Horse farm feeding robot navigation system and feeding method thereof

Abstract

The invention discloses a fixed-point navigation method for a horse house feeding robot, and aims to improve the positioning precision and task execution stability of the robot in a horse house environment. According to the method, a two-dimensional code recognition technology and a traditional sensor system are combined, and accurate positioning and dynamic pose adjustment of the robot in a complex and dynamically-changing breeding environment can be achieved. The method is implemented by the following steps of: firstly, acquiring a grid map by using a two-dimensional laser radar in a horse house environment; the robot plans a walking path between feeding troughs according to the grid map; after the feeding trough point is reached, the robot adjusts the movement path in real time by recognizing a preset two-dimensional code of the feeding trough, and it is ensured that the discharging port can be aligned with the feeding trough for feeding; and accurate quantitative feeding is carried out according to the preset feeding time. According to the method, a robot mapping navigation technology and a computer vision method are fused, the problem that feed is wasted due to the fact that a feeding robot cannot accurately position a feeding trough is solved, and the technical vacancy of a horse house feeding robot navigation technology and feeding trough positioning is filled.

Description

Technical Field

[0001] The present invention relates to a navigation and feeding control technology for horse stable feeding robots, in particular to a control method for horse stable feeding robots based on fixed-point navigation, which is applicable to the automated breeding industry. Background Art

[0002] With the rapid development of automation and intelligent technologies, the application of intelligent robots in the breeding industry has gradually increased. Especially in animal breeding environments such as horse stables, horse stable feeding robots have gradually become an important tool for improving production efficiency and reducing labor costs. Traditional horse stable feeding robots mainly rely on built-in navigation systems (such as lidar, sensors, etc.) for path planning and positioning. However, these systems often have certain limitations, especially in complex and irregular breeding environments, where it is difficult to achieve high-precision positioning and dynamic adjustment.

[0003] Currently, many horse stable feeding robots use technologies such as lidar, vision sensors, and inertial measurement units (IMUs) to perceive the surrounding environment and perform positioning. These systems can map the environment and plan paths, but in some horse stables with complex and rapidly changing environments, problems such as inaccurate positioning and path planning failure may occur due to factors such as environmental occlusion, sensor errors, and complex terrain. In addition, traditional positioning systems often rely on pre-set environmental features and static maps and cannot adapt to the minute changes in the horse stable environment in real time, resulting in a decrease in positioning accuracy and stability. When a horse stable feeding robot performs a task, it often needs to make dynamic adjustments according to the actual situation of the task. For example, when the robot reaches near the feeding point, due to environmental interference, equipment failure, or other unforeseen factors, the position of the robot may deviate. If relying on traditional sensor positioning systems, the robot often has difficulty immediately detecting and correcting the deviation. Through a QR code recognition system, the robot can quickly compare with the target position after receiving the QR code information and automatically adjust the path and pose according to the current state. This process not only improves the working accuracy of the robot but also enables the robot to continuously and efficiently operate in a complex environment. Summary of the Invention

[0004] The present invention aims to solve the problem of how to achieve fixed-point navigation and pose adjustment for accurate feeding of a feeding robot in a horse stable, and provides a navigation system and a feeding method for a horse stable feeding robot. The feeding robot avoids obstacles in the road based on the environmental information obtained by sensors in the horse stable aisle and performs real-time path planning to prevent collisions. When the robot reaches the feeding trough, it uses a QR code to assist in accurately adjusting the pose to ensure that the discharge port is aligned with the feeding trough.

[0005] Technical Solution:

[0006] The present invention first discloses a navigation system for a horse ranch feeding robot, including a feeding robot and a navigation system, wherein:

[0007] The feeding robot includes: a feed bin, a discharge port, a lidar, a depth camera, and a robot chassis. An edge device is fixed inside the robot chassis as the main control of the entire system. The feed bin is arranged above the robot chassis. The robot chassis is equipped with four DC motors, which use belt drive. The DC motors are directly controlled by a motion controller, and the controller is controlled by the PWM wave generated by an ESP32s3 single-chip microcomputer. By adjusting the amplitude and frequency of the waveform, the rotation speed of the motors of the robot chassis is controlled, realizing the displacement of the feeding robot; the discharge port is tubular and extends from the feed bin. There is a screw conveyor connecting the discharge port and the feed bin. The screw conveyor uses a motor as the power source, and the motor is also powered by the chassis power system. When powered on, the screw conveyor rotates to drive the feed to be transported from the feed bin to the discharge port. The lidar and the depth camera are arranged at the front end of the feeding robot, used to obtain environmental point clouds and image information for positioning the AprilTag code;

[0008] The navigation system includes:

[0009] Hardware:

[0010] (1) 3D lidar

[0011] (2) IMU built in the lidar

[0012] (3) Depth camera

[0013] Software:

[0014] (1) AMCL positioning system

[0015] (2) Navigation2 navigation plugin

[0016] (3) AprilTag two-dimensional code marking system

[0017] Among them, the hardware part is centrally installed on the front side of the robot chassis and rigidly connected to the chassis. The 3D lidar is connected to the chassis main control through an RJ485 network cable, which is used for power supply of the motor inside the lidar and data transmission on the local area network. The motor inside the lidar drives the laser emitter to rotate, and the internal laser receiver is used to receive the reflected laser. By calculating the time difference between emission and reception, the relative position between the lidar and the object is calculated. The lidar is built-in with an inertial sensor (IMU) to correct the errors caused by sensor drift and tire slippage during SLAM mapping. The depth camera is installed directly in front of the chassis and connected to the chassis main control through a USB-Type-c data cable, and its viewing angle is parallel to the ground. When the robot moves near the AprilTag code, the depth camera will transmit the acquired image to the chassis main control for pose fine-tuning. The software part involves Adaptive Monte Carlo Localization (AMCL) and the Navigation2 plugin. AMCL is a localization algorithm based on particle filtering, which is used to estimate the position and pose of the robot in a known map. It represents the position hypothesis of the robot by generating a large number of particles. Each particle has a position and a weight, and the weight represents the matching degree between this position and the actual observation. Through sensor data, AMCL gradually adjusts the distribution of particles and finally determines the most likely pose of the robot. Navigation2 is a navigation framework in ROS2, which provides a complete set of functions to help mobile robots achieve autonomous navigation. It includes core functions such as path planning, local obstacle avoidance, map construction, and localization, and supports the navigation of the robot from one point to another. Navigation2 is compatible with different hardware platforms, sensors (such as lidar, IMU, etc.) and controllers. It is a key component for realizing the autonomous navigation of the robot in the ROS2 system. AprilTag is a two-dimensional barcode marking system, which is used for robot vision positioning and pose estimation. It consists of a series of black and white patterns, and each tag has a unique identifier. By identifying these tags with a camera, the robot can accurately estimate the position and direction of the tags.

[0018] The present invention also discloses a navigation method for a horse farm feeding robot, comprising the following steps:

[0019] S1. Use a lidar to perform SLAM mapping in the horse barn to obtain a two-dimensional grid map of the horse barn environment;

[0020] S2. Start the horse barn feeding navigation task;

[0021] S3. The feeding robot arrives near the feeding point, and performs fine-tuning of the feeding pose and position of the robot;

[0022] S4. Feed horse feed;

[0023] S5. Detect whether all feeding target point feeding tasks have been completed. If completed, return to the starting position of the robot and wait for the start of the next feeding task. If not completed, continue to navigate to the next feeding target point until the task is completed.

[0024] Specifically, the step S1 specifically includes:

[0025] S1-1. Enable the lidar to scan the horse stable environment and construct a two-dimensional grid map;

[0026] S1-2. Manually correct the obtained grid map, eliminate the noise points in the map, and save the corrected map in the map file.

[0027] Specifically, the step S2 specifically includes:

[0028] S2-1. Load the grid map obtained in step S1 into the map server of the Navigation2 navigation plugin;

[0029] S2-2. Initialize the lidar and its built-in IMU sensor and depth camera;

[0030] S2-3. Calibrate the initial pose of the robot chassis, give the initial pose of "2D Pose" in the Navigation2 plugin, and align the obstacles recognized by the radar with the obstacles in the grid map;

[0031] S2-4. Given the navigation target point, call the A* algorithm module in the Navigation2 navigation plugin to plan the global path, and update the path in real time when encountering obstacles to implement obstacle avoidance. In the A* algorithm, the heuristic function can guide the graph search algorithm to move towards the target point faster. For a grid map, when the robot can move in eight directions on a plane, the heuristic function usually selects the Chebyshev distance:

[0032] h(node) = Chebyshev(node) = max(abs(x goal -x node ), abs(y goal -y node ))

[0033] Where xgoal and xnode represent the x coordinates of the target point and the current node, and ygoal and ynode represent the y coordinates of the target point and the current node. In the A* algorithm, the global cost f(node) is defined as follows:

[0034] f(node) = g(node) + h(node)

[0035] Among them, g(node) is the actual cost from the current node to the target node, which can be calculated through the information pre-stored in the two-dimensional grid map. Through iterative calculation, the global cost map can be obtained. Based on the global cost map, the robot can generate the optimal path from the starting point to the target point.

[0036] Specifically, step S3 specifically includes:

[0037] S3-1. Start the depth camera for scanning, adopt the AprilTag two-dimensional code marking recognition system, and the chassis main control obtains the robot's spatial pose information through the decoding algorithm provided by the AprilTag library;

[0038] S3-2. Adjust the position and pose of the robot according to the AprilTag code information near the feeding target point, and place the discharge port directly above the trough.

[0039] Specifically, in S3-1, the spatial pose information of the robot relative to the AprilTag code is calculated by the pose calculation algorithm, including the relative spatial distance and the relative spatial angle. The calculation formulas for the relative spatial angle and the relative spatial distance are as follows;

[0040]

[0041] θ x =tanh -1 (R 32 / R 33 )

[0042]

[0043] θ z =tanh -1 (R 21 / R 11 )

[0044] t = ωt'

[0045] Among them, θ is the relative angle of the camera coordinate system relative to the AprilTag code coordinate system, t is the distance from the camera coordinate system to the AprilTag code coordinate system, R is the obtained rotation matrix, t' is the distance from the camera to the AprilTag code with an actual side width of a unit square in the same direction, and ω is the known actual side width of the AprilTag code.

[0046] Specifically, in S3-2, the relative position r and pose θ' between the robot coordinate system F C and the trough coordinate system F are obtained through coordinate transformation according to the calculated AprilTag code information near the feeding target point. The specific solution process is as follows:

[0047] Let the homogeneous transformation matrix between the robot coordinate system F solved by the AprilTag and after coordinate transformation and the feeder trough coordinate system F be r and the feeder trough coordinate system F C be

[0048]

[0049] where the vector represents the translation of F r relative to F C , that is, the relative position); the matrix R = [R ij represents the rotation relationship between the two coordinate systems; the relative pose θ’ is represented in Euler angles and is calculated according to the following formula:

[0050] Yaw angle, rotation around the z-axis:

[0051] θ z = arctan2(R 21 , R 11 )

[0052] Pitch angle, rotation around the y-axis:

[0053]

[0054] Roll angle, rotation around the x-axis:

[0055] θ x = arctan2(R 32 , R 33 )

[0056] θ’ = (θ x , θ y , θ z ) is the relative rotation information between F r and F C ;

[0057] Then, use the PID negative feedback adjustment algorithm to adjust the pose and position of the robot. The adjustment of the PID algorithm is as follows:

[0058]

[0059] Δθ’(t) = θ’ m - θ’(t)

[0060]

[0061] where μ(t) is used as the control variable of the motor, is the target position, θ′ m is the target pose angle, is the position deviation, Δθ’(t) is the pose angle deviation, Kp , K i , K d are the proportional, integral, and derivative controller coefficients of the PID algorithm respectively; finally, the motor control variable is input into the motion controller to adjust the spatial state of the robot so that the discharge port of the robot is directly above the feed trough.

[0062] Specifically, step S4 specifically includes:

[0063] S4-1. When the task of S3 is completed, start the feeding motor to drive the auger to feed and put the feed into the manger;

[0064] S4-2. Select the preset feeding time according to the decoding information obtained in step S3. When the feeding time meets the set target feeding time, immediately stop the motor rotation to complete the feeding, so as to achieve the on-demand feeding of the mangers with different AprilTag code information.

[0065] Specifically, in S4-2, select the corresponding feeding time t0 according to the decoding information obtained in step S3. When the feeding time t reaches the set target feeding time t0, immediately stop the motor rotation to stop the feeding. At this time, the amount of feed fed is η:

[0066]

[0067] where ρ is the average density of horse feed and δ is the cross-sectional area of the auger discharge port, and finally the amount of feed η for each fed trough is calculated.

[0068] Advantages of the present invention

[0069] The present invention applies the multi-target tracking algorithm to the automatic feeding of horse barn feed, improves the positioning accuracy and navigation reliability of the horse barn feeding robot, and ensures that the robot can accurately reach the designated feeding point. The technology disclosed by the present invention provides technical support for the automatic feeding direction in the field of intelligent horse breeding, can reduce manual intervention, save labor, and improve breeding efficiency. Description of the drawings

[0070] Figure 1 is the mapping, navigation, and feeding flow chart of the horse barn feeding robot based on the laser-vision sensor of the present invention.

[0071] Figure 2 is the structure diagram of the feeding robot of the present invention. Specific embodiments

[0072] The following combines the drawings and embodiments to further illustrate the present invention. In the embodiments of the present invention, a feeding robot navigation and feeding method is provided, which uses a laser-vision sensor to construct a two-dimensional grid map of the horse barn environment and uses AprilTag two-dimensional codes to assist the robot in fine-tuning its pose and quantitatively feeding.

[0073] The feeding robot is as Figure 2 shown, including: a feed bin 2, a discharge port 1, a lidar 3, a depth camera 4, and a robot chassis 5. An edge device is fixed inside the robot chassis as the main controller of the entire system. The feed bin 2 is arranged above the robot chassis 5. The robot chassis 5 is equipped with four DC motors, which use belt drive. The DC motors are directly controlled by a motion controller, and the controller is controlled by the PWM wave generated by an ESP32s3 single-chip microcomputer. By adjusting the amplitude and frequency of the waveform, the rotation speed of the robot chassis motor is controlled to achieve the displacement of the feeding robot; the discharge port 1 is tubular and extends from the feed bin 2. There is a screw conveyor connecting the discharge port 1 and the feed bin 2. The screw conveyor is powered by a motor, and the motor is also powered by the chassis power system. When powered on, the screw conveyor rotates to drive the feed from the feed bin 2 to the discharge port 1. The lidar 3 and the depth camera 4 are arranged at the front end of the feeding robot to obtain environmental point clouds and image information for positioning the AprilTag code.

[0074] The overall flowchart of the feeding method of the feeding robot is as Figure 1 shown, mainly including the following steps (S1 to S5):

[0075] S1. Use the Figure 2 3D lidar installed on the front side of the chassis as shown to perform SLAM mapping in the horse stable to obtain a two-dimensional grid map of the horse stable environment;

[0076] S1-1. Enable the lidar to scan the horse stable environment and construct a two-dimensional grid map;

[0077] S1-2. Select a truncation height, project the three-dimensional point cloud map onto a two-dimensional plane to obtain an original two-dimensional grid map;

[0078] S1-3. Manually correct the obtained grid map, eliminate the noise points in the map, mark the feeding locations in the map and obtain the coordinates of the feeding locations, and finally save the corrected map in the map file;

[0079] S2. Start the horse stable feeding navigation task;

[0080] S2-1. Load the grid map obtained in step S1 into the map server of the Navigation2 navigation plugin;

[0081] S2-2. Initialize sensors such as the lidar, depth camera, and IMU;

[0082] S2-3. Use the Navigation2 plugin to calibrate the initial pose of the robot chassis, give the initial pose of "2D Pose", and align the obstacles recognized by the radar with the obstacles in the grid map;

[0083] S2-4. Given a navigation target point, the Navigation2 navigation plugin plans a global path based on the A* algorithm, and updates the path in real time when encountering obstacles to implement obstacle avoidance;

[0084] S3. When approaching the feeding point, finely adjust the pose and position of the robot at the feeding position;

[0085] S3-1. When the robot chassis reaches near the feeding trough, an AprilTag code appears in the field of view of the depth camera located on the front side of the chassis. The chassis main control calculates the spatial position information of the robot relative to the AprilTag code using an attitude solution algorithm. The calculation formulas for the relative spatial angle and relative spatial distance are as follows;

[0086] θ x =tanh -1 (R 32 / R 33 )

[0087]

[0088] θ z =tanh -1 (R 21 / R 11 )

[0089] t = ωt'

[0090] where θ is the relative angle of the camera coordinate system relative to the AprilTag code coordinate system, t is the distance from the camera coordinate system to the AprilTag code coordinate system, R is the obtained rotation matrix, t' is the distance from the camera to the AprilTag code with an actual side width of a unit square in the same direction, and ω is the known actual side width of the AprilTag code.

[0091] S3-2. Near the feeding target point, obtain the relative position r between the robot coordinate system F C and the feeding trough coordinate system F and the pose θ' through coordinate transformation according to the calculated AprilTag code information, and then use the PID negative feedback adjustment algorithm to adjust the pose and position of the robot. The adjustment of the PID algorithm is as follows;

[0092]

[0093] Δθ'(t) = θ' m -θ'(t)

[0094]

[0095] where μ(t) is used as the control variable of the motor, is the target position, θ' m is the target pose angle, is the position deviation, Δθ'(t) is the pose angle deviation, K p , K i , K d are the proportional, integral, and derivative controller coefficients of the PID algorithm respectively; finally, the motor control variable is input to the motion controller to adjust the spatial state of the robot so that the discharge port of the robot is directly above the feed trough.

[0096] S4. Feed the horse feed in the robot's feed trough at the designated feeding point.

[0097] S4-1. After completing step S3, start the feeding motor. The rotation of the motor drives the spiral blade of the auger to rotate through the belt. The rotation speed of the auger spiral blade is collected by the encoder, and the feed in the robot's silo is controlled to be lifted to the discharge port at a uniform speed v0.

[0098] S4-2. Select the corresponding feeding time t0 according to the decoding information obtained in step S3. When the feeding time t reaches the set target feeding time t0, immediately stop the motor rotation, that is, stop feeding. At this time, the amount of feed fed is η.

[0099]

[0100] where ρ is the average density of the horse feed, δ is the cross-sectional area of the auger discharge port, and finally the amount of feed η for each fed feed trough is calculated.

[0101] S5. Detect whether all feeding target point feeding tasks have been completed. If completed, return to the starting position of the robot and wait for the next feeding task to start. If not completed, continue to navigate to the next feeding target point until the task is completed.

[0102] The specific embodiments described in this article are only illustrative of the spirit of the present invention. Those skilled in the art of the present invention can make various modifications or supplements to the described specific embodiments or use similar methods to replace them, but will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.

Claims

1. A navigation system for a horse ranch feeding robot, characterized in that It includes a feeding robot and a navigation system, where: The feeding robot includes: a feed bin (2), a discharge port (1), a lidar (3), a depth camera (4), and a robot chassis (5). An edge device is fixed inside the robot chassis as the main controller of the entire system. The feed bin (2) is arranged above the robot chassis (5). The robot chassis (5) is equipped with four DC motors, which use belt drive. The DC motors are directly controlled by a motion controller, and the controller is controlled by the PWM wave generated by an ESP32s3 single-chip microcomputer. By adjusting the amplitude and frequency of the waveform, the rotation speed of the robot chassis motor is controlled to achieve the displacement of the feeding robot. The discharge port (1) is tubular and extends from the feed bin (2). There is a screw conveyor connecting the discharge port (1) and the feed bin (2). The screw conveyor uses a motor as the power source, and the motor is also powered by the chassis power system. When powered on, the screw conveyor rotates to drive the feed to be transported from the feed bin (2) to the discharge port (1). The lidar (3) and the depth camera (4) are arranged at the front end of the feeding robot and are used to obtain environmental point clouds and image information for positioning the AprilTag code. The lidar (3) is built-in with an inertial sensor IMU to correct the errors caused by sensor drift and tire slippage during SLAM mapping. The lidar (3) and the depth camera (4) are the hardware parts of the navigation system and collect signals for the navigation system. The software part of the navigation system includes an AMCL positioning system, a Navigation2 navigation plugin, and an AprilTag two-dimensional code marking system, where: The AMCL positioning system is used to estimate the position and pose of the feeding robot in a known map. The Navigation2 navigation plugin is used to achieve autonomous navigation of the robot. The AprilTag two-dimensional code marking system is used for visual positioning and pose estimation of the feeding robot to assist the feeding robot in accurately estimating the position and direction of the tag.

2. A feeding method based on a navigation system of a horse ranch feeding robot, characterized in that It includes the following steps: S1. Use a lidar to perform SLAM mapping in the horse barn to obtain a two-dimensional grid map of the horse barn environment. S2. Start the horse barn feeding navigation task. S3. The feeding robot arrives near the feeding point, and the feeding pose and position of the robot are finely adjusted. S4. Feed horse feed. S5. Detect whether all feeding target point feeding tasks have been completed. If completed, return to the starting position of the robot and wait for the next feeding task to start. If not completed, continue to navigate to the next feeding target point until the task is completed.

3. The method according to claim 2, wherein The specific steps of step S1 include: S1-1. Enable the lidar to scan the horse barn environment and construct a two-dimensional grid map. S1-2. Manually correct the obtained grid map to eliminate noise points in the map, and save the corrected map in the map file.

4. The method according to claim 2, wherein The specific steps of step S2 include: S2-1. Load the grid map obtained in step S1 into the map server of the Navigation2 navigation plugin. S2-2. Initialize the lidar and its built-in IMU sensor and depth camera. S2-3. Calibrate the initial pose of the robot chassis. Given the initial pose of "2D Pose" in the Navigation2 plugin, align the environmental obstacles recognized by the radar with the obstacles in the grid map. S2-4. Given the navigation target point, call the A* algorithm module in the Navigation2 navigation plugin to plan the global path, and update the path in real time when encountering obstacles to implement obstacle avoidance.

5. The method according to claim 4, characterized in that In the A* algorithm, the heuristic function guides the graph search algorithm to move towards the target point faster. For the grid map, when the feeding robot moves in eight directions on the plane, the Chebyshev distance is selected as the heuristic function: h(node) = Chebyshev(node) = max(abs(x goal - x node ), abs(y goal - y node )) where xgoal and xnode represent the x coordinates of the target point and the current node, and ygoal and ynode represent the y coordinates of the target point and the current node. In the A* algorithm, the global cost f(node) is defined as follows: f(node) = g(node) + h(node) where g(node) is the actual cost from the current node to the target node, which is calculated through the information pre-stored in the two-dimensional grid map. Through iterative calculation, the global cost map can be obtained. According to the global cost map, the feeding robot can generate the optimal path from the starting point to the target point.

6. The method according to claim 2, wherein The specific steps of step S3 include: S3-1. Start the depth camera for scanning, and use the AprilTag QR code marking recognition system. The chassis main control obtains the robot's spatial pose information through the decoding algorithm provided by the AprilTag library. S3-2. Adjust the position and pose of the robot according to the AprilTag code information near the feeding target point, and place the discharge port directly above the feed trough.

7. The method according to claim 6, wherein In S3-1, the spatial pose information of the robot relative to the AprilTag code is calculated using the pose solution algorithm, including the relative spatial distance and relative spatial angle. The calculation formulas for the relative spatial angle and relative spatial distance are as follows: θ x =tanh -1 (R 32 / R 33 ) θ z =tanh -1 (R 21 / R 11 ) t = ωt′ where θ is the relative angle between the camera coordinate system and the AprilTag code coordinate system, t is the distance from the camera coordinate system to the AprilTag code coordinate system, R is the obtained rotation matrix, t′ is the distance from the camera to the AprilTag code with an actual side width of one unit square in the same direction, and ω is the known actual side width of the AprilTag code.

8. The method according to claim 6, wherein In S3-2, the robot coordinate system F is obtained through coordinate transformation based on the calculated AprilTag code information near the feeding target point r and the relative position C with the feeding trough coordinate system F and pose θ', and the specific solution process is as follows: Let the homogeneous transformation matrix between the robot coordinate system F solved by the AprilTag and after coordinate transformation and the feeder coordinate system F r be C ​ Among them, the vector represents the translation of F r relative to F C , that is, the relative position; the matrix R = [R ij represents the rotation relationship between two coordinate systems; the relative pose θ’ is represented in the form of Euler angles and is calculated according to the following formula: Yaw angle, rotation around the z-axis: θ z = arctan2(R 21 , R 11 ) Pitch angle, rotation around the y-axis: Roll angle, rotation around the x-axis: θ x = arctan2(R 32 , R 33 ) θ’ = (θ x , θ y , θ z ) is the relative rotation information between F r and F C ; Then use the PID negative feedback adjustment algorithm to adjust the pose and position of the robot. The adjustment of the PID algorithm is as follows: Δθ′(t) = θ′m - θ’(t) where μ(t) is the control variable of the motor, is the target position, and θ’ m is the target pose angle, is the position deviation, Δθ’(t) is the pose angle deviation, and K p , K i , K d are the proportional, integral, and derivative controller coefficients of the PID algorithm respectively; finally, the motor control variable is input to the motion controller to adjust the spatial state of the robot so that the discharge port of the robot is directly above the trough.

9. The method according to claim 2, wherein The specific steps of step S4 include: S4-1. When the S3 task is completed, start the feeding motor to drive the auger to feed and put the feed into the manger. S4-2. Select the preset feeding time according to the decoding information obtained in step S3. When the feeding time meets the set target feeding time, immediately stop the motor rotation to complete the feeding, achieving on-demand feeding of the mangers with different AprilTag code information.

10. The method according to claim 9, wherein In S4-2, select the corresponding feeding time t0 according to the decoding information obtained in step S3. When the feeding time t reaches the set target feeding time t0, immediately stop the motor rotation, that is, stop feeding. At this time, the amount of feed fed is η: Where ρ is the average density of horse feed and δ is the cross-sectional area of the auger discharge port. Finally, the amount of feed η in each fed trough is calculated.

Images