Cover cutting type corn row inter-plant weeding robot and control method
Through the combination of hood-cut structure and extended Kalman filtering and self-correction controller, combined with deep learning technology, the precise weeding of corn weeding robots in corn fields is achieved, and the problems of low weeding efficiency and plant damage in the existing technology are solved. , improving agricultural production efficiency and mechanization level.
Patent Information
- Application Number
- CN202411315372.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-20
- Publication Date
- 2024-11-08
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing corn weeding robots have difficulty performing weeding operations accurately in narrow corn fields, resulting in inefficiency and damage to corn plants, and it is difficult to achieve inter-row and inter-plant weeding at the same time, increasing the number of repeated operations and restricting the weeding efficiency.
Using a hood-cut structure and an extended Kalman filtering combined with a self-correction controller, the precise protection and miscellaneous corn seedlings are achieved through the mobile chassis, positioning cover seedling robotic arm and a hood-cutting inter-plant weeding end effector. Effective cleaning of grass, deep learning and image segmentation technology for seedling recognition and robotic arm control.
It improves the accuracy and efficiency of weeding in corn fields, reduces damage to corn plants, reduces labor intensity and environmental pollution, and improves the level of agricultural mechanization.
Smart Images

Figure CN118901377A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of agricultural robot technology, and in particular to a weeding robot for cutting between rows of corn plants and its control method. Background Technology
[0002] For a long time, agricultural production has been severely affected by weeds, as they compete with crops for sunlight, water, and nutrients, leading to decreased crop yields and quality, and causing huge losses to the agricultural economy. Corn, in particular, as one of the world's most important food crops, is frequently disrupted by weeds during its growing season. Traditional weeding methods require a large amount of manpower and are inefficient, potentially damaging crops. Therefore, the development of corn weeding robots has become a topic of great interest. These robots can automatically perform weeding operations in narrow cornfields, thereby reducing manual labor intensity, improving weeding efficiency, and reducing reliance on pesticides, meeting the requirements of sustainable agricultural development.
[0003] Compared to traditional manual and chemical weeding, mechanical weeding is more efficient, capable of completing large-scale weeding operations in a short time, reducing reliance on pesticides, and thus contributing to reduced environmental pollution and aligning with the requirements of sustainable agricultural development. However, the design of corn weeding robots faces several challenges. First, the robot needs to perform precise weeding operations in confined cornfields, ensuring effective weed removal while minimizing damage to corn plants, which places high demands on the robot's operational performance. Second, existing end effectors struggle to simultaneously perform weeding between rows and between individual plants, increasing the number of repetitive operations and severely limiting weeding efficiency. Therefore, the rational design of the end effector for corn row-to-plant weeding robots is crucial. An optimized end effector design ensures precise weeding operations in narrow cornfields, guaranteeing effective weed removal while minimizing damage to corn plants. Such an optimized design can effectively improve weeding operations in cornfields, increase agricultural production efficiency, reduce labor intensity, and minimize negative environmental impacts, thereby promoting the advancement of agricultural mechanization. Summary of the Invention
[0004] The purpose of this invention is to provide a weeding robot for corn rows and plants with a cover-cutting mechanism and a control method. The weeding robot is dexterous and can move in cornfields without damaging the seedlings while performing weeding between rows and plants. It adopts a cover-cutting structure to protect corn seedlings from damage and reduce the risk of adjacent corn plants being injured by the weeding blade. It innovatively uses a combination of extended Kalman filtering and a self-calibrating controller to control the movement of the robotic arm.
[0005] This invention provides a weeding robot for inter-row weeding of corn and its control method, including a mobile chassis for the weeding robot, a positioning and covering mechanical arm, an end effector for inter-row weeding with a covering cut, and a control system. The mobile chassis for the weeding robot is installed at the bottom, the positioning and covering mechanical arm is installed above the mobile chassis for the weeding robot, the end effector for inter-row weeding with a covering cut is installed below the positioning and covering mechanical arm, and the control system is located on one side of the mobile chassis for the weeding robot.
[0006] Preferably, the mobile chassis of the weeding robot includes four wheel-driven steering mechanisms and a chassis support frame, with the four wheel-driven steering mechanisms hinged at the four end positions of the chassis support frame.
[0007] Preferably, the wheel-driven steering mechanism includes a traveling wheel, a wheel connecting axle, a wheel fixing frame, a traveling drive motor, a steering fixing shaft, a bearing positioning plate, a steering fixing bearing, a steering motor fixing plate, and a steering drive motor. The traveling drive motor is bolted to the outside of the wheel fixing frame. The wheel fixing frame and the steering fixing shaft are integrally formed. The traveling wheel is fixed together with the wheel connecting axle and connected to the motor shaft of the traveling drive motor through the wheel connecting axle. The steering drive motor is bolted to the top of the steering motor fixing plate, and the side of the steering motor fixing plate is welded to the chassis support frame. The bearing positioning plate is welded to the outside of the chassis support frame. Two steering fixing bearings are screwed onto the upper and lower sides of the bearing positioning plate. The steering fixing shaft passes from bottom to top through the center hole of the steering fixing bearing and the bearing positioning plate and is connected to the motor hole of the steering drive motor through a key.
[0008] Preferably, the seedling positioning and covering robotic arm includes a robotic arm support frame, a seedling longitudinal translation mechanism, a seedling lateral translation mechanism, and a seedling vertical telescopic mechanism; the seedling longitudinal translation mechanism is installed above the robotic arm support frame, the seedling lateral translation mechanism is installed above the seedling longitudinal translation mechanism, and the seedling vertical telescopic mechanism is installed below the seedling lateral translation mechanism.
[0009] Preferably, the longitudinal translation mechanism for the seedling cover includes a pulley positioning bearing housing, a synchronous pulley, a pulley fixing shaft, a synchronous belt, a bearing housing fixing block, a slider connecting plate, a slide rail, a slide rail slider, a motor drive shaft, a coupling, a motor fixing plate, a longitudinal translation motor, and a longitudinal translation unit; the bearing housing fixing block is fixedly connected to the upper part of the robotic arm support frame, and the pulley positioning bearing housing is screwed to the upper part of the bearing housing fixing block; on one side of the longitudinal translation motor, the longitudinal translation motor is bolted to the motor fixing plate, and the motor fixing plate is fixedly connected to the upper part of the robotic arm support frame, moving longitudinally... The translational motor extends to both sides, followed by a coupling, a motor drive shaft, a synchronous pulley, and a pulley positioning bearing seat. The motor shaft of the longitudinal translational motor is connected to the motor drive shaft via the coupling. The motor drive shaft passes through the synchronous pulley and is fixed by the positioning bearing seat. The synchronous belt and the synchronous pulleys on both sides form a synchronous belt drive. From top to bottom, the slider connecting plate, the lower synchronous belt, and the slide rail slider are fixedly connected. The slide rail slider is installed on the slide rail track, which is fixed above the robotic arm support frame. The two ends of the longitudinal translational unit are respectively fixed above the slider connecting plates on both sides.
[0010] Preferably, the vertical telescopic mechanism of the seedling cover includes an electric push rod fixing plate and a vertical telescopic electric push rod. The vertical telescopic electric push rod is fixedly connected to the electric push rod fixing plate, and the electric push rod fixing plate is fixed below the horizontally moving unit by bolt connection.
[0011] Preferably, the control system includes a RealSense D435i camera, a seedling positioning sensor, an infrared distance sensor, a robotic arm motor drive, a Raspberry Pi 4B, an STM32F411, a Beidou navigation and positioning module, a Wi-Fi module, a gyroscope, and a speed sensor.
[0012] Preferably, in step S1, weeding begins, and the weeding robot moves forward along the direction of corn weeding operation under the electric drive of the mobile chassis.
[0013] S2. When a Realsense D435i depth camera is positioned in front of the weeding robot, the Realsense D435i depth camera detects weeds in the direction of travel using a SAM-based weed and seedling joint recognition model. The weeding robot then continues to move forward, and the weed status is set to 1. Then, the seedling positioning sensor located on top of the weeding end effector detects seedlings in its field of view and sends the data back to the host computer, causing the weeding robot to stop moving forward.
[0014] The S3, seedling positioning sensor, infrared sensor, and robot's own posture sensor gyroscope acquire information and process it through an extended Kalman filter. The STM32f411 uses an adaptive control strategy to control the seedling-finding robotic arm to move horizontally, vertically, and longitudinally. The movement of the robotic arm keeps the center of the camera image on the corn seedling, thus continuously aligning it with the corn seedling.
[0015] S4. The seedling protection mechanism of the inter-row weeding end effector covers the corn seedlings under the contraction of the telescopic opening and closing mechanism, thus achieving the function of surrounding and protecting the corn seedlings.
[0016] S5. Driven by the weeding drive mechanism, the segmented telescopic weeding blade of the rotary cutting mechanism can rotate and cut the weeds next to the corn seedlings, thus clearing the weeds between the rows.
[0017] S6. The telescopic opening and closing mechanism opens and separates from the corn plant, completing one weed clearing operation.
[0018] Therefore, the present invention adopts the above-mentioned cover-cutting inter-row weeding robot and control method for corn. The weeding robot is dexterous and can move in the cornfield without damaging the seedlings, while performing inter-row weeding. The cover-cutting structure protects the corn seedlings from damage and reduces the risk of adjacent corn plants being injured by the weeding blade. The invention innovatively adopts a combination of extended Kalman filtering and self-calibrating controller to control the movement of the robotic arm.
[0019] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0020] Figure 1 This is a schematic diagram showing the location of the key components of a corn row weeding robot and its control method, which is a cover-cutting type.
[0021] Figure 2 (a) is a schematic diagram of the structure of the mobile chassis 1 of the weeding robot of the present invention, which is a weeding robot and control method for inter-row weeding of corn.
[0022] Figure 2 (b) is a schematic diagram of the wheel drive steering mechanism 101 of the corn row weeding robot and control method of the present invention.
[0023] Figure 3 (a) is a schematic diagram of the positioning and covering mechanical arm 2 of the corn row weeding robot and control method of the present invention;
[0024] Figure 3 (b) is a schematic diagram of the longitudinal translation mechanism 202 of the corn row weeding robot and control method of the present invention.
[0025] Figure 3 (c) is a schematic diagram of the lateral translation mechanism 203 of the corn row weeding robot and control method of the present invention.
[0026] Figure 3(d) is a schematic diagram of the vertical telescopic mechanism 204 for the corn row weeding robot device and method of the present invention.
[0027] Figure 4 This is a schematic diagram of the end effector for the inter-row weeding robot and control method of corn row-to-row weeding according to the present invention.
[0028] Figure 5 This is a schematic diagram of the weeding robot's travel route in the corn row-to-row weeding robot and its control method of the present invention.
[0029] Figure 6 This is a control flowchart of the mobile chassis 1 of the weeding robot, which is a corn row-to-row weeding robot and control method of the present invention.
[0030] Figure 7 (a) is a control block diagram of the seedling-finding robotic arm 2 in the horizontal plane xoy of the corn row-to-row weeding robot and control method of the present invention.
[0031] Figure 7 (b) is a control block diagram of the seedling-finding robotic arm 2 in the horizontal plane xoy of the corn row-to-row weeding robot and control method of the present invention.
[0032] Figure 8 The middle part is a control block diagram of the seedling-finding robotic arm 2 in the horizontal plane xoy of the corn row-to-row weeding robot and control method of the present invention.
[0033] Figure 9 This is a flowchart illustrating the overall system control of the corn row-to-row weeding robot and its control method according to the present invention.
[0034] Figure 10 (a) is the camera view of the front camera when the robot moves forward, which is the control method of the corn row weeding robot of the present invention.
[0035] Figure 10 (b) shows the seedling segmentation results based on SAM for the Slicing Robot and Control Method for Inter-row Weeding of Corn in the present invention.
[0036] Figure Labels
[0037] 1. Weeding robot chassis; 2. Positioning and seedling-covering robotic arm; 3. Cutting-type inter-row weeding end effector; 4. Control system; 101. Wheel drive steering mechanism; 102. Chassis support frame; 111. Walking wheels; 112. Wheel connecting axle; 113. Wheel fixing frame; 114. Walking drive motor; 115. Steering fixing axle; 116. Bearing positioning plate; 117. Steering fixing bearing; 118. Steering motor fixing plate; 119. Steering drive motor; 201 202. Mechanical arm support frame; 203. Seedling cover longitudinal translation mechanism; 204. Seedling cover lateral translation mechanism; 205. Seedling cover vertical telescopic mechanism; 216. Pulley positioning bearing seat; 217. Synchronous pulley; 218. Pulley fixed shaft; 219. Synchronous belt; 210. Bearing seat fixing block; 2110. Slider connecting plate; 212. Slide rail; 213. Slide rail slider; 214. Motor connecting shaft; 225. Coupling; 226. Motor fixing plate; 227. Longitudinal translation motor 223. Longitudinal translational unit; 224. Lateral translational motor; 225. Lateral translational unit; 226. Electric push rod fixing plate; 227. Vertical telescopic electric push rod; 301. Seedling protection mechanism; 302. Contour-following weeding mechanism; 303. End effector support frame; 304. Weed cutting drive mechanism; 311. Contour-following telescopic weeding knife; 312. Seedling cover opening and closing mechanism; 313. Cutting rotary disc; 314. Contour-following curved disc; 315. Contour-following weeding bracket; 316. 317. Seedling protection motor; 318. Robotic arm connecting frame; 409. Contour weeding motor; 4001. RealSense D435i camera; 401. Seedling positioning sensor (K210 camera); 402. Infrared distance sensor; 403. Robotic arm motor drive; 404. Raspberry Pi 4B; 405. STM32F411; 406. Beidou navigation and positioning module; 407. WiFi module; 408. Gyroscope; 419. Speed sensor. Detailed Implementation
[0038] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.
[0039] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.
[0040] Example 1
[0041] like Figures 1-10As shown, the present invention discloses a weeding robot device and method for inter-row weeding of corn, including a weeding robot mobile chassis 1, a seedling positioning and covering mechanical arm 2, a weeding end effector 3 for inter-row weeding, and a control system 4. The weeding robot mobile chassis 1 is installed at the bottom, the seedling positioning and covering mechanical arm 2 is installed above the weeding robot mobile chassis 1, the weeding end effector 3 for inter-row weeding is installed below the seedling positioning and covering mechanical arm 2, and the control system 4 is located on one side of the weeding robot mobile chassis.
[0042] The mobile chassis 1 of the weeding robot includes four wheel-driven steering mechanisms 101 and a chassis support frame 102. The four wheel-driven steering mechanisms 101 are hinged at the four end positions of the chassis support frame 102.
[0043] The wheel-driven steering mechanism 101 includes a traveling wheel 111, a wheel connecting axle 112, a wheel mounting bracket 113, a traveling drive motor 114, a steering fixed axle 115, a bearing positioning plate 116, a steering fixed bearing 117, a steering motor mounting plate 118, and a steering drive motor 119. The traveling drive motor 114 is bolted to the outside of the wheel mounting bracket 113. The wheel mounting bracket 113 and the steering fixed axle 115 are integrally formed. The traveling wheel 111 is fixedly connected to the wheel connecting axle 112 and connected to the wheel via the wheel connecting axle 112. The drive motor 114 is connected to the motor shaft, and the steering drive motor 119 is fixed to the steering motor mounting plate 118 by bolts. The steering motor mounting plate 118 is welded to the chassis support frame 102 on the side. The bearing positioning plate 116 is welded to the outside of the chassis support frame 102. Two steering fixed bearings 117 are screwed to the upper and lower sides of the bearing positioning plate 116. The steering fixed shaft 115 passes through the center hole of the steering fixed bearing 117 and the bearing positioning plate 116 from bottom to top, and is connected to the motor hole of the steering drive motor 119 by a key.
[0044] The seedling positioning and covering robotic arm includes a robotic arm support frame 201, a seedling longitudinal translation mechanism 202, a seedling lateral translation mechanism 203, and a seedling vertical telescopic mechanism 204. The seedling longitudinal translation mechanism 202 is installed above the robotic arm support frame 201, the seedling lateral translation mechanism 203 is installed above the seedling longitudinal translation mechanism 202, and the seedling vertical telescopic mechanism 204 is installed below the seedling lateral translation mechanism 203.
[0045] The longitudinal translation mechanism 202 includes a pulley positioning bearing seat 211, a synchronous pulley 212, a pulley fixing shaft 213, a synchronous belt 214, a bearing seat fixing block 215, a slider connecting plate 216, a slide rail 217, a slide rail slider 218, a motor drive shaft 219, a coupling 220, a motor fixing plate 221, a longitudinal translation motor 222, and a longitudinal translation unit 223. The bearing seat fixing block 215 is fixedly connected to the top of the robotic arm support frame 201, and the pulley positioning bearing seat 211 is screwed onto the top of the bearing seat fixing block 215. On one side of the longitudinal translation motor 222, the longitudinal translation motor 222 is bolted to the motor fixing plate 221, which is fixedly connected to the top of the robotic arm support frame 201. Extending from the longitudinal translation motor 222 to both sides are, in sequence, the coupling 220, the motor drive shaft 219, the synchronous pulley 212, and the pulley positioning bearing seat 211. The longitudinal translation motor 222's motor shaft is connected to the motor drive shaft 219 via a coupling 220. The motor drive shaft 219 passes through the synchronous pulley 212 and is finally fixed by the positioning bearing seat 211. The synchronous belt 214 and the synchronous pulleys 212 on both sides form a synchronous belt drive. The slider connecting plate 216, the lower synchronous belt 214, and the slide rail slider 218 are fixed together from top to bottom to achieve synchronous longitudinal translation. The slide rail slider 218 is installed on the slide rail track 217 and can achieve longitudinal reciprocating sliding. The slide rail track 217 is fixed above the robotic arm support frame 201. The two ends of the longitudinal translation unit 223 are respectively fixed above the slider connecting plates 216 on both sides.
[0046] The lateral translation mechanism 203 includes a pulley positioning bearing seat 211, a synchronous pulley 212, a pulley fixing shaft 213, a synchronous belt 214, a bearing seat fixing block 215, a slider connecting plate 216, a slide rail 217, a slide rail slider 218, a coupling 220, a motor fixing plate 221, a lateral translation motor 224, and a lateral translation unit 225. The bearing seat fixing block 215 is fixedly connected above the longitudinal translation unit 223, and the pulley positioning bearing seat 211 is screwed above the bearing seat fixing block 215. On one side of the lateral translation motor 224, the lateral translation motor 224 is bolted to the motor fixing plate 221, which is fixedly connected above the longitudinal translation unit 223. Extending from the lateral translation motor 224 to both sides are couplings. The device consists of a 220, a pulley fixed shaft 213, a pulley positioning bearing seat 211, a synchronous pulley 212, and a pulley positioning bearing seat 211. The motor shaft of the transverse translation motor 224 is connected to the pulley fixed shaft 213 via a coupling 220. After passing through the synchronous pulley 212, the pulley fixed shaft 213 is finally fixed by the positioning bearing seats 211 on both sides of the synchronous pulley 212. The synchronous belt 214 and the synchronous pulleys 212 on both sides form a synchronous belt drive. The slider connecting plate 216, the lower synchronous belt 214, and the slide rail slider 218 are fixed together from top to bottom. The slide rail slider 218 is installed on the slide rail track 217. The slide rail track 217 is fixed above the longitudinal translation unit 223. The two ends of the transverse translation unit 225 are respectively fixed above the slider connecting plates 216 on both sides.
[0047] The vertical telescopic mechanism 204 for seedlings includes an electric push rod fixing plate 226 and a vertical telescopic electric push rod 227; wherein, the vertical telescopic electric push rod 227 is fixedly connected to the electric push rod fixing plate 226, and the electric push rod fixing plate 226 is fixed below the transverse translational unit 225 by bolt connection.
[0048] The control system includes a RealSense D435i camera 401, a seedling positioning sensor 402, an infrared distance sensor 403, a robotic arm motor drive 404, a Raspberry Pi 4B 405, an STM32F411 406, a Beidou navigation and positioning module 407, a Wi-Fi module 408, a gyroscope, and a speed sensor 409.
[0049] In the process of locating and covering corn seedlings, the weeding robot operates on uneven soil surfaces. Because the soil is not compacted, its tires are prone to getting stuck in grooves, sometimes sinking deeper. This results in constantly changing 6D attitudes, with uncertain pitch and rotation angles. Combined with water droplets and other uncertainties in the field, this creates numerous disturbances when accurately locating corn seedlings and weeds and controlling the robot arm's movement. During control, an Extended Kalman Filter (EKF) is first used to estimate the system state to handle measurement noise and system uncertainties. The EKF is a variant of the Kalman Filter, offering better performance for state estimation in nonlinear systems. It approximates the operation of the Kalman Filter by linearizing the nonlinear model and is faster than other nonlinear Kalman Filters, resulting in better efficiency in information processing. It has two phases: the prediction phase and the update phase.
[0050] The first stage is the prediction phase. In this stage, EKF uses a nonlinear system model to predict the current state and error covariance. The state prediction is as follows (1):
[0051]
[0052] in, It is the predicted state at time k. It is the estimated state at time k-1, u k-1 It controls the input.
[0053] The covariance of the prediction error is shown in equation (2) below:
[0054]
[0055] in, It is the prediction error covariance, F k Is The Jacobian matrix of the state transition function obtained at point P k-1 It is the error covariance estimate of the previous time step, Q k It is the process noise covariance.
[0056] Then comes the update phase, in which the EKF updates the predicted state and error covariance based on the observed data. This requires observation updates, Kalman gain calculations, state updates, and error covariance updates. The update formula for the observations is shown in equation (3) below:
[0057]
[0058] Among them, z k Here, h is the actual observed value, and v is the observation model function. k It is observation noise. Kalman gain calculation K k As shown in equation (4):
[0059]
[0060] Among them, H k Is The Jacobian matrix of the observation model function obtained at point R k It is the observation noise covariance.
[0061] The state update is shown in equation (5) below:
[0062]
[0063] in, It is the estimated state at time k, K k This is the Kalman gain calculation at time k.
[0064] The error covariance is updated as shown in equation (6):
[0065]
[0066] Where I is the identity matrix.
[0067] After continuous state estimation using the extended Kalman filter, an adaptive control algorithm is needed to adjust the controller parameters based on the state estimates to optimize the system response, ultimately outputting the controller's control input. Adaptive control dynamically adjusts the control strategy based on these estimates, ensuring control performance is unaffected by internal system dynamics and external disturbances. The adaptive control algorithm of this invention is a self-tuning controller (STR), which automatically adapts to changing system dynamics and adjusts its control parameters based on real-time feedback, improving operational robustness. In the self-tuning controller, the output of the extended Kalman filter serves as the current system state to adjust the control parameters. Based on this state, recursive least squares (RLS) is used to estimate the controller parameters θ. The RLS formula is as follows:
[0068]
[0069] Among them, y k It is a control target. It is by Derived design matrix.
[0070] Simultaneously, gain updates and covariance updates are performed, as shown in equations (8) and (9), respectively:
[0071]
[0072] The control law is then updated based on the parameters estimated by RLS. The updated control law is as follows:
[0073]
[0074] Performance monitoring continues to use system output y k To adjust future control behavior based on deviations from the control objective:
[0075] e k =r k -y k (11)
[0076] Where r k This is the reference output.
[0077] Figure 7 (a) describes the control process of the seedling-finding robotic arm 2 searching for corn seedlings in the xoy plane. A K210 camera was selected as the seedling position sensor because the complex, layered corn seedlings made direct 6D estimation and depth estimation unsuitable. Therefore, only the K210 camera was used to capture the object's position on the horizontal plane, calculating its distance to the horizontal center point of the end effector (structure 3) for row-to-row weeding, ensuring the center of the corn seedling is aligned with the center of the end effector. The camera is installed in the top recess of the cone-shaped support frame 303 of the end effector. When the end effector is lowered to the working position, the camera is 200-400mm above the ground; when the end effector is in the initial working position, the camera probe is 300-1000mm above the ground. Here, (x0, y0) are the pixel center coordinates of the K210 camera, (x... c ,y c (x) represents the current 2D position of the center point of the corn seedling within the camera, transmitted via communication between the camera and the STM32F411 controller. c ,y cThe information is then transmitted to the robotic arm motor drive 404 via the self-calibrating controller (STR) to control the speed of the longitudinal and lateral translation motors. This, in turn, changes the horizontal position of the weeding end effector in the 3-section weeding row-to-row weeding mechanism through the seedling covering longitudinal and lateral translation mechanisms. While controlling the positioning and seedling covering robotic arm to align the weeding end effector horizontally with the corn seedling, it is also necessary to control the end effector to stop within a certain height z0 above the seedling. The distance sensor at the bottom of the end effector is an infrared sensor 403 (Sharp GP2Y0A21YK0F), installed in the top groove of the cone-shaped part of the end effector support frame 303, near the K210 camera mounting position. Its measurement range is 10-80 cm, suitable for close-range measurement. z0 is the preset distance to the robotic arm placement height above the seedling top (set to 50 mm). c Current distance. (Passage) Figure 7 (b) The robot arm control process along the z-axis completes the height descent control.
[0078] At this time, the system status is:
[0079] x k =[x c ,y c ,z c ] T (12)
[0080] The reference output is:
[0081] r k =[x0,y0,z0] T (13)
[0082] Through the above control process, after the weeding end effector 3 is aligned with the corn seedling and descends to a certain height z0, it wraps around the seedling and closes the opening and closing mechanism.
[0083] The overlay-type inter-row weeding end effector 3 is installed below the positioning and overlaying robotic arm 2, and is fixedly connected to the bottom end of the vertical telescopic electric actuator 227 via a hole on the upper part of the robotic arm connecting frame 317. The overlay-type inter-row weeding end effector 3 includes a overlay protection mechanism 301, a contour weeding mechanism 302, an end effector support frame 303, and a weeding drive mechanism 304. The overlay protection mechanism 301 is installed inside the end effector support frame 303, which remains stationary. The rotation of the overlay protection mechanism 301 is controlled by the overlay protection motor 316. The contour weeding mechanism 302 is installed outside the end effector support frame 303, and its rotation is controlled by the contour weeding motor 318 in the weeding drive mechanism 304. The weeding drive mechanism 304 is installed above the contour weeding mechanism 302 and is fixed together by the vertical telescopic electric actuator 227 and the robotic arm connecting frame 317.
[0084] The seedling protection mechanism 301 includes a seedling opening and closing mechanism 312 and a seedling protection motor 316; wherein, by controlling the rotational movement of the motor shaft of the seedling protection motor 316, the seedling opening and closing mechanism 312 is extended and retracted to wrap the seedling.
[0085] The contour weeding mechanism 302 includes a contour telescopic weeding blade 311, a cutting rotary wheel 313, and a contour weeding support 315. The cutting rotary wheel 313 is fixedly connected to the contour weeding support 315. The contour telescopic weeding blade 311 is installed below the cutting rotary wheel 313 and rotates with the cutting rotary wheel 313. It converts the change in the track shape of the contour curved disk 314 into the extension and retraction of the blade head of the contour telescopic weeding blade 311, thereby realizing contour cutting.
[0086] The grass-cutting drive mechanism 304 includes a robotic arm connecting frame 317 and a contour-following weeding motor 318. The contour-following weeding motor 318 is fixedly connected above the robotic arm connecting frame 317, and the other end of the robotic arm connecting frame 317 is fixedly connected to the vertical telescopic electric actuator 227, thereby connecting the end effector 3 for weeding between rows and plants and the seedling-finding robotic arm 2. Through mechanical transmission, the rotational motion of the output shaft of the contour-following weeding motor 318 can be converted into the rotational cutting motion of the contour-following weeding mechanism 302.
[0087] After the seedling protection mechanism 301 covers the corn seedlings, the seedling opening and closing mechanism 312 retracts under the counterclockwise rotation of the seedling protection motor 316, thus protecting the corn seedlings. Next, driven by the weeding drive mechanism 304, the high-speed rotating contour-following telescopic weeding blade 311 of the contour-following weeding mechanism 302 cuts the weeds surrounding the corn seedlings, clearing weeds between rows. Finally, the seedling protection motor 316 rotates clockwise, and the contour-following weeding mechanism 302 opens and separates from the corn seedlings, completing one weeding operation. The specific weeding process can be seen in [the diagram / document / description]. Figure 8 The specific control flow diagram of the overlay-type inter-row weeding end effector 3.
[0088] Considering the different soil compaction levels when the weeding blade enters or does not enter the soil, different types of blades can be used for weeding operations. Preferably, the segmented telescopic weeding blade 207 can be designed as an arc blade, V-shaped blade, L-shaped blade, straight blade, spiral blade, or curved / curved blade, etc.
[0089] Based on the above-mentioned weeding robot chassis 1 movement control, seedling-finding and weeding robotic arm 2 positioning control, and weeding process of the weeding row-cutting end effector 3, the overall system control flow can be summarized. Figure 9 .
[0090] In addition, this weeding robot is equipped with a Beidou navigation and positioning module and a Wi-Fi module. The Beidou navigation module obtains the robot's real-time location in the field and transmits the information to the Raspberry Pi, which is especially useful in vast fields, helping operators to obtain real-time weeding progress and current location of the robot. The Wi-Fi module is connected to the Raspberry Pi and transmits the real-time image of the RealSense camera 401 located in front of the weeding robot to the local area network, making it convenient for operators to view the weeding situation of the robot in front and behind on a PC or mobile phone.
[0091] A method for using a hood-cutting inter-row weeding robot for corn includes the following steps:
[0092] S1. Weeding begins. Driven by the electric drive of the mobile chassis 1 of the weeding robot, the weeding robot moves forward along the direction of corn weeding operation.
[0093] S2. When the weeding robot is moving forward, a Realsense D435i depth camera 401 is installed in front of it. After the Realsense D435i depth camera 401 detects that there are weeds in the direction of movement using the SAM-based weed seedling joint recognition model, the weeding robot continues to move forward and the weed status is set to 1. Then, the seedling positioning sensor 402 (K210 camera) located on the top of the weeding end effector captures the seedlings in its field of view and feeds them back to the host computer (Raspberry Pi and STM32F411).
[0094] First, the weeding robot needs to jointly identify corn seedlings and weeds, as shown in the image within its field of view. Figure 10 (a). And SAM (Segment Auything Model) , The Segmentation Model (SAM) can generate high-quality masks based on input prompts (such as points and bounding boxes), achieving zero-shot segmentation for more tasks. It is currently an advanced implementation method in the field of image segmentation, and the application scenario of this invention is well-suited to this model. A semantic segmentation model for seedlings and weeds is trained using SAM. A RealSense depth camera positioned in front of the weeding robot transmits real-time images to a Raspberry Pi. A lightweight, self-trained semantic segmentation model for weeds and seedlings is deployed on the Raspberry Pi. After image segmentation using this SAM-based model, the images are then labeled as corn seedlings using a YOLOv8 image detection model or directly based on features such as fixed growth location, pose, and area. The growth area of the corn seedlings is estimated, and green areas outside the estimated location are considered as weeds. The result illustration is shown below. Figure 10 As shown in (b).
[0095] The mobile control flowchart of the weeding robot's mobile chassis 1 is as follows: Figure 6As shown. First, the chassis moves forward continuously along the seedling cultivation line. When the RealSense D435i depth camera 401, after learning from a few samples (a combined sample of a small number of corn seedlings and weeds), detects weeds within the field of view using its SegmentAnything network, the weed status f... p Set to 1. When the weed state f p When the setting is 1, the seedling positioning sensor 402 (i.e., the K210 camera) located at the top of the weeding actuator detects a seedling within its field of view. The chassis 1 will immediately stop moving and enter the control and adjustment process of the seedling-finding robotic arm 2. After the weeding is completed, the chassis will continue moving forward.
[0096] The weeding robot has two sensors, 401 (Realsense D435i depth cameras): 401-a and 401-b. Their fixed installation positions are at the front and rear center of the robot, respectively, but their orientation (i.e., the robot's direction of travel) can change. When 401-a is in front of the robot, it primarily performs semantic segmentation on weeds and seedlings, detecting the presence of weeds near seedlings. The rear center sensor 401-b is used for gap detection, ensuring the robot has cleared all weeds from the path it has traversed. When 401-a is behind the robot, the roles of the two sensors are reversed.
[0097] In practical work, such as Figure 5 (Path diagram of chassis 1) As shown, the weeding robot first begins weeding along the rows of corn seedlings. After reaching the last seedling in that row (specifically, in robot language, this is represented by a 5-second period where no corn seedling is visible in the image of depth camera 401-a in front of the robot), the weeding robot performs a translation between rows. The 119 steering drive motor activates, and the weeding robot changes direction from forward to right until the seedling is centered in the return image of depth camera 401-b, which was originally behind the robot. The robot then adjusts its direction again, changing from right to backward, reversing the front and back directions. Depth camera 401-b, which was originally behind the robot, becomes the new camera in front, and depth camera 401-a, which was originally in front, becomes the new camera in back. This process repeats continuously until the robot completes the weeding work in the designated area.
[0098] Considering the agronomical requirements for row spacing Lh and plant spacing Lz in corn planting, to ensure that the wheel-driven steering mechanisms 101 on both sides of the hood-cutting corn inter-row weeding robot can simultaneously land on the ground without trampling on the corn seedlings, and to provide sufficient turning space, preferably, the width W of the chassis support frame 102 is designed to be 600-1500mm, and the length L of the chassis support frame 102 is designed to be 600-1500mm.
[0099] After the S3, seedling positioning sensor 402 (K210 camera), infrared sensor 403, and robot's own posture sensor gyroscopes 409 and 410 acquire information, the information is processed through an extended Kalman filter. The STM32f411 controls the seedling-finding robotic arm 2 to move laterally, longitudinally, and vertically through an adaptive control strategy. The movement of the robotic arm 2 keeps the center of the camera image on the corn seedling, thereby continuously aligning it with the corn seedling.
[0100] S4. The seedling protection mechanism 301 of the overlay-type inter-row weeding end effector 3 covers the corn seedlings under the contraction of the telescopic opening and closing mechanism 304, thereby achieving the function of surrounding and protecting the corn seedlings.
[0101] S5. Driven by the weeding drive mechanism 303, the segmented telescopic weeding blade 305 of the rotary cutting mechanism 302 can rotate and cut the weeds next to the corn seedlings, thus clearing the weeds between the rows.
[0102] S6, the telescopic opening and closing mechanism 304 opens and separates from the corn plant, completing one weed clearing operation.
[0103] Therefore, this invention employs the aforementioned cover-cutting inter-row weeding robot device and method for corn. The weeding robot is dexterous, enabling it to move through cornfields without damaging seedlings. It can simultaneously perform inter-row weeding, and its cover-cutting structure not only protects the covered corn seedlings from damage but also reduces the risk of adjacent corn plants being injured by the weeding blades. It innovatively uses a novel method based on deep learning and image segmentation—the SAM network—to segment and identify seedling instances. Furthermore, it innovatively combines extended Kalman filtering and a self-calibrating controller to control the movement of the robotic arm.
[0104] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.
Claims
1. A hood-cutting corn weeding robot, characterized in that: It includes a mobile chassis of the weeding robot, a positioning seedling covering mechanical arm, a covering and cutting type row-to-row weeding end effector and a control system. The mobile chassis of the weeding robot is installed at the bottom, the positioning seedling covering mechanical arm is installed above the mobile chassis of the weeding robot, the covering and cutting type row-to-row weeding end effector is installed below the positioning seedling covering mechanical arm, and the control system is located on one side of the mobile chassis of the weeding robot.
2. The hood-cutting corn weeding robot according to claim 1, characterized in that: The mobile chassis of the weeding robot comprises four wheel-driven steering mechanisms and a chassis support frame, wherein the four wheel-driven steering mechanisms are hinged at four end points of the chassis support frame.
3. The hood-cutting corn inter-row weeding robot according to claim 2, characterized in that: The wheel drive steering mechanism includes a travel wheel, a wheel connecting shaft, a wheel fixing frame, a travel drive motor, a steering fixed shaft, a bearing positioning plate, a steering fixed bearing, a steering motor fixing plate and a steering drive motor; the travel drive motor is fixedly connected to the outer side of the wheel fixing frame by bolts, the wheel fixing frame and the steering fixed shaft are integrally formed, the travel wheel is fixedly connected to the wheel connecting shaft, and is connected to the motor shaft of the travel drive motor through the wheel connecting shaft, the steering drive motor is fixedly connected above the steering motor fixing plate by bolts, and the side of the steering motor fixing plate is welded to the chassis support frame; the bearing positioning plate is welded to the outer side of the chassis support frame, two steering fixed bearings are oppositely screwed on the upper and lower sides of the bearing positioning plate, the steering fixed shaft passes through the center holes of the steering fixed bearing and the bearing positioning plate from bottom to top, and is connected to the motor hole of the steering drive motor by a key.
4. The hood-cutting corn weeding robot according to claim 1, characterized in that: The positioning seedling covering mechanical arm includes a mechanical arm support frame, a longitudinal translation mechanism for covering seedlings, a transverse translation mechanism for covering seedlings, and a vertical telescopic mechanism for covering seedlings; the longitudinal translation mechanism for covering seedlings is installed above the mechanical arm support frame, the transverse translation mechanism for covering seedlings is installed above the longitudinal translation mechanism for covering seedlings, and the vertical telescopic mechanism for covering seedlings is installed below the transverse translation mechanism for covering seedlings.
5. The hood-cutting corn weeding robot according to claim 4, characterized in that: The longitudinal translation mechanism of the seedling cover includes a pulley positioning bearing seat, a synchronous pulley, a pulley fixing shaft, a synchronous belt, a bearing seat fixing block, a slider connecting plate, a slide rail, a slide rail slider, a motor drive shaft, a coupling, a motor fixing plate, a longitudinal translation motor and a longitudinal translation monomer; the bearing seat fixing block is fixedly connected to the top of the mechanical arm support frame, and the pulley positioning bearing seat is screwed to the top of the bearing seat fixing block; on one side of the longitudinal translation motor, the longitudinal translation motor is connected to the motor fixing plate by bolts, and the motor fixing plate is fixedly connected to the top of the mechanical arm support frame, from the longitudinal translation The motor extends to both sides, followed by a coupling, a motor drive shaft, a synchronous pulley and a pulley positioning bearing seat. The motor shaft of the longitudinal translation motor is connected to the motor drive shaft via a coupling. The motor drive shaft passes through the synchronous pulley and is fixed by the positioning bearing seat. The synchronous belt and the synchronous pulleys on both sides form a synchronous belt drive. The slider connecting plate, the lower synchronous belt and the slide rail slider are fixed from top to bottom. The slide rail slider is installed on the slide rail track. The slide rail track is fixed on the top of the robot arm support frame. The two ends of the longitudinal translation unit are respectively fixed on the top of the slider connecting plates on both sides.
6. The hood-cutting corn weeding robot according to claim 1, characterized in that: The vertical telescopic mechanism of the seedling cover includes an electric push rod fixing plate and a vertical telescopic electric push rod. The vertical telescopic electric push rod is fixedly connected to the electric push rod fixing plate, and the electric push rod fixing plate is fixed below the transverse translational monomer by bolt connection.
7. The hood-cutting corn weeding robot according to claim 1, characterized in that: The control system includes the camera RealsenseD435i, seedling positioning sensor, infrared distance sensor, robotic arm motor driver, Raspberry Pi 4B, STM32F411, Beidou navigation and positioning module, wifi module, gyroscope and speed sensor.
8. The control method of the hood-cutting corn inter-row weeding robot according to any one of claims 1 to 8, characterized in that: The following steps are involved: S1, weeding starts. Under the action of the electric drive of the mobile chassis of the weeding robot, the weeding robot moves forward in the direction of corn weeding operation; S2, when a Realsense D435i depth camera is set in front of the weeding robot, after the Realsense D435i depth camera detects weeds in the forward direction using the weed seedling joint recognition model based on SAM, the weeding robot continues to move forward and the weed status is set to 1; Then, the seedling positioning sensor located on the top of the weeding end effector captures the seedlings in its field of view and feeds back to the host computer, and the weeding robot stops moving forward; S3, seedling positioning sensor and infrared sensor and robot's own attitude sensor gyroscope and obtain information and process it through extended Kalman filter. STM32F411 controls the positioning seedling cover mechanical arm to move horizontally, longitudinally and vertically respectively through adaptive control strategy. Through the movement of the mechanical arm, the center of the camera screen is the corn seedling, so as to continuously aim at the corn seedling; S4, the seedling cover and anti-cutting mechanism of the end effector of the cover-cutting type inter-row weeding covers the corn seedlings under the action of the contraction of the telescopic opening and closing mechanism, thereby achieving the surrounding and protective effect on the corn seedlings; S5. Under the drive of the weeding drive mechanism, the segmented telescopic weeding knife of the rotary cutting mechanism can rotate and cut off the weeds next to the corn seedlings, thereby cleaning the weeds between the rows and plants; S6. The telescopic opening and closing mechanism opens and separates from the corn plant, completing a weed clearing operation.
Citation Information
Patent Citations
Environment-friendly intelligent weeding robot
CN107509399A
Opening-closing type intelligent weeding device for acquiring seedling and grass information based on binocular vision
CN109005693A
Among-plant weeding mechanism with self-adaption adjustment function
CN109618572A
Corn seedling field weeding tool based on intelligent identification
CN110786087A
Reciprocating type seedling avoiding and weeding device between soybean plants and control method
CN117016066A
Cited By
Lawn pruning and shaping device
CN119586418A
Weeding robot and method for soybean and corn compound planting based on intelligent agriculture
CN120240024A
A soybean and corn compound planting weeding robot and method based on intelligent agriculture
CN120240024B
