertical drive component can be selected based on application requirements. For example, a vertical drive component might be an electric linear actuator. An electric linear actuator includes a drive motor, a reducer, a lead screw, and a push rod. When the drive motor rotates, it drives the lead screw to rotate through the reducer. The rotational motion of the lead screw is converted into the linear motion of a nut, which is fixedly connected to the push rod, allowing the push rod to extend or retract. The fixed end of the electric linear actuator (i.e., the motor and lead screw housing) is hinged or fixedly mounted on the robot's mobile chassis, while its movable end (i.e., the end of the push rod) is connected to the horizontal slide rail. Electric linear actuators have the advantages of compact structure, easy installation, and simple control, making them suitable for applications with low load requirements.

[0137] Optionally, the vertical drive component is a servo electric cylinder. The servo electric cylinder works on a similar principle to an electric actuator, but employs a high-precision ball screw and a high-performance servo motor, with a built-in high-resolution encoder, enabling millimeter-level or even higher precision position and speed control. Servo electric cylinders are suitable for applications requiring high positioning accuracy and smooth motion, such as scenarios where the clamping assembly 204 needs to be precisely aligned with the garbage can clamping ring.

[0138] The operation of the vertical drive component can include: when it is necessary to lift the clamping assembly 204, the control module 210 sends a positive command to the vertical drive component, causing the movable end of the vertical drive component to extend and push the horizontal slide rail upward along the vertical slide rail; when it is necessary to lower the clamping assembly 204, the control module 210 sends a reverse command to the vertical drive component, causing the movable end of the vertical drive component to retract and pull the horizontal slide rail downward along the vertical slide rail. During the movement, the control module 210 determines in real time whether the target position has been reached based on the position signal fed back by the position detection element, and issues a stop command when the target position is reached.

[0139] In some embodiments,

[0140] The clamping assembly 204 includes a horizontal guide rail, a left clamping arm, a right clamping arm, and a drive assembly;

[0141] The left clamping arm and the right clamping arm are symmetrically arranged and slide in cooperation with the horizontal guide rail respectively; the driving component is connected to the left clamping arm and the right clamping arm for driving the left clamping arm and the right clamping arm to move towards each other or away from each other along the horizontal guide rail;

[0142] The ends of the left clamping arm and the right clamping arm are respectively connected to arc-shaped clamping plates, and the inner side of the arc-shaped clamping plates is fitted with an elastic buffer pad.

[0143] The radius of curvature of the arc-shaped clamp is adapted to the outer wall of the object to be clamped;

[0144] The elastic buffer pad is used to increase clamping friction, absorb clamping impact, and accommodate the shape tolerances and placement errors of the object to be clamped.

[0145] The drive assembly includes a drive motor and a synchronous transmission mechanism;

[0146] The synchronous transmission mechanism includes a double-sided output screw mechanism or a synchronous belt transmission mechanism; the drive motor drives the left clamping arm and the right clamping arm to move synchronously in opposite directions or in opposite directions along the horizontal guide rail through the synchronous transmission mechanism.

[0147] The horizontal guide rail is fixedly mounted on the horizontal slide rail of the transmission assembly 202. Specifically, the base of the horizontal guide rail is fixed to the upper or front surface of the horizontal slide rail by bolts or other detachable connection methods, ensuring a rigid connection between the horizontal guide rails. The horizontal guide rail extends horizontally and is perpendicular to the forward direction of the robot's mobile chassis. The horizontal guide rail can be arranged in a dual-rail parallel configuration: two parallel horizontal guide rails are installed on the left and right sides of the horizontal slide rail respectively, forming a stable guiding system. Each horizontal guide rail is equipped with one or more sliders, and the left and right gripping arms are fixedly mounted on the corresponding sliders. The dual-rail structure can effectively resist the deflection torque generated by the gripping arms during the gripping process, ensuring the stability of the gripping movement. Limit blocks can be set at both ends of the horizontal guide rail (i.e., the left and right ends) to limit the maximum opening position of the left and right gripping arms and prevent the gripping arms from detaching from the ends of the guide rail. The inner side of the limit stop can be fitted with a buffer pad (such as a rubber pad) to absorb the impact when the clamping arm moves to its limit position.

[0148] The left and right clamping arms are symmetrically arranged and slide in conjunction with the horizontal guide rail. Specifically, the left clamping arm is slidably connected to the left side of the horizontal guide rail via a left slider, and the right clamping arm is slidably connected to the right side of the horizontal guide rail via a right slider. The left and right sliders move independently on the horizontal guide rail, but under the action of the drive component, they achieve synchronous opposite or backward movements.

[0149] The ends of the left and right clamping arms (i.e., the ends furthest from the horizontal guide rail) are each connected to an arc-shaped clamping plate. The arc-shaped clamping plates can be fixed to the clamping arms via bolts, welding, or integral molding. An elastic cushioning pad is fitted to the inner side of the arc-shaped clamping plate (i.e., the side facing the object to be clamped). The elastic cushioning pad can be fixed to the arc-shaped clamping plate by adhesive, vulcanization, or snap-fit, ensuring it will not fall off during use.

[0150] The radius of curvature of the curved clamp is adapted to the outer wall of the object to be clamped. For example, the object to be clamped is a 120L or 240L trash can. The diameter of a 120L trash can is approximately 470 mm to 490 mm, and the diameter of a 240L trash can is approximately 570 mm to 610 mm. To accommodate both sizes, the radius of curvature of the curved clamp can be designed to be between these values ​​(e.g., approximately 280 mm), allowing for a larger contact area with both 120L and 240L trash cans. When clamping a 120L trash can, the top and bottom edges of the curved clamp may contact the can wall first; when clamping a 240L trash can, the central area of ​​the curved clamp may contact the can wall first. The elastic deformation of the elastic cushioning pad compensates for this difference, achieving a tight fit.

[0151] Elastic buffer pads are used to increase gripping friction, absorb gripping impact, and accommodate shape tolerances and placement errors of the object to be gripped. Specifically: the high coefficient of friction surface of the elastic buffer pad prevents the trash can from slipping due to vibration during robot movement; the elastic deformation capacity of the elastic buffer pad can absorb impact energy at the moment of gripping, reducing damage to the surface of the trash can; when the trash can has slight deformation (such as dents or bulges) or is not placed absolutely vertically, the elastic buffer pad can achieve a tight fit through local compression deformation, improving the system's fault tolerance.

[0152] The drive assembly is connected to the left and right clamping arms for driving them to move in opposite directions or away from each other along the horizontal guide rail. The drive assembly includes a drive motor and a synchronous transmission mechanism.

[0153] The drive motor provides power for the clamping action. The drive motor can be a servo motor, stepper motor, or brushless DC motor. It is mounted on a base of a horizontal slide rail or guide rail, and its output shaft is connected to a synchronous transmission mechanism. This mechanism converts the rotational motion of the drive motor into linear motion of the left and right clamping arms, ensuring synchronization. The synchronous transmission mechanism can take the form of a double-sided output screw mechanism or a synchronous belt drive mechanism.

[0154] The implementation of a dual-output lead screw mechanism can include: a lead screw with two sections of threads in opposite directions (i.e., left-hand and right-hand threads). The left section of the thread engages with the nut seat of the left clamping arm, and the right section of the thread engages with the nut seat of the right clamping arm. When the drive motor rotates the lead screw, the left and right nut seats move in opposite directions along the lead screw axis, thereby causing the left and right clamping arms to close or open synchronously. The advantages of the dual-output lead screw mechanism are: it ensures absolute synchronization of the left and right clamping arms from a mechanical structure perspective; it is simple to control (only the direction and speed of the motor need to be controlled); it has high synchronization accuracy; and it is suitable for applications with high synchronization requirements.

[0155] The implementation of a synchronous belt drive mechanism can include: a driving synchronous pulley is mounted on the output shaft of the drive motor; the left and right clamping arms are connected to the synchronous belt via left and right driven pulleys, respectively. The synchronous belt moves under the drive pulley, simultaneously driving the left and right driven pulleys to rotate in opposite directions at the same speed. By converting rotational motion into linear motion (e.g., through a gear, rack, or lead screw mechanism), the synchronous opposite or backward movement of the left and right clamping arms is achieved.

[0156] In some embodiments, a limit switch is installed on the side of the left clamping arm and / or the right clamping arm facing the object to be clamped; the limit switch is used to be triggered when the left clamping arm and / or the right clamping arm reaches a preset clamping position and to send a trigger signal to the object.

[0157] The clamping assembly 204 also includes a bottom guide rail, which is fixedly installed on the robot's mobile chassis and located below the left clamping arm and the right clamping arm. The bottom guide rail is used to guide the object to be clamped into the clamping area and limit the position of the object to be clamped. A damping buffer is provided at the end of the vertical slide rail, which is used to absorb the impact when the horizontal slide rail moves to its limit position.

[0158] The limit switch can be a mechanical micro switch, photoelectric switch, Hall effect switch, or inductive proximity switch. For example, a waterproof mechanical micro switch can be used, featuring simple structure, high reliability, strong anti-interference capability, and suitability for outdoor environments. The limit switch works as follows: when the clamping assembly 204 moves to the preset position, the object to be clamped (such as the wall of a trash can) or the mechanical triggering device pushes the trigger rod (or striker, roller) of the limit switch, causing the internal contacts of the limit switch to actuate and output an electrical signal (e.g., changing from a low level to a high level, or from a normally open state to a normally closed state). The control module 210 detects this electrical signal to determine whether the clamping assembly 204 has reached the preset position.

[0159] The preset clamping position refers to the position corresponding to the standard clamping depth that the clamping assembly 204 is expected to reach when clamping the object to be clamped. The design of the preset clamping position needs to comprehensively consider the standard size of the object to be clamped, the clamping stability requirements, and the equipment safety margin. For example, if the clamping is too shallow (i.e., not reaching the preset position), the clamping assembly 204 may not fully wrap the trash can, posing a risk of slippage during lifting or transportation; if the clamping is too deep (i.e., exceeding the preset position), the clamping assembly 204 may excessively squeeze the trash can, causing deformation or damage to the can body. The preset clamping position corresponds to the position where the curved clamping plate pushes the trash can to the standard clamping depth, at which point the trash can wall just triggers the limit switch.

[0160] The bottom guide rail can be a guide structure located below the left and right gripping arms. The shape of the bottom guide rail can be designed according to the bottom characteristics of the object to be gripped. Specifically, the bottom guide rail is U-shaped (i.e., a U-shaped bottom guide rail), with its opening facing the rear of the robot (i.e., the direction in which the robot moves backward towards the trash can). The functions of the U-shaped bottom guide rail include: a guiding function, that is, when the robot moves backward to bring the trash can into the gripping area, the U-shaped bottom guide rail can guide the bottom rollers or the body of the trash can to smoothly enter the predetermined position, preventing the trash can from deviating due to uneven ground or robot positioning deviation; and a limiting function, that is, when the trash can is lifted, the U-shaped bottom guide rail supports or restricts the bottom of the trash can from below, preventing the trash can from swaying back and forth or left and right during the robot's movement, thus improving transportation stability.

[0161] A damping buffer is a buffer element installed at the end of a vertical slide rail (i.e., at the extreme position of the horizontal slide rail) to absorb impact energy when the horizontal slide rail reaches its extreme position. Damping buffers can be of various types, including rubber buffer blocks, polyurethane buffer pads, hydraulic buffers, or pneumatic buffers. Specifically, a hydraulic buffer is used, filled with hydraulic oil. When the piston is impacted, the hydraulic oil generates damping force through the damping orifice, converting the impact kinetic energy into heat energy, thereby achieving smooth deceleration and impact absorption. When the vertical drive component causes the horizontal slide rail to exceed its travel due to control errors or malfunctions, the damping buffer can absorb the impact energy before or simultaneously with the mechanical hard limit, preventing a violent collision between the horizontal slide rail and the mechanical hard limit, thus protecting the vertical slide rail, the horizontal slide rail, and the clamping assembly 204 mounted on them from impact damage, while also reducing collision noise.

[0162] Limit switches are installed on the side (inner side) of the left and / or right clamping arms facing the object to be clamped. These limit switches are triggered when the left and / or right clamping arms reach a preset clamping position and send a trigger signal to the control module 210. Specifically, one limit switch is installed on the inner side of each of the left and right clamping arms, employing a dual-sided independent detection design. The left limit switch is installed on the inner side of the left clamping arm, and the right limit switch is installed on the inner side of the right clamping arm. The trigger signals from both limit switches are transmitted to the control module 210, which can independently read the status of each limit switch. Dual-sided independent detection can identify unilateral abnormal conditions. For example, if the left limit switch has been triggered but the right limit switch has not, it indicates that the right clamping arm has not reached the preset position, possibly due to the trash can being misplaced (leaning to the left) or a foreign object being clamped on the right side. This ability to identify unilateral abnormalities is not available in single-sided limit switch solutions.

[0163] During the closing process of the clamping arm, the arc-shaped clamping plate first contacts the barrel wall, and the elastic buffer pad begins to compress. As the clamping arm continues to close, the position of the barrel wall relative to the clamping arm gradually changes. When the clamping arm moves to the preset clamping position, the barrel wall (or the trigger block installed on the clamping arm) pushes the trigger rod of the limit switch, and the internal contacts of the limit switch actuate, sending a trigger signal to the control module 210 (e.g., changing from a low level to a high level). After receiving this signal, the control module 210 determines the clamping completion status based on the current value.

[0164] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0165] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A robot gripping system, characterized in that, The system includes: a transmission assembly, a clamping assembly, an environmental sensing module, a body sensing module, and a control module; wherein... The transmission assembly is slidably connected to the clamping assembly; the control module is communicatively connected to the environmental sensing module, the body sensing module, the transmission assembly, and the clamping assembly, respectively. The transmission component is used to drive the clamping component to move; The clamping assembly is used to clamp the object to be clamped; the clamping assembly includes a horizontal guide rail, a left clamping arm, a right clamping arm, and a drive assembly; the left clamping arm and the right clamping arm are symmetrically arranged and respectively slide in cooperation with the horizontal guide rail; the drive assembly includes a drive motor and a synchronous transmission mechanism; the drive motor drives the left clamping arm and the right clamping arm to move synchronously towards or away from each other along the horizontal guide rail through the synchronous transmission mechanism; The environmental perception module is used to collect environmental images and point cloud data of the environment in which the clamping component is located, and send the environmental images and point cloud data to the control module; The body sensing module is used to detect the force and motion state of the clamping assembly, and send the force and motion state to the control module; the body sensing module includes a current detector and a limit switch; the current detector is used to detect the drive current output by the drive motor; the limit switch is used to detect the target event of the clamping arm moving to a preset clamping position; The control module is used to determine the target position of the object to be clamped based on the environmental image and the point cloud data. The control module is also used to control the transmission component to drive the clamping component to move to a preset range of the target position; The control module is also used to control the clamping component to perform a clamping action on the object to be clamped; The control module is further configured to determine the clamping completion status of the clamping component on the object to be clamped based on the force and motion status returned by the body sensing module during the execution of the clamping action; wherein, the control module determines the clamping completion status of the clamping component on the object to be clamped based on the force and motion status returned by the body sensing module during the execution of the clamping action; the control module, in response to the current value rising from a preset no-load current value, compares the current value of the drive motor with a preset clamping current threshold; when the current current value reaches the clamping current threshold, it obtains the current triggering state of the limit switch; if the current triggering state indicates that the limit switch has been triggered, it is determined that the clamping arm has successfully clamped the object to be clamped; if the current triggering state indicates that the limit switch has not been triggered, it is determined that the clamping arm has abnormally clamped the object to be clamped. The control module is also used to control the transmission assembly and the clamp assembly according to the completion status of the clamp.

2. The system according to claim 1, characterized in that, The control module is also used for: The environmental image is identified by a pre-trained deep learning object detection model to obtain the two-dimensional bounding box corresponding to the object to be grabbed in the environmental image. The point cloud data is projected onto the environmental image according to the calibration extrinsic parameters corresponding to the environmental perception module to obtain the point cloud projection map corresponding to the clamping component, and the corresponding candidate point cloud clusters are extracted from the point cloud projection map according to the two-dimensional bounding box. Outlier filtering is performed on the candidate point cloud cluster, and the three-dimensional spatial pose of the object to be clamped relative to the clamping component is obtained from the filtered point cloud as the target position. The target location is converted into a navigation target point in the global map coordinate system, and the transmission component is controlled to navigate to the preset range of the navigation target point; Adjust the pose of the clamping component so that the angle between the clamping component and the object to be clamped is less than a preset angle threshold.

3. The system according to claim 1, characterized in that, The control module is also used for: The environmental image is input into the feature extraction network of the deep learning object detection model. The feature extraction network includes multiple sequentially connected convolutional layers and pooling layers, which are used to output multi-scale feature maps of the environmental image. The multi-scale feature map is input into the region generation network of the deep learning object detection model. The region generation network is used to generate multiple candidate anchor boxes on the multi-scale feature map and output the confidence that each candidate anchor box contains the object to be hugged. Based on the confidence level, a target anchor box is selected from the multiple candidate anchor boxes, and the position and size of the target anchor box are adjusted by the bounding box regression layer of the deep learning object detection model to output the two-dimensional bounding box.

4. The system according to claim 2, characterized in that, The control module is also used for: Control the movement of the transmission component and pause it at a preparatory position at a distance threshold from the object to be clamped; The environmental perception module is controlled to acquire the environmental image and point cloud data at the preparatory position, and the target position is updated according to the environmental image and point cloud data corresponding to the preparatory position. The pose of the clamping component is adjusted according to the updated target position, so that the clamping component reaches a clamping preparation posture with a preset relative pose between itself and the object to be clamped. In the clamping preparation posture, the angle between the clamping component and the object to be clamped is less than the preset angle threshold.

5. The system according to claim 1, characterized in that, The control module is also used to: control the transmission component to drive the clamping component to move until the clamping center height of the clamping component matches the height of the clamping part on the object to be clamped, based on the height information of the object to be clamped; The drive motor is controlled to drive the clamping arm to close in a direction closer to the object to be clamped; The current detector is controlled to obtain the current value of the drive motor during the closing process of the clamping arm as the force condition; The trigger state of the target event output by the limit switch during the closing process of the clamping arm is obtained as the motion state.

6. The system according to claim 1, characterized in that, If it is determined that the robot's gripping of the object to be gripped is abnormal, the control module is further configured to: execute a preset fault handling strategy; the fault handling strategy includes at least one of controlling the drive motor to stop driving, sending a preset alarm message, and controlling the gripping component to retreat a preset distance and then re-grip; If the clamping component is still deemed to be clamping abnormal after the number of times it re-clamps reaches a preset threshold, the control module is further configured to: record the current position of the clamping component and report the current clamping abnormality to a preset management entity.

7. The system according to claim 1, characterized in that, The transmission assembly includes: Vertical slide rails are mounted on the chassis; A horizontal slide rail is slidably engaged with the vertical slide rail, and the clamping assembly is disposed on the horizontal slide rail; A vertical drive component is connected to the horizontal slide rail for driving the horizontal slide rail to move along the vertical slide rail.

8. The system according to claim 7, characterized in that, The ends of the left clamping arm and the right clamping arm are respectively connected to arc-shaped clamping plates, and the inner side of the arc-shaped clamping plates is fitted with an elastic buffer pad. The radius of curvature of the arc-shaped clamp is adapted to the outer wall of the object to be clamped; The elastic buffer pad is used to increase clamping friction, absorb clamping impact, and accommodate the shape tolerances and placement errors of the object to be clamped.

9. The system according to claim 8, characterized in that, The drive assembly includes a drive motor and a synchronous transmission mechanism; the synchronous transmission mechanism includes a double-sided output screw mechanism or a synchronous belt drive mechanism; the drive motor drives the left clamping arm and the right clamping arm to move synchronously in opposite directions or in opposite directions along the horizontal guide rail through the synchronous transmission mechanism.

10. The system according to claim 8, characterized in that, Limit switches are installed on the side of the left clamping arm and / or the right clamping arm facing the object to be clamped; the limit switches are used to be triggered when the left clamping arm and / or the right clamping arm reaches the preset clamping position and send a trigger signal to the object. The clamping assembly also includes a bottom guide rail, which is fixedly installed below the left clamping arm and the right clamping arm to guide the object to be clamped into the clamping area and limit the object to be clamped; the end of the vertical slide rail is provided with a damping buffer, which is used to absorb the impact when the horizontal slide rail moves to the limit position.

Images