A cage chicken farm dead chicken picking robot and a control method thereof

Abstract

The application discloses a cage laying hen farm dead chicken picking robot and a control method thereof, and relates to the technical field of agricultural robots. The robot comprises a moving platform, a lead screw guide rail, a dead chicken recycling device, a laser navigation system and a control system which are installed on the moving platform, a six-axis mechanical arm which is installed on the lead screw guide rail and can move up and down, and a depth camera which is installed at the end joint of the six-axis mechanical arm. The robot can navigate in the cage laying hen farm through the laser navigation system. The combined arrangement of the lead screw guide rail and the six-axis mechanical arm can realize image data acquisition of different heights. The control system can realize automatic control of dead chicken automatic identification and picking. The robot can realize automatic identification and picking of dead chickens in the cage laying hen farm.

Description

Technical Field

[0001] This application relates to the field of agricultural robot technology, and in particular to a robot for picking up dead chickens in caged laying hen farms and its control method. Background Technology

[0002] In the egg-laying hen farming industry, many problems remain to be solved, especially in the removal of dead chickens. Manual labor is still required to remove dead or abnormal chickens discovered during daily inspections. This process is labor-intensive, and the act of handling dead chickens can easily cause stress reactions in healthy chickens in the same cage. Furthermore, direct contact between humans and dead chickens increases the risk of zoonotic diseases. Left-behind dead chickens not only create unpleasant odors in the chicken coop but also pose a risk of further viral spread and larger-scale chicken deaths. Manually retrieving dead chickens from high-rise cages is not only physically demanding but also carries the risk of falls. To address these problems, a device capable of identifying and retrieving dead chickens is urgently needed. Summary of the Invention

[0003] The purpose of this application is to provide a robot for picking up dead chickens in caged layer hen farms and its control method, so as to realize the automatic identification and picking up of dead chickens in caged layer hen farms.

[0004] To achieve the above objectives, this application provides the following solution.

[0005] In the first aspect, this application provides a dead chicken collection robot for caged layer chicken farms, comprising: a mobile platform, a lead screw guide rail, a dead chicken collection device, a laser navigation system and a control system mounted on the mobile platform, a six-axis robotic arm mounted on the lead screw guide rail and capable of moving up and down, and a depth camera mounted at the end joint of the six-axis robotic arm. Both the laser navigation system and the depth camera are connected to the control system, which is connected to the control end of the mobile platform, the control end of the lead screw guide rail, and the control end of the six-axis robotic arm. The laser navigation system is used to scan caged egg-laying chicken farms to obtain laser scanning data of the caged egg-laying chicken farms; The depth camera is used to acquire image data of each cage in the caged layer hen farm; The control system is used to control the laser navigation system to scan the caged layer hen farm, obtain laser scan data of the caged layer hen farm, construct a grid map of the caged layer hen farm based on the laser scan data, control the mobile platform to move within the caged layer hen farm based on the grid map, and control the depth camera to capture image data of each cage in the caged layer hen farm. Based on the captured image data, control the six-axis robotic arm to pick up dead chickens; the improved Kalman filter algorithm is obtained by introducing dynamic grid velocity state into the Kalman filter algorithm.

[0006] Secondly, this application provides a control method for the aforementioned dead chicken scavenging robot in a caged layer hen farm, comprising the following steps: The laser navigation system that controls the dead chicken collection robot in the caged layer chicken farm scans the caged layer chicken farm to obtain laser scanning data of the caged layer chicken farm; Based on the laser scanning data, a grid map of a caged layer hen farm is constructed using an improved Kalman filter algorithm; the improved Kalman filter algorithm is obtained by introducing dynamic grid velocity state into the Kalman filter algorithm; The robot's mobile platform is controlled to move within the caged egg-laying hen farm according to the grid map. The robot's depth camera is controlled to capture image data of each cage in the caged egg-laying hen farm. Based on the captured image data, the robot's six-axis robotic arm is controlled to collect dead chickens.

[0007] According to the specific embodiments provided in this application, this application has the following technical effects.

[0008] This application provides a robot for collecting dead chickens in a caged layer hen farm and its control method. The robot includes a mobile platform, a lead screw guide rail, a dead chicken collection device, a laser navigation system, and a control system mounted on the mobile platform, a six-axis robotic arm mounted on the lead screw guide rail and capable of vertical movement, and a depth camera mounted at the end joint of the six-axis robotic arm. The laser navigation system enables the robot to navigate within the caged layer hen farm. The combined configuration of the lead screw guide rail and the six-axis robotic arm allows for the acquisition of image data at different heights. The control system enables automatic identification and collection of dead chickens. The robot of this application achieves automatic identification and collection of dead chickens in a caged layer hen farm. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 The present application provides a structural schematic diagram of a robot for picking up dead chickens in a caged laying hen farm, as described in one embodiment of the present application.

[0011] Figure 2 This is a flowchart illustrating the workflow of a dead chicken retrieval robot for a caged layer chicken farm, as provided in one embodiment of this application.

[0012] Figure 3A flowchart illustrating a control method for a dead chicken scavenging robot in a caged layer hen farm, as provided in one embodiment of this application.

[0013] Explanation of reference numerals in the attached figures: 1. Lead screw guide rail; 2. Lead screw slider; 3. Dead chicken recovery device; 4. Control panel; 5. Moving platform; 6. Lead screw motor and transmission base; 7. Laser navigation system; 8. Depth camera; 9. Gripper-type end effector; 10. Six-axis robotic arm. Detailed Implementation

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0015] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0016] In one exemplary embodiment, a robot for collecting dead chickens in caged egg-laying hen farms is provided, such as... Figure 1 As shown, it includes: a mobile platform 5, and a lead screw guide rail 1, a dead chicken recovery device 3, a laser navigation system 7, and a control system mounted on the mobile platform 5. A six-axis robotic arm 10, which is mounted on the lead screw guide rail 1 and capable of vertical movement, and a depth camera 8 mounted at the end joint of the six-axis robotic arm 10.

[0017] The control system is connected to control panel 4.

[0018] The mobile platform 5 adopts a two-wheel differential chassis, with two differential wheels as drive wheels and follower wheels, enabling it to move forward and backward and turn vertically. The laser navigation system 7 uses the SICK LMS100-10000 laser navigation system, which is installed at the front of the mobile platform 5. The SICK LMS100-10000 laser navigation system can provide a horizontal scanning angle of 270°, a scanning frequency of up to 25Hz or 50Hz, an angular resolution of 0.25° or 0.5°, and a working area of ​​0.5m to 20m. It scans the surrounding environment, performs obstacle detection and avoidance, thereby achieving accurate path planning and navigation.

[0019] The lead screw guide rail 1 is installed at the front end of the mobile platform 5, and its bottom is fixed to the mobile platform 5 by two right-angle connectors. The six-axis robotic arm 10 is fixed on the lead screw slider 2, which is installed on the lead screw guide rail 1 and can move along the lead screw guide rail 1 to drive the six-axis robotic arm 10 to perform vertical linear motion. A gripper-type end effector 9 is installed at the end of the six-axis robotic arm 10 to complete the tasks of opening and closing the door frame and picking up dead chickens. The depth camera 8 is a Realsense D435i depth camera, installed above the gripper-type end effector 9, and fixed to the end joint of the six-axis robotic arm 10 by a bracket, which can realize real-time identification and positioning of the door frame and dead chickens. The control panel 4 is fixed to the rear end of the mobile platform 5 by a support frame and is connected to the control system. The control system is located in the mobile platform 5.

[0020] The control system controls the laser navigation system 7 to scan the caged layer hen farm, obtain laser scan data of the caged layer hen farm, construct a grid map of the caged layer hen farm based on the laser scan data, control the mobile platform 5 to move within the caged layer hen farm based on the grid map, and control the depth camera 8 to capture image data of each chicken cage in the caged layer hen farm. Based on the captured image data, control the six-axis robotic arm 10 to pick up dead chickens; the improved Kalman filter algorithm is obtained by introducing dynamic grid velocity state into the Kalman filter algorithm.

[0021] Existing dead chicken collection equipment has the following drawbacks: It lacks a visual and convenient user interface and is deficient in terms of intelligence. Without a lifting mechanism, it cannot meet the task of collecting dead chickens in high-rise cage-raised layer chicken farms; Due to its large size, it is not suitable for operation in the narrow aisles of high-rise cage-raised layer chicken farms; The end effector structure is not suitable for most dead chicken retrieval scenarios; The cage opening and picking actions were not integrated into a single robotic arm system, resulting in structural redundancy.

[0022] The dead chicken retrieval robot in the above embodiments of this application adopts laser navigation and collects laser scanning data through SICKLMS100-10000 single-line laser radar; it identifies dead chickens and door frames based on image data, and integrates the single robotic arm operation mode of opening cages and picking up dead chickens. It can adapt well to the dead chicken retrieval environment of high-rise caged egg farms and complete the action of picking up dead chickens.

[0023] In another exemplary embodiment, such as Figure 2As shown, the working process of the dead chicken retrieval robot in the aforementioned caged layer hen farm includes: system startup and environmental scanning, map building and navigation, dead chicken identification, door frame identification, cage opening, dead chicken retrieval, and cage closing. Its specific workflow is: system startup → environmental scanning (LiDAR) → map building (Kalman filtering + grid division) → global path planning (A / B). Algorithm) → Dead Chicken Detection (YOLOv8 + Depth Camera) → Door Frame Localization (3D Coordinate Transformation) → Robotic Arm Control (Homogeneous Transformation Matrix) → Retrieval and Loop Closure Detection.

[0024] In another exemplary embodiment, during system startup and environmental scanning, the entire machine is started by manually opening the control panel 4, and laser scanning data is collected by the SICK LMS100-10000 laser navigation system to obtain an outline map of the internal environment of the egg farm, including distance and angle information of obstacles, providing basic data for map construction.

[0025] In another exemplary embodiment, during the map building and navigation process, a 2D raster map of the chicken farm is constructed using the GMAPping algorithm, particle filtering is initialized, and a map is generated. Each particle contains the robot's pose. Initial weights To update the motion model, odometry data is used. (linear velocity) angular velocity Predict particle pose and calculate particle weights: ; ; In the formula, For time intervals, For motion noise, For the first Individual particle weights For LiDAR scanning data, For the first A grid map corresponding to each particle.

[0026] Grid occupancy probability is updated using log odds: ; ; ; In the formula, For grid The logarithm of the probability, Historical observation data is used to provide prior confidence levels for the raster state. For sensor observation models, The distance between the endpoint of the laser beam and the nearest obstacle on the map is the Euclidean distance. Sensor noise parameters For random noise probability, This represents the probability of occupancy after the transformation.

[0027] according to Weighted resampling of particles retains high-weight particles to approximate the true posterior distribution. The particle map with the highest weight is selected. As the initial raster map.

[0028] After constructing the initial raster map, an improved Kalman filter is used to optimize the initial data, enhancing the map's accuracy and robustness. Dynamic raster velocity states are introduced. To optimize the ability to distinguish between static or dynamic obstacles such as chicken coops, people, and escaped chickens.

[0029] ; ; In the formula, For the first The predicted state vector at time t. This is the state vector from the previous time step. This is process noise.

[0030] ; In the formula, Based on occupancy probability and velocity noise, This is the dynamic noise gain coefficient.

[0031] LiDAR provides grid occupancy observation ,when When the grid is occupied (due to an obstacle), when The grid is unoccupied (unobstructed), and its observation equation is: ; In the formula, For the first The observed value at time, For the observation matrix, For the first The true state vector at time t, For observation noise. Its dynamic observation noise adjustment method: ; In the formula, Based on observation noise, The dust influence coefficient is calculated using the sensor's grayscale variance. estimate.

[0032] To enable the improved Kalman filter to dynamically adjust the estimation of grid state (occupancy probability and dynamic velocity) based on lidar observation information, its Kalman gain is calculated as follows: ; ; ; ; ; In the formula, For Kalman gain, To estimate the covariance matrix a priori, This is the transpose of the observation matrix, used to transform the covariance of the predicted state to the observation space, and is related to the observation noise covariance. Combined with the calculated Kalman gain, This is the state update equation, which corrects the predicted state based on the observed values. For the covariance update equation, the updated state Covariance As input for the next moment.

[0033] ; Static grid Stable, dynamic grid and Larger, output optimized map To improve map robustness.

[0034] The completed map is stored in the control system. The mobile platform 5 performs global path planning based on the established global two-dimensional map and initializes A. Algorithm parameters: Set the starting point, which is the robot's initial inspection position. The endpoint is each inspection point. .

[0035] ; In the formula, The first row and first column of the chicken coop position. The width of a single-row chicken coop. Width of a single chicken coop The distance between two adjacent sets of chicken cages. Width is 5 for mobile platforms.

[0036] In another exemplary embodiment, regarding the aspect of constructing a grid map of a caged layer hen farm using an improved Kalman filter algorithm based on the laser scanning data, the control system specifically includes: An initial grid map is constructed using the GMAPing algorithm based on the laser scanning data. An improved Kalman filter algorithm is used to update and iterate based on the initial grid map until the iteration termination condition is met to obtain the grid map. The iteration termination condition is: the occupancy probability of each grid tends to be stable, the velocity state of each grid is greater than 0, and the velocity noise is greater than the velocity noise threshold. ; ; ; ; ; in, Let be the prior state vector at time k. for The posterior state vector at time t. , for The first moment The probability of each grid cell being occupied. for The first moment The speed status of each grid cell For the number of grid cells, The state transition matrix of the raster is... , The time interval between two adjacent moments. for The noise matrix at time step, , Based on the probability of occupancy, For speed noise, The velocity state of the i-th grid cell. This is the dynamic noise gain coefficient. The process noise at time k, The Kalman gain at time k, Let be the prior estimate of the covariance matrix at time k. This is the observation matrix, where the superscript T denotes transpose. Let k be the covariance update equation at time k. , Based on observation noise, The dust impact coefficient is... The grayscale variance of the sensor; for The posterior state vector at time t. For the first The observed value at time, , for Observation noise at any given moment for The posterior estimated covariance matrix at time t. It is an identity matrix.

[0037] In another exemplary embodiment, in controlling the mobile platform to move within a caged layer hen farm according to the grid map, controlling the depth camera to capture image data of each cage within the caged layer hen farm, and controlling the six-axis robotic arm to collect dead chickens based on the image data, the control system is specifically used for: The mobile platform is controlled to move within the caged layer chicken farm according to the grid map, and the first RGB image of each chicken cage is obtained. Dead chickens are identified based on the first RGB image of each chicken cage, and the chicken cage containing dead chickens is taken as the target chicken cage. Based on the grid map, the mobile platform is controlled to move to the detection point position corresponding to the target chicken cage, and the depth camera is controlled to capture a second RGB image of the target chicken cage. Door frame recognition is performed based on the second RGB image of the target chicken cage to obtain the door frame recognition result; Based on the door frame recognition result, control the depth camera to capture a first depth image and a third RGB image of the door frame area; Based on the first depth image and the third RGB image of the door frame area, control the six-axis robotic arm to grip and pull open the door frame of the target chicken coop, and record the position of the gripping point of the door frame when it is fully opened; The depth camera is controlled to acquire a second depth image and a fourth RGB image inside the target chicken coop. Based on the second depth image and the fourth RGB image, control the six-axis robotic arm to pick up the dead chicken and put it into the dead chicken recycling device. Based on the position of the door frame clamping point when it is fully opened, control the six-axis robotic arm to clamp and close the door frame of the target chicken cage.

[0038] In another exemplary embodiment, during the process of dead chicken identification based on the first RGB images of each chicken cage, a dead chicken identification model is established using the YOLO_V8 algorithm. When the mobile platform 5 reaches an inspection point, the control system issues a command as an input signal to the lead screw motor and transmission seat 6, causing the lead screw motor and transmission seat 6 to drive the lead screw guide rail 1 to rotate, thereby driving the lead screw slider 2 to make a vertical linear motion until the six-axis robotic arm 10 reaches the target cage layer. The depth camera 8 located above the gripper-type end effector 9 acquires images of all positions in the target cage layer and uploads the acquired image data to the control system, where the dead chicken identification is performed using the trained YOLO_V8 algorithm dead chicken identification model.

[0039] In another exemplary embodiment, door frame recognition is performed based on the second RGB image of the target chicken coop. During the process of obtaining the door frame recognition result (i.e., door frame recognition), a door frame recognition model is established using the YOLO_V8 algorithm. After a dead chicken is identified, the depth camera 8 on the six-axis robotic arm 10 again acquires images of the corresponding chicken coop door frame and transmits the acquired data to the control system, where the trained YOLO_V8 algorithm door frame recognition model is used for door frame recognition.

[0040] In another exemplary embodiment, the process of controlling a six-axis robotic arm to grip and pull open the door frame of the target chicken cage based on a first depth image and a third RGB image of the door frame area, and recording the position of the door frame gripping point (i.e., opening the cage) in the fully opened state, includes the following steps 101-102.

[0041] Step 101: Based on the first depth image and the third RGB image, determine the three-dimensional coordinates of the clamping point of the target chicken cage's door frame in the base coordinate system of the six-axis robotic arm.

[0042] After identifying the door frame, in order to obtain the door frame's location in the camera coordinate system, depth camera 8 acquires the first depth image and the third RGB image of the door frame area, and extracts the pixel coordinates of the door frame clamping points using the YOLO_V8 algorithm. and depth value The pixel coordinates are converted to normalized coordinates using camera intrinsics. : ; In the formula, For camera focal length, Let be the coordinates of the camera's optical center on the door frame image plane.

[0043] Normalized coordinates combined with depth values The three-dimensional coordinates of the door frame clamping point in the camera coordinate system are obtained by the following formula: ; In the formula, The coordinates of the door frame clamping point relative to the camera coordinate system.

[0044] The homogeneous transformation matrix from the camera coordinate system to the base coordinate system of the six-axis robotic arm 10 is obtained through hand-eye calibration. The method is as follows: ; In the formula, This is a 3×3 rotation matrix describing the rotation relationship from the camera coordinate system to the robot arm base coordinate system. It is a 3×1 translation vector, representing the position of the camera origin in the base coordinate system of the six-axis robotic arm.

[0045] In actual operation, the movement of the mobile platform will be bumpy, causing the camera to have a slight displacement relative to the base or target. Therefore, dynamic calibration compensation based on visual servoing is introduced. Cooperative markers are set at fixed positions in the environment. The dynamic pose data is obtained and the dynamic pose correction of the current camera coordinate system relative to the world coordinate system is calculated in real time to correct the transformation matrix in real time and achieve high-precision positioning and grasping.

[0046] In one exemplary embodiment, to improve the positioning accuracy of the mobile platform under bumpy conditions on a chicken farm road, this application introduces a dynamic pose compensation mechanism based on visual servoing on top of static hand-eye calibration. This is achieved by pre-setting cooperative markers such as ArUco codes at fixed environmental locations, such as the side of the chicken cage frame. The three-dimensional coordinates of these markers are obtained in advance through measurement and calibration. The Perspective-n-Point (PnP) algorithm is then used to calculate the rotation matrix of the camera coordinate system relative to the marker coordinate system (i.e., the world coordinate system) at the current moment. Translation vector This allows us to obtain the camera's pose in the world coordinate system. .

[0047]

[0048]

[0049] In the formula, To establish the relationship between the coordinate system of the marker and the coordinate system of the robot arm base. for The rotation matrix represents the rotation relationship between the marker coordinate system and the base coordinate system. for The translation vector represents the position of the origin of the marker coordinate system in the base coordinate system. for The rotation matrix represents the rotation relationship from the world coordinate system to the camera coordinate system. for The translation vector represents the position of the origin of the world coordinate system in the camera coordinate system.

[0050] The actual pose of the camera relative to the robot arm base coordinate system at the current moment is calculated using dynamic correction values. for:

[0051]

[0052] Right now,

[0053] The actual pose and the static calibration value The deviation between them is the dynamic correction amount. for:

[0054] Right now:

[0055] Simplified to:

[0056] In the formula, To correct the rotation vector, we represent the real-time rotation angle of the camera's optical center relative to its ideal pose during static calibration. The correction translation vector represents the real-time translation of the camera's optical center relative to its ideal position during static calibration.

[0057] The coordinates of the door frame clamping point relative to the camera coordinate system The formula for converting to world coordinates relative to the base coordinate system of the six-axis robotic arm is as follows:

[0058] Right now:

[0059] When expanded, it appears as follows:

[0060] Step 102: Based on the three-dimensional coordinates of the clamping point of the target chicken cage door frame in the base coordinate system of the six-axis robotic arm and the three-dimensional coordinates of the end of the six-axis robotic arm in the base coordinate system of the six-axis robotic arm, control the six-axis robotic arm to clamp and pull open the door frame of the target chicken cage, and record the position of the clamping point of the door frame in the fully opened state.

[0061] The coordinates of the end effector of the six-axis robotic arm in the base coordinate system Calculate the depth distance between the door frame clamping point and the end effector of the six-axis robotic arm in the base coordinate system. The control system drives the gripper-type end effector 9 on the six-axis robotic arm 10 to adjust its position, grip and pull open the door frame to realize the cage opening action. At the same time, it records the coordinate information of the door frame relative to the moving platform 5 after the cage is opened and stores it in the control system. The calculation formula is as follows: ; In the formula, The transformed coordinates are those of the center point of the door frame relative to the coordinate system of the six-axis robotic arm's base. Let be the cosine of the angle between the base's X-axis and the camera's X-axis. Let be the cosine of the angle between the X-axis of the base and the Y-axis of the camera. Let be the cosine of the angle between the X-axis of the base and the Z-axis of the camera. Let be the cosine of the angle between the Y-axis of the base and the X-axis of the camera. Let be the cosine of the angle between the Y-axis of the base and the Y-axis of the camera. Let be the cosine of the angle between the Y-axis of the base and the Z-axis of the camera. Let be the cosine of the angle between the base's Z-axis and the camera's X-axis. Let be the cosine of the angle between the base's Z-axis and the camera's Y-axis. The cosine of the angle between the base's Z-axis and the camera's Z-axis is given. This represents the offset of the camera coordinate system origin along the X-axis of the robotic arm base coordinate system. The offset of the camera coordinate system origin from the Y-axis of the robotic arm base coordinate system. The offset of the camera coordinate system origin from the Z-axis of the robotic arm base coordinate system.

[0062] In another exemplary embodiment, the process of controlling a six-axis robotic arm to pick up a dead chicken (i.e., dead chicken retrieval) based on a second depth image and a fourth RGB image includes the following steps 201-202.

[0063] Step 201: Determine the three-dimensional coordinates of the dead chicken's center point in the base coordinate system of the six-axis robotic arm based on the fourth RGB image and the second depth image.

[0064] Depth camera 8 re-captures a depth image of the dead chicken and extracts the pixel coordinates of the center point of the dead chicken. and depth value The pixel coordinates are converted to normalized coordinates using camera intrinsics. Normalized coordinates combined with depth values The three-dimensional coordinates of the dead chicken's center point in the camera coordinate system are obtained using the following formula: ; ; In the formula, The coordinates of the dead chicken's center point relative to the camera coordinate system. For camera focal length, Let be the coordinates of the camera's optical center on the dead chicken image plane.

[0065] From the above formula, we can see that the three-dimensional coordinates of the dead chicken's center point in the camera coordinate system are: The coordinates are then transformed to the base coordinate system relative to the six-axis robotic arm. The transformation method is as follows:

[0066] When unfolded, it is as follows:

[0067] Step 202: Based on the three-dimensional coordinates of the dead chicken's center point in the base coordinate system of the six-axis robotic arm and the three-dimensional coordinates of the end point of the six-axis robotic arm in the base coordinate system of the six-axis robotic arm, control the six-axis robotic arm to pick up the dead chicken and put it into the dead chicken recycling device.

[0068] The coordinates of the end effector of the six-axis robotic arm at this moment in the base coordinate system Calculate the depth distance between the center point of the dead chicken and the end effector of the six-axis robotic arm in the base coordinate system. After locating the center of the dead chicken, the control system drives the hand-type end effector 9 on the six-axis robotic arm 10 to adjust its posture, grab the dead chicken and put it into the dead chicken collection device 3 to complete the dead chicken collection.

[0069] ; In the formula, The coordinates of the dead chicken's center point relative to the coordinate system of the six-axis robotic arm's 10-base base are transformed. The translation vector is 3×1, representing the position of the camera origin in the coordinate system of the six-axis robotic arm 10 base when the Realsense D435i depth camera 8 acquires the dead chicken image target.

[0070] In another exemplary embodiment, during the process of controlling the six-axis robotic arm to clamp and close the door frame of the target chicken cage (i.e., close the cage) based on the position of the door frame clamping point in the fully opened state, the control system pushes the door frame to close the chicken cage according to the recorded position of the door frame clamping point in the fully opened state, and the six-axis robotic arm 10 returns to the initialization state.

[0071] Compared with the prior art, the beneficial effects of the above embodiments of this application are as follows: This application uses a control panel as a visual operation interface, which enhances human-computer interaction and improves the intelligence level of the product; The embodiments of this application use lifting devices such as lead screws and guide rails to drive the movement of the robotic arm, which expands the spatial operating range of the robotic arm to a certain extent, enabling it to better adapt to the environment of dead chicken collection in high-rise cage-raising chicken farms. This application embodiment integrates the actions of opening the cage and picking up dead chickens into a single robotic arm system, reducing structural redundancy and effectively reducing product development costs. The embodiments of this application are based on a robot dead chicken retrieval process using a single robotic arm system, which enables the robot to complete the dead chicken retrieval task more efficiently, reduces human intervention, and improves the safety and efficiency of the operation.

[0072] The robot in this embodiment occupies a small volume and uses a narrow mobile platform, which can better handle operations in the narrow passageways of high-rise cage-raised layer chicken farms.

[0073] Based on the same inventive concept, this application also provides a control method for the aforementioned dead chicken scavenging robot in a caged layer hen farm. The solution provided by this control method is similar to the solution described in the robot above; therefore, the specific limitations in one or more control method embodiments provided below can be found in the limitations of the robot described above, and will not be repeated here.

[0074] In one exemplary embodiment, a control method for a robot that collects dead chickens in a caged layer hen farm is provided, such as... Figure 3 As shown, it includes the following steps 301-303.

[0075] Step 301: Control the laser navigation system of the dead chicken collection robot in the caged egg-laying chicken farm to scan the caged egg-laying chicken farm and obtain laser scanning data of the caged egg-laying chicken farm.

[0076] Step 302: Based on the laser scanning data, a grid map of the caged layer chicken farm is constructed using an improved Kalman filter algorithm; the improved Kalman filter algorithm is obtained by introducing dynamic grid velocity state into the Kalman filter algorithm.

[0077] Step 303: Control the mobile platform of the dead chicken collection robot in the caged egg-laying hen farm to move within the caged egg-laying hen farm according to the grid map, and control the depth camera of the dead chicken collection robot in the caged egg-laying hen farm to capture image data of each chicken cage in the caged egg-laying hen farm, and control the six-axis robotic arm of the dead chicken collection robot in the caged egg-laying hen farm to collect dead chickens according to the captured image data.

[0078] In another exemplary embodiment, step 303 described above can be replaced by the following steps.

[0079] The mobile platform is controlled to move within the caged layer chicken farm according to the grid map, and the first RGB image of each chicken cage is acquired; dead chickens are identified based on the first RGB image of each chicken cage, and the chicken cages containing dead chickens are taken as target chicken cages. Based on the grid map, the mobile platform is controlled to move to the detection point position corresponding to the target chicken cage, and the depth camera is controlled to capture a second RGB image of the target chicken cage. Door frame recognition is performed based on the second RGB image of the target chicken coop to obtain the door frame recognition result; Based on the door frame recognition result, control the depth camera to capture a first depth image and a third RGB image of the door frame area; Based on the first depth image and the third RGB image of the door frame area, control the six-axis robotic arm to grip and pull open the door frame of the target chicken coop, and record the position of the gripping point of the door frame when it is fully opened; The depth camera is controlled to acquire a second depth image and a fourth RGB image inside the target chicken coop. Based on the second depth image and the fourth RGB image, control the six-axis robotic arm to pick up the dead chicken and put it into the dead chicken recycling device. Based on the position of the door frame clamping point when it is fully opened, control the six-axis robotic arm to clamp and close the door frame of the target chicken cage.

[0080] 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 specification.

[0081] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A robot for collecting dead chickens in caged egg-laying hen farms, characterized in that, The dead chicken collection robot for caged laying hen farms includes: a mobile platform, a lead screw guide rail, a dead chicken collection device, a laser navigation system and a control system mounted on the mobile platform, a six-axis robotic arm mounted on the lead screw guide rail and capable of moving up and down, and a depth camera mounted at the end joint of the six-axis robotic arm. Both the laser navigation system and the depth camera are connected to the control system, which is connected to the control end of the mobile platform, the control end of the lead screw guide rail, and the control end of the six-axis robotic arm. The laser navigation system is used to scan caged egg-laying chicken farms to obtain laser scanning data of the caged egg-laying chicken farms; The depth camera is used to acquire image data of each cage in the caged layer hen farm; The control system is used to control the laser navigation system to scan the caged layer hen farm, obtain laser scan data of the caged layer hen farm, construct a grid map of the caged layer hen farm based on the laser scan data, control the mobile platform to move within the caged layer hen farm based on the grid map, and control the depth camera to capture image data of each cage in the caged layer hen farm. Based on the captured image data, control the six-axis robotic arm to pick up dead chickens; the improved Kalman filter algorithm is obtained by introducing dynamic grid velocity state into the Kalman filter algorithm.

2. The dead chicken retrieval robot for caged layer chicken farms according to claim 1, characterized in that, In terms of constructing a grid map of a caged layer hen farm using an improved Kalman filter algorithm based on the laser scanning data, the control system specifically includes: An initial grid map is constructed using the GMAPing algorithm based on the laser scanning data. An improved Kalman filter algorithm is used to update and iterate based on the initial grid map until the iteration termination condition is met to obtain the grid map. The iteration termination condition is: the occupancy probability of each grid tends to be stable, the velocity state of each grid is greater than 0, and the velocity noise is greater than the velocity noise threshold. ； ； ； ； ； in, Let be the prior state vector at time k. for The posterior state vector at time t. , for The first moment The probability of each grid cell being occupied. for The first moment The speed status of each grid cell For the number of grid cells, The state transition matrix of the raster is... , The time interval between two adjacent moments. for The noise matrix at time step, , Based on the probability of occupancy, For speed noise, The velocity state of the i-th grid cell. This is the dynamic noise gain coefficient. The process noise at time k, The Kalman gain at time k, Let be the prior estimate of the covariance matrix at time k. This is the observation matrix, where the superscript T denotes transpose. Let k be the covariance update equation at time k. , Based on observation noise, The dust impact coefficient is... The grayscale variance of the sensor; for The posterior state vector at time t. For the first The observed value at time, , for Observation noise at any given moment for The posterior estimated covariance matrix at time t. It is an identity matrix.

3. The dead chicken retrieval robot for caged layer chicken farms according to claim 1, characterized in that, In terms of controlling the mobile platform to move within the caged layer hen farm according to the grid map, controlling the depth camera to capture image data of each cage within the caged layer hen farm, and controlling the six-axis robotic arm to collect dead chickens based on the image data, the control system is specifically used for: The mobile platform is controlled to move within the caged layer hen farm according to the grid map, and the first RGB image of each cage is obtained; Dead chickens are identified based on the first RGB image of each chicken cage, and the chicken cage containing dead chickens is taken as the target chicken cage. Based on the grid map, the mobile platform is controlled to move to the detection point position corresponding to the target chicken cage, and the depth camera is controlled to capture a second RGB image of the target chicken cage. Door frame recognition is performed based on the second RGB image of the target chicken coop to obtain the door frame recognition result; Based on the door frame recognition result, control the depth camera to capture a first depth image and a third RGB image of the door frame area; Based on the first depth image and the third RGB image of the door frame area, control the six-axis robotic arm to grip and pull open the door frame of the target chicken coop, and record the position of the gripping point of the door frame when it is fully opened; The depth camera is controlled to acquire a second depth image and a fourth RGB image inside the target chicken coop. Based on the second depth image and the fourth RGB image, control the six-axis robotic arm to pick up the dead chicken and put it into the dead chicken recycling device. Based on the position of the door frame clamping point when it is fully opened, control the six-axis robotic arm to clamp and close the door frame of the target chicken cage.

4. The dead chicken retrieval robot for caged layer chicken farms according to claim 3, characterized in that, Based on the first depth image and the third RGB image of the door frame area, the six-axis robotic arm is controlled to grip and pull open the door frame of the target chicken coop, specifically including: Based on the first depth image and the third RGB image, determine the three-dimensional coordinates of the clamping point of the target chicken cage's door frame in the base coordinate system of the six-axis robotic arm; Based on the three-dimensional coordinates of the clamping point of the target chicken cage door frame in the base coordinate system of the six-axis robotic arm and the three-dimensional coordinates of the end of the six-axis robotic arm in the base coordinate system of the six-axis robotic arm, the six-axis robotic arm is controlled to clamp and pull open the door frame of the target chicken cage, and the position of the clamping point of the door frame in the fully opened state is recorded.

5. The dead chicken retrieval robot for caged layer chicken farms according to claim 4, characterized in that, Based on the first depth image and the third RGB image, the three-dimensional coordinates of the door frame clamping point of the target chicken cage in the base coordinate system of the six-axis robotic arm are determined, specifically including: Based on the third RGB image, the door frame clamping point is identified, and the pixel coordinates of the door frame clamping point are determined. The depth value of the door frame clamping point is determined based on the pixel coordinates of the clamping point and the first depth image. Based on the pixel coordinates and depth value of the door frame clamping point, determine the three-dimensional coordinates of the door frame clamping point in the camera coordinate system; The three-dimensional coordinates of the door frame clamping point in the camera coordinate system are transformed to obtain the three-dimensional coordinates of the door frame clamping point in the base coordinate system of the six-axis robot arm.

6. The dead chicken retrieval robot for caged layer chicken farms according to claim 4, characterized in that, The model for identifying dead chickens based on the first RGB image of each chicken cage is called the dead chicken recognition model, and the model for identifying door frames based on the second RGB image of the target chicken cage is called the door frame recognition model. Both the dead chicken recognition model and the door frame recognition model are based on the YOLO_V8 algorithm.

7. The dead chicken retrieval robot for caged layer chicken farms according to claim 4, characterized in that, Based on the second depth image and the fourth RGB image, a six-axis robotic arm is controlled to pick up the dead chicken and place it into a dead chicken recycling device, specifically including: The three-dimensional coordinates of the dead chicken's center point in the base coordinate system of the six-axis robotic arm are determined based on the fourth RGB image and the second depth image. Based on the three-dimensional coordinates of the dead chicken's center point in the base coordinate system of the six-axis robotic arm and the three-dimensional coordinates of the end point of the six-axis robotic arm in the base coordinate system of the six-axis robotic arm, the six-axis robotic arm is controlled to pick up the dead chicken and put it into the dead chicken recycling device.

8. The dead chicken retrieval robot for caged layer chicken farms according to claim 7, characterized in that, The three-dimensional coordinates of the dead chicken's center point in the base coordinate system of the six-axis robotic arm are determined based on the fourth RGB image and the second depth image, specifically including: Based on the fourth RGB image, the center point of the dead chicken is identified, and the pixel coordinates of the center point of the dead chicken are determined. The depth value of the dead chicken's center point is determined based on the second depth image and the pixel coordinates of the dead chicken's center point. Based on the pixel coordinates and depth value of the dead chicken's center point, determine the three-dimensional coordinates of the dead chicken's center point in the camera coordinate system; The three-dimensional coordinates of the dead chicken's center point in the camera coordinate system are transformed to determine the three-dimensional coordinates of the dead chicken's center point in the base coordinate system of the six-axis robotic arm.

9. A control method for a dead chicken retrieval robot in a caged layer hen farm as described in any one of claims 1-8, characterized in that, Includes the following steps: The laser navigation system that controls the dead chicken collection robot in the caged layer chicken farm scans the caged layer chicken farm to obtain laser scanning data of the caged layer chicken farm; Based on the laser scanning data, a grid map of a caged layer hen farm is constructed using an improved Kalman filter algorithm; the improved Kalman filter algorithm is obtained by introducing dynamic grid velocity state into the Kalman filter algorithm; The robot's mobile platform is controlled to move within the caged egg-laying hen farm according to the grid map. The robot's depth camera is controlled to capture image data of each cage in the caged egg-laying hen farm. Based on the captured image data, the robot's six-axis robotic arm is controlled to collect dead chickens.

10. The control method for the dead chicken retrieval robot in a caged layer hen farm according to claim 9, characterized in that, The robot's mobile platform is controlled to move within the caged egg-laying hen farm according to the grid map. The robot's depth camera is controlled to capture image data of each cage within the farm. Based on the captured image data, the robot's six-axis robotic arm is controlled to collect dead chickens. Specifically, this includes: The mobile platform is controlled to move within the caged layer hen farm according to the grid map, and the first RGB image of each cage is obtained; Dead chickens are identified based on the first RGB image of each chicken cage, and the chicken cage containing dead chickens is taken as the target chicken cage. Based on the grid map, the mobile platform is controlled to move to the detection point position corresponding to the target chicken cage, and the depth camera is controlled to capture a second RGB image of the target chicken cage. Door frame recognition is performed based on the second RGB image of the target chicken coop to obtain the door frame recognition result; Based on the door frame recognition result, control the depth camera to capture a first depth image and a third RGB image of the door frame area; Based on the first depth image and the third RGB image of the door frame area, control the six-axis robotic arm to grip and pull open the door frame of the target chicken coop, and record the position of the gripping point of the door frame when it is fully opened; The depth camera is controlled to acquire a second depth image and a fourth RGB image inside the target chicken coop. Based on the second depth image and the fourth RGB image, control the six-axis robotic arm to pick up the dead chicken and put it into the dead chicken recycling device. Based on the position of the door frame clamping point when it is fully opened, control the six-axis robotic arm to clamp and close the door frame of the target chicken cage.

Images