A smart weeding robot for water chestnut fields and its control method
The intelligent weeding robot for water chestnut fields, which combines a dual-float structure and a multi-source sensing system with an adaptive control algorithm, solves the problems of poor passability and precise control of weeding equipment in paddy fields. It achieves coordinated operation of low-resistance movement, precise weeding, and ecological weed suppression, protecting water chestnut plants and meeting the needs of efficient weed control in water chestnut fields.
Patent Information
- Application Number
- CN202611179844.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-05
- Publication Date
- 2026-11-17
AI Technical Summary
Existing paddy field weeding equipment has poor maneuverability, is prone to damaging crops, and lacks precise control, making it difficult to achieve green mechanical physical weeding and ecological weed suppression, and failing to meet the demand for efficient weed control in water chestnut fields.
Employing a dual-buoy structure, a spiral propulsion mechanism, a disc cutter height control device, a muddy water weed suppression device, and a multi-source sensing system, combined with an adaptive control algorithm, the robot achieves low-resistance movement, precise weed removal, and ecological weed suppression in paddy fields.
In paddy fields, it achieves low-resistance movement, precise obstacle avoidance, and coordinated operation of physical weeding and ecological weed suppression, avoiding the use of chemical agents, protecting water chestnut plants, and ensuring high efficiency and ecological safety of operations.
Smart Images

Figure CN123100382A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of agricultural machinery technology, and in particular to an intelligent weeding robot for water chestnut fields and its control method. Background Technology
[0002] As an important aquatic vegetable, water bamboo's planting area has continued to expand with the development of industrial scale. Weed control in the fields has become a key agronomic link to ensure yield and quality. At present, weeding in water bamboo fields mainly relies on three methods: manual weeding, chemical spraying, and general agricultural machinery. Manual methods are labor-intensive and inefficient, making it difficult to meet the needs of large-scale planting. Although chemical weeding can improve efficiency, pesticide residues can easily pollute the aquatic environment, which contradicts the concept of green agriculture. While existing wheeled or tracked agricultural machinery can replace manual labor to some extent, its structure is mostly designed for dryland crops, and its direct application in paddy fields has obvious limitations.
[0003] Existing paddy field machinery has poor maneuverability in muddy paddy fields, easily getting stuck in the mud and damaging crops. Furthermore, it lacks precise navigation and control algorithms for tall aquatic crops, making it difficult to achieve green mechanical physical weeding. Specifically, wheeled or tracked walking mechanisms experience excessive ground pressure in soft mud, easily sinking and increasing resistance, thus reducing mobility. Simultaneously, existing equipment lacks a sensing and control system adapted to the row spacing and plant height of aquatic crops, making it prone to touching or crushing water chestnut plants during operation, causing mechanical damage. In addition, the working width and cutting height of existing weeding devices are fixed, making it difficult to adaptively adjust according to weed distribution and plant height. They also lack the function of coordinating physical weeding and ecological weed suppression, failing to achieve long-term stable weed control. Therefore, a smart weeding robot for water chestnut fields and its control method are needed. Summary of the Invention
[0004] To address the technical problems of existing paddy field weeding equipment, such as poor maneuverability, easy damage to crops, and lack of precise control, this invention provides an intelligent weeding robot for water chestnut fields and its control method.
[0005] In a first aspect, the present invention provides an intelligent weeding robot for water chestnut fields, which adopts the following technical solution: A smart weeding robot for water chestnut fields includes: Chassis support frame, screw propulsion mechanism, disc cutter height control device, disc cutter weeding end effector, muddy water weed suppression device and control system; The spiral propulsion mechanism is installed at the bottom of the chassis support frame, the disc cutter height control device is located at the top of the chassis support frame, the disc cutter weeding end effector is installed below the disc cutter height control device and is connected to the disc cutter height control device in a transmission manner, and a muddy water weed suppression device is provided in the middle of the chassis support frame. The control system is installed on one side of the chassis support frame and is electrically connected to the screw propulsion mechanism, the disc cutter height control device, the muddy water weed suppression device, and the disc cutter weeding end effector. The chassis support frame is a symmetrical double-float structure, and the outer perimeter of the chassis support frame is wrapped with a foam buoyancy layer.
[0006] Furthermore, the spiral propulsion mechanism includes a spiral device drive shaft, a spiral device fixed cylinder, spiral device blades, and a travel drive motor; The drive shaft of the spiral device is interference-fitted with the fixed cylinder of the spiral device, the blades of the spiral device are disposed on the outer peripheral wall of the fixed cylinder of the spiral device, and the drive shaft of the spiral device is connected to the chassis support frame through bearings. The walking drive motor is fixed to the outside of the chassis support frame by bolts. The output end of the walking drive motor is connected to the drive shaft of the screw device. The motor mounting plate of the walking drive motor is welded to the chassis support frame.
[0007] Furthermore, the disc cutter height control device includes a second synchronous pulley, a weeding device connecting shaft, a screw thread, a screw thread fixing bracket, a radar identification unit, a screw thread drive motor, and a drive motor fixing bracket; The drive motor mounting bracket is fixedly connected to the chassis support bracket. The screw drive motor is mounted on the drive motor mounting bracket. The screw is driven by the output end of the screw drive motor. The screw is fixedly mounted above the chassis support bracket through the screw mounting bracket. The weeding device connecting shaft is driven by the screw and reciprocates vertically under the rotation drive of the screw. The second synchronous pulley is mounted on the connecting shaft of the weeding device and is used to drive the disc cutter weeding end effector to rotate synchronously; The radar identification unit is mounted on the disc cutter height control device and is used to scan the height information of the environment ahead.
[0008] Furthermore, the disc cutter weeding end effector includes a front blade drive shaft, a blade connector, a front weed cutting blade, and a protective shell; The front weed cutting blade is bolted to the blade connector, the front blade drive shaft is welded to the blade connector, and the front blade drive shaft is connected to the weeding device connecting shaft through a coupling so as to drive the front weed cutting blade to rotate under the drive of the second synchronous pulley. The protective shell is disposed on the outside of the front weed cutting blade.
[0009] Furthermore, the muddy water weed suppression device includes a muddy water weed suppression device fixing frame, a motor drive component, a synchronous belt, a synchronous pulley, a drive shaft, a stirring device fixing frame, and a muddy water stirring component; The muddy water weed suppression device mounting frame is fixedly connected to the chassis support frame. The motor drive component is welded to the muddy water weed suppression device mounting frame and is connected to the synchronous belt pulley via a synchronous belt. The drive shaft is driven by the synchronous belt pulley. The drive shaft is connected to the muddy water weed suppression device mounting frame via bearings. The stirring device mounting frame is interference-fitted to the drive shaft. The muddy water stirring component is installed on the stirring device mounting frame.
[0010] Furthermore, it also includes a sensing system and an energy system, both of which are electrically connected to the control system; The sensing system includes a lidar module, an IMU module, a camera, and a GPS-RTK module; The lidar module is installed on the top of the chassis support frame and is used to scan the surrounding environment in real time. The IMU module is mounted on the chassis support frame and is used to measure the robot's posture; The camera is mounted on the front end of the chassis support frame and is used to collect visual information; The GPS-RTK module is located inside the chassis support frame and is used to provide an absolute positioning reference.
[0011] Furthermore, the energy system includes a lithium battery pack, flexible solar panels, and an MPPT controller; The lithium battery pack is installed inside the chassis support frame to provide the main power storage; The flexible solar panel is installed on the top of the chassis support frame to replenish energy in real time; The MPPT controller is electrically connected to the lithium battery pack and the flexible solar panel respectively, and is used to optimize solar energy conversion efficiency and monitor battery status.
[0012] Secondly, a control method for an intelligent weeding robot in a water chestnut field includes: S1. The control system drives the spiral propulsion mechanism to operate, enabling the robot to move along a preset path; S2. The sensing system collects multi-source sensor data in real time, fuses the multi-source sensor data, and identifies the row lines of water chestnuts and the distribution information of weeds. S3. Based on the information of the water chestnut row, run the vision-assisted row control algorithm, calculate the lateral deviation and generate control commands, adjust the speed and direction of the left and right helical propulsion mechanisms, so that the robot can move accurately along the crop row; S4. After arriving at the work area, the execution system starts the disc cutter weeding end actuator to cut off the weeds, and at the same time drives the turbid water weed suppression device to stir up the water layer to increase the turbidity of the water and inhibit the photosynthesis of weeds. S5. After the task is completed, the control system judges the degree of completion of the current row based on the perception data, controls the robot to turn and enter the next work row.
[0013] Further, step S3 includes: Based on the robot posture data collected by the IMU module, the adaptive slip factor in the soft mud environment is estimated in real time using the recursive least squares method. Based on the Ackermann steering geometry model, the adaptive slip factor for the mud environment is introduced into the calculation of the equivalent steering radius, and the theoretical steering radius is corrected for mud slip. The corrected turning radius is mapped to the speed difference command of the left and right helical propulsion mechanisms to compensate for the pose drift caused by the lateral slippage of the slime, enabling the robot to move accurately along the crop row. The corrected turning radius satisfies the following formula: , in, This is the corrected turning radius. The wheelbase is the projected distance in the horizontal plane of the line connecting the centers of the left and right helical propulsion mechanisms. For the equivalent front wheel steering angle, It is an adaptive slip factor for soft mud environments.
[0014] Furthermore, step S3 also includes: The state estimation of the robot is performed on the multi-source sensor data collected by the perception system based on the extended Kalman filter, and the current state of the robot is output through the prediction stage and the update stage. Based on the self-calibrating controller, the slip rate factor of the soft mud is estimated in real time using the recursive least squares method according to the current state of the robot. Based on the lateral alignment deviation and the slip ratio factor of the soft mud, an adaptive speed difference control command is constructed, which satisfies the following formula: , in, The difference in speed between the left and right helical propulsion motors. This is the proportional gain coefficient. This refers to the lateral alignment deviation. This is the slip compensation gain coefficient. It is the slip ratio factor of mud.
[0015] In summary, the present invention has the following beneficial technical effects: 1. This invention employs a symmetrical double-buoy structure and wraps a foam buoyancy layer around the outer perimeter of the chassis support frame. Combined with a spiral propulsion mechanism installed at the bottom of the chassis support frame, it enables the robot to move with low resistance and float stably in the muddy environment of paddy fields, thereby improving the robot's passability in paddy fields and reducing the risk of sinking.
[0016] 2. This invention integrates a disc cutter weeding end effector with a turbid water weed suppression device, achieving physical removal of visible weeds and ecological inhibition of underwater weed photosynthesis. It effectively controls weeds while avoiding the use of chemical agents, thus achieving green physical weeding and water ecological protection.
[0017] 3. This invention uses an extended Kalman filter to estimate the state of multi-source sensor data and combines it with the recursive least squares method to estimate the adaptive slip factor in the soft mud environment in real time. It corrects the theoretical turning radius of the Ackerman steering geometry model for soft mud slip, thereby realizing adaptive speed difference control of the robot in the soft mud substrate. This effectively compensates for the pose drift caused by lateral slip and ensures that the robot body is always in the center of the channel.
[0018] 4. This invention achieves dynamic adjustment of the vertical position of the disc cutter by linking the screw rod in the disc cutter height control device with the radar identification unit, so that the cutter only operates in the weed area without damaging the water chestnut root system, thus achieving precise obstacle avoidance and plant protection in space. Attached Figure Description
[0019] Figure 1 This is a schematic diagram of the overall structure of the intelligent weeding robot for water chestnut fields according to an embodiment of the present invention.
[0020] Figure 2 This is a schematic diagram of the mobile chassis of the weeding robot according to an embodiment of the present invention.
[0021] Figure 3 This is a front view of the weeding robot according to an embodiment of the present invention.
[0022] Figure 4 This is a diagram of the chassis support frame installation structure according to an embodiment of the present invention.
[0023] Figure 5 This is a schematic diagram of the structure of the central muddy water weed suppression device in an embodiment of the present invention.
[0024] Figure 6 This is a control flowchart of the mobile chassis of a control method for an intelligent weeding robot in a water chestnut field according to an embodiment of the present invention; Figure 7This is a horizontal control block diagram of a control method for an intelligent weeding robot in a water chestnut field according to an embodiment of the present invention. Figure 8 This is a schematic diagram of the chassis steering structure of a control method for an intelligent weeding robot in a water chestnut field according to the present invention. Figure 9 This is a flowchart illustrating the overall system control method of a smart weeding robot for water chestnut fields according to the present invention. The components include: 1. Spiral propulsion mechanism; 101. Spiral device drive shaft; 102. Spiral device fixing cylinder; 103. Spiral device blades; 2. Spiral screw drive motor; 3. Turbid water weed suppression device; 301. Turbid water weed suppression device fixing frame; 302. Synchronous pulley; 303. Turbid water stirring component; 304. Stirring device fixing frame; 305. Synchronous belt; 306. Motor drive component; 4. Drive motor fixing frame; 5. Front weed cutting blade; 6. Blade connector; 7. Spiral screw; 8. Radar identification unit; 9. Control system; 10. Second synchronous pulley; 11. Chassis support frame. Detailed Implementation
[0025] The present invention will be further described in detail below with reference to the accompanying drawings.
[0026] Example 1 Reference Figure 1 This embodiment of an intelligent weeding robot for water chestnut fields includes: like Figure 1 , Figure 2 As shown, this embodiment provides an intelligent weeding robot for water chestnut fields, including a chassis support frame 11, a spiral propulsion mechanism 1, a disc cutter height control device, a disc cutter weeding end effector, a muddy water weed suppression device 3, a control system 9, a sensing system, and an energy system; The chassis support frame 11 is a symmetrical double-pontoon structure with an overall rectangular platform appearance. The outer perimeter of the chassis support frame 11 is wrapped with a foam buoyancy layer made of lightweight, high-buoyancy material. The double-pontoon structure forms a stable waterline surface through symmetrical arrangement, lowering the center of gravity of the entire machine and enhancing its anti-overturning ability. As an auxiliary buoyancy element, the foam buoyancy layer increases the overall buoyancy reserve on top of the main buoyancy support provided by the double pontoons. Even if some pontoons are damaged, it can still maintain basic buoyancy, thereby ensuring stable operation of the robot in water environments with a depth of 15cm to 30cm.
[0027] like Figure 4As shown, the screw propulsion mechanism 1 is installed at the bottom of the chassis support frame 11 and distributed at the four end positions of the bottom of the chassis support frame 11; the disc cutter height control device is located at the top of the chassis support frame 11, and the disc cutter weeding end effector is installed below the disc cutter height control device and is connected to the disc cutter height control device in a transmission manner; the muddy water weed suppression device 3 is located in the middle of the chassis support frame 11, and the control system 9 is installed on one side of the chassis support frame 11 and is electrically connected to the screw propulsion mechanism 1, the disc cutter height control device, the muddy water weed suppression device 3, and the disc cutter weeding end effector; the sensing system and the energy system are both electrically connected to the control system 9.
[0028] The screw propulsion mechanism 1 includes a screw drive shaft 101, a screw fixing cylinder 102, screw blades 103, a travel drive motor, and a motor bracket. The screw drive shaft 101 and the screw fixing cylinder 102 are fixedly connected by an interference fit. The screw blades 103 are fixedly mounted on the outer peripheral wall of the screw fixing cylinder 102, and the diameter of the screw blades 103 is 250mm. The screw drive shaft 101 is rotatably connected to the chassis support frame 11 via bearings. The thrust generated when the screw blades rotate is transmitted to the chassis support frame 11 through the screw drive shaft 101, thereby driving the entire machine to move.
[0029] like Figure 3 As shown, the walking drive motor is fixed to the float via a motor bracket. The output end of the walking drive motor is connected to the drive shaft 101 of the screw device. The motor mounting plate of the walking drive motor is welded to the chassis support frame 11 to form a rigid transmission and support structure. The left and right sets of screw propulsion mechanisms 1 generate a speed difference under the differential control of the control system 9, realizing the robot's turning and correction in soft mud. It provides 2×260N of propulsion force for stable movement in water depths of 15cm to 30cm and soft mud environments of 20cm to 40cm, avoiding the sinking problem caused by excessive ground pressure in muddy paddy fields of traditional wheeled or tracked structures.
[0030] The disc cutter height control device includes a second synchronous pulley 10, a weeding device connecting shaft, a screw 7, a screw 7 mounting bracket, a radar identification unit 8, a screw 7 drive motor 2, and a drive motor mounting bracket 4. The drive motor mounting bracket 4 is fixedly connected to the chassis support frame 11, and the screw 7 drive motor 2 is mounted on the drive motor mounting bracket 4. The screw 7 is driven by the output end of the screw 7 drive motor 2, and the screw 7 is fixedly mounted above the chassis support frame 11 via the screw 7 mounting bracket. The weeding device connecting shaft is driven by the screw 7 and reciprocates vertically under the rotational drive of the screw 7. The second synchronous pulley 10 is mounted on the weeding device connecting shaft and is used to drive the disc cutter weeding end effector to rotate synchronously. The radar identification unit 8 is mounted on the disc cutter height control device and is used to scan the height information of the surrounding environment and feed the height data back to the control system 9 in real time. After receiving instructions from the control system 9, the screw 7 drives the motor 2 to convert the rotational motion into linear motion of the connecting shaft of the weeding device through the screw 7, thereby precisely adjusting the vertical working position of the end effector of the disc cutter weeding device to adapt to different water depths and weed heights.
[0031] The disc cutter weeding end effector includes a front-end blade drive shaft, a blade connector 6, multiple rotating disc blades, and a protective shell. The multiple rotating disc blades are bolted to the blade connector 6, and the power is 540W. The front-end blade drive shaft and blade connector 6 are welded together. The front-end blade drive shaft is connected to the weeding device connecting shaft via a coupling, driving the multiple rotating disc blades to rotate under the drive of the second synchronous pulley 10. The protective shell is located on the outside of the multiple rotating disc blades to isolate them from the water chestnut plants during operation, preventing damage to the crop. During operation, the multiple rotating disc blades rotate at high speed driven by the power transmitted from the travel drive motor via the second synchronous pulley 10, quickly cutting into and removing visible weeds above the water surface using the cutting action of the blades. The protective shell limits the effective radius of the blades, ensuring that the cutting area only covers the weed growth area and does not touch the stems and roots of the water chestnut plants on both sides.
[0032] like Figure 5As shown, the turbid water weed suppression device 3 includes a turbid water weed suppression device mounting frame 301, a motor drive component 306, a synchronous belt 305, a synchronous pulley 302, a drive shaft, a stirring device mounting frame 304, and a turbid water stirring component 303. The turbid water weed suppression device mounting frame 301 is fixedly connected to the chassis support frame 11. The motor drive component 306 is welded to the turbid water weed suppression device mounting frame 301, and the output end of the motor drive component 306 is connected to the synchronous pulley 302 via the synchronous belt 305. The drive shaft is driven by the synchronous pulley 302, and the drive shaft is rotatably connected to the turbid water weed suppression device mounting frame 301 via bearings. The stirring device mounting frame 304 is interference-fitted with the drive shaft, and the turbid water stirring component 303 is fixedly installed on the stirring device mounting frame 304. In this embodiment, the turbid water stirring component 303 is preferably composed of nylon rods. During operation, the motor drive component 306 drives the drive shaft to rotate via the synchronous belt 305 and synchronous pulley 302. The stirring device mounting bracket 304 rotates synchronously with the drive shaft, causing the turbid water stirring component 303 to continuously agitate the water, generating uniform eddies and ripples, disrupting the light transmittance of the water surface, thereby inhibiting the photosynthesis of underwater weeds and achieving long-term ecological weed suppression. The turbid water stirring component 303, made of nylon rods, is corrosion-resistant and lightweight, which reduces the drive load and extends its service life.
[0033] The perception system includes a lidar module, an IMU module, a camera, and a GPS-RTK module. The lidar module, mounted on the top of the chassis support frame 11, scans the surrounding environment in real time, measuring the distance to obstacles by emitting and receiving laser pulse signals, providing high-precision distance measurement and obstacle detection. The IMU module, mounted on the chassis support frame 11, performs high-frequency attitude measurement, capturing the robot's acceleration and angular velocity data via accelerometers and gyroscopes, providing a real-time attitude reference for the control system 9. The camera, mounted at the front of the chassis support frame 11, collects visual information about the field ahead, and uses image processing algorithms to identify the rows of water chestnuts and the distribution of weeds, calculating the lateral deviation between the rows and the image center. The GPS-RTK module, located inside the chassis support frame 11, provides an absolute positioning reference, employing RTK wireless mode in high-density plant areas to reduce signal interference and achieve centimeter-level positioning accuracy.
[0034] The energy system includes a lithium battery pack, flexible solar panels, and an MPPT controller. The lithium battery pack is installed inside the chassis support frame 11 to provide primary power storage. The flexible solar panels are installed on top of the chassis support frame 11 for real-time energy replenishment. The MPPT controller is electrically connected to both the lithium battery pack and the flexible solar panels to track the maximum power point of the flexible solar panels in real time, optimize solar energy conversion efficiency, and monitor the battery charging and discharging status for intelligent charging and discharging management. The lithium battery pack and flexible solar panels form a hybrid power supply scheme. During periods of ample sunlight, the MPPT controller prioritizes solar energy replenishment, while on cloudy days or under high-load conditions, the lithium battery pack provides a stable power output, ensuring continuous operation of the robot for extended periods.
[0035] Control system 9 is built on the NVIDIA Jetson Nano platform and runs the ROS 2 Humble operating system. Mounted on one side of the chassis support frame 11, control system 9 receives multi-source data collected by the sensing system, performs real-time processing and decision-making, and generates control commands to drive the spiral propulsion mechanism 1, the disc cutter height control device, and the muddy water weed suppression device 3 to work collaboratively. Control system 9 uses an extended Kalman filter to fuse multi-source data from the LiDAR module, IMU module, camera, and GPS-RTK module to handle measurement noise and system uncertainty, outputting an estimate of the robot's current state. Based on a vision-assisted alignment control algorithm, it calculates the lateral deviation based on the water chestnut row information identified by the camera, and combines this with IMU module attitude data to generate control commands, achieving precise alignment navigation.
[0036] like Figure 6 As shown, during field operations, the energy system manages the energy output of the lithium battery pack and flexible solar panels through the MPPT controller, supplying power to the control system 9, sensing system, spiral propulsion mechanism 1, disc cutter height control device, disc cutter weeding end effector, and muddy water weed suppression device 3. At the start of operations, the control system 9 generates control commands based on the water chestnut row information and environmental height information fed back by the sensing system. These commands drive the spiral propulsion mechanism 1 to adjust the speed and direction of the left and right walking drive motors, enabling the robot to move along the crop rows in the soft mud.
[0037] When the robot arrives at the work area, the control system 9, based on the real-time height data fed back by the radar identification unit 8, controls the screw 7 to drive the motor 2. The screw 7 adjusts the vertical position of the connecting shaft of the weeding device, placing multiple rotating disc blades at a suitable working height. Simultaneously, the walking drive motor drives the front blade drive shaft to rotate via the second synchronous pulley 10, thereby causing multiple rotating disc blades to rotate at high speed, cutting visible weeds. While the multiple rotating disc blades are operating, the turbid water weed suppression device 3, driven by the motor drive component 306, drives the drive shaft to rotate via the synchronous belt 305 and synchronous pulley 302. This causes the turbid water stirring component 303 to continuously agitate the water layer, increasing water turbidity and inhibiting weed photosynthesis, achieving synergistic weed control through physical cutting and ecological weed suppression.
[0038] During operation, radar identification unit 8 continuously scans the height information of the environment ahead, and control system 9 dynamically adjusts the vertical position of multiple rotating disc blades based on real-time height data to ensure that the blades only operate in weedy areas without damaging the roots of the water chestnuts. When the perception system detects that the current row has been completed, control system 9 controls the spiral propulsion mechanism 1 to perform a turning action, allowing the robot to enter the next working row to continue the weeding operation.
[0039] Example 2 The difference between this embodiment and Embodiment 1 is that this embodiment provides a control method for an intelligent weeding robot in a water chestnut field; Specifically, a control method for an intelligent weeding robot in a water chestnut field includes the following steps: S1. The control system drives the spiral propulsion mechanism to operate, enabling the robot to move along a preset path.
[0040] like Figure 8 , Figure 9 As shown, specifically, the control system generates initial movement commands based on the ROS 2 Humble operating system, outputs reference speed signals to the left and right walking drive motors respectively, and the spiral propulsion mechanism generates propulsion force under the drive of the walking drive motors, so that the robot moves forward along the preset path in the paddy field.
[0041] During the initial movement phase, the left and right walking drive motors maintain synchronized speeds, and the robot enters the work area along a preset straight path. The control system obtains the robot's absolute position through the GPS-RTK module and compares it with the preset path coordinates to generate initial heading information; the LiDAR module scans the environment ahead in real time to build a local obstacle map; the IMU module captures initial attitude data and measures the robot's acceleration and angular velocity; and the camera acquires initial field images ahead. The above multi-source sensor data is initially frame-aligned and coordinate-system-unified by the control system, providing a reference state for subsequent data fusion processing of the extended Kalman filter and the vision-assisted alignment control algorithm, ensuring that the robot has completed pose initialization and path alignment before entering the work area.
[0042] S2. The sensing system collects multi-source sensor data in real time, fuses and processes the multi-source sensor data, and identifies the row lines of water chestnuts and the distribution information of weeds. During the robot's movement, the perception system simultaneously initiates multi-source data acquisition. The lidar module, mounted top-mounted, performs a 360-degree scan of the surrounding environment, emitting and receiving laser pulse signals to acquire obstacle distance information and build a real-time point cloud map. The IMU module captures the robot's three-axis acceleration and three-axis angular velocity data at a high-frequency sampling rate, outputting high-frequency attitude information. The camera, mounted at a forward-tilted angle, acquires images of the field ahead, covering the crop row area. The GPS-RTK module, located inside the robot, receives satellite signals, providing centimeter-level absolute positioning coordinates in open areas and switching to RTK wireless mode in high-density plant areas to reduce multipath interference. The lidar distance data, IMU attitude data, camera image data, and GPS-RTK positioning data are timestamped and aligned with the coordinate system by the control system, forming a multi-source sensor data stream.
[0043] The control system employs an extended Kalman filter to fuse multi-source sensor data, eliminating noise from individual sensors and estimating the robot's current state. The extended Kalman filter processing is divided into a prediction phase and an update phase. In the prediction phase, the current state is predicted based on a nonlinear system model, and the state prediction satisfies the following equation: , in, Let k be the predicted state. The estimated state at time k-1, To control the input, the covariance of the prediction error satisfies the following equation: ; in, The prediction error covariance, Let be the Jacobian matrix of the state transition function obtained at k-1. This is an estimate of the error covariance from the previous time step. To account for the process noise covariance, during the update phase, the predicted state is corrected based on the actual observation data, and the observation update satisfies the following equation: , in, These are actual observed values. For the observation model function, For observation noise. The Kalman gain calculation satisfies the following equation: , in, For Kalman gain, Let be the Jacobian matrix of the observation model function obtained at k. To observe the noise covariance, the state update satisfies the following equation: , Error covariance update satisfies the following equation: , in, The identity matrix is used. Through continuous iterations in the prediction and update phases, the extended Kalman filter outputs an estimate of the robot's current state, including position, velocity, and attitude information, providing a filtered baseline state for subsequent control decisions.
[0044] While fusing multi-source data, field images captured by cameras are processed using visual processing algorithms to identify the rows of water chestnuts and the distribution of weeds. The image processing workflow includes: converting the original image to grayscale to reduce computational load; segmenting the crop and background areas using adaptive threshold binarization; extracting the edges of the water chestnut rows using Hough line detection or lightweight semantic segmentation models; calculating the lateral pixel deviation between the center lines of the left and right rows and the center line of the image; and converting this deviation into a lateral alignment deviation in the actual physical coordinate system. When the camera detects weeds in the direction of travel based on the image processing algorithm, it sets the weed status and feeds it back to the control system; when the camera captures the water chestnut rows within its field of view, it simultaneously feeds back the row coordinate information and the lateral alignment deviation to the control system. The fused robot state estimation, lateral alignment deviation, weed distribution status, and obstacle distance information together form the input data basis for subsequent vision-assisted row alignment control algorithms, enabling the control system to generate precise control commands based on real-time environmental perception.
[0045] S3. Based on the information of the water chestnut row, run the vision-assisted row control algorithm, calculate the lateral deviation and generate control commands, adjust the speed and direction of the left and right helical propulsion mechanisms, so that the robot can move accurately along the crop row; Based on the robot's current state estimate from the extended Kalman filter output in step S2, the control system initiates a self-tuning controller to adaptively adjust control parameters. The self-tuning controller uses the state estimate from the extended Kalman filter output as the system's current state, and employs recursive least squares to estimate controller parameters online and estimate the slip factor of the soft mud in real time. The parameter update using recursive least squares satisfies the following equation: , in, Here are the estimated controller parameters at time k. Here are the estimated controller parameters at time k-1. In order to control the target, For the transpose of the derived design matrix, For Kalman gain.
[0046] Simultaneously, the control system, based on the robot's posture data acquired by the IMU module, uses the recursive least squares method to estimate the adaptive slip factor in the mud environment in real time. Based on the Ackermann steering geometry model, this adaptive slip factor is incorporated into the calculation of the equivalent steering radius, and the theoretical steering radius is corrected for mud slippage. The corrected steering radius satisfies the following equation: , in, This is the corrected turning radius. The wheelbase is the projected distance in the horizontal plane of the line connecting the centers of the left and right helical propulsion mechanisms. For the equivalent front wheel steering angle, The adaptive slip factor for the soft mud environment is estimated online using the recursive least squares method based on attitude data collected in real time by the IMU module.
[0047] The corrected turning radius is mapped to the theoretical speed difference benchmark of the left and right helical propulsion mechanisms. This theoretical speed difference benchmark characterizes the ideal differential speed relationship of the robot in a soft mud environment without sideslip, and satisfies the following equation: , in, Based on the theoretical speed difference, For travel speed, The center distance between the left and right helical propulsion mechanisms. It represents the propulsion efficiency coefficient of the propeller blades in the mud-water mixture.
[0048] Based on this, the self-calibrating controller constructs an adaptive speed difference control command according to the lateral alignment deviation and the mud slip rate factor. This adaptive speed difference control command serves as the final control quantity output to the left and right travel drive motors, satisfying the following equation: , in, The difference in speed between the left and right helical propulsion motors. This is the proportional gain coefficient. This refers to the lateral alignment deviation. This is the slip compensation gain coefficient. This is the slip ratio factor of the mud estimated in real time based on the recursive least squares method. The control law update satisfies the following equation: , in, Let k be the control output at time k. Performance monitoring adjusts future control behavior based on the deviation between the system output and the control target, where the deviation satisfies the following equation: , in, To control deviation, For reference output, To control the target.
[0049] Through the above two steps of calculation, the theoretical speed difference reference is obtained. Provides kinematic constraint references in muddy environments and adaptive speed difference control commands. Based on this, a closed-loop control system for lateral correction and slip compensation is implemented. The control system distributes adaptive speed difference control commands to the left and right helical propulsion mechanisms to compensate for the pose drift caused by the lateral slippage of the soft mud, enabling the robot to move precisely along the crop row.
[0050] Simultaneously, based on real-time height data fed back by the radar identification unit, the control system dynamically adjusts the vertical position of the end effector of the disc cutter weeding device via a screw-driven motor. The vertical position satisfies the following formula: , in, The working height of the disc cutter weeding end effector. Given the current water depth, The height of the weeds, This is the height ratio coefficient. As a safety margin constant, it ensures that the cutter only operates in the weed area without damaging the protected root system of the water chestnut, thus achieving precise obstacle avoidance in space.
[0051] S4. After arriving at the work area, the execution system starts the disc cutter weeding end actuator to cut off the weeds, and at the same time drives the turbid water weed suppression device to stir up the water layer to increase the turbidity of the water and inhibit the photosynthesis of weeds. like Figure 7As shown, after arriving at the work area, the control system, based on the positioning information output by the visual-assisted row control algorithm in step S3 and the real-time height data fed back by the radar identification unit, synchronously outputs a start signal to the disc cutter height control device and the muddy water weed suppression device, and the execution system enters the collaborative weeding operation state.
[0052] The start-up process of the disc cutter weeding end effector is as follows: Based on the current water depth and weed height data, the control system controls the screw drive motor to operate. The screw rotates under the constraint of the screw fixing frame, converting the rotational motion into linear motion of the weeding device connecting shaft. This causes the weeding device connecting shaft to be vertically adjusted according to the formula... A predetermined operating height ensures that the front weed-cutting blade is positioned vertically to cover only the area where weeds grow. Once the vertical position is adjusted, the drive motor drives the front blade drive shaft to rotate via the second synchronous belt pulley. The front blade drive shaft is connected to the weeding device connecting shaft via a coupling, transmitting power to the blade connector, which in turn drives the front weed-cutting blade to rotate at high speed. Utilizing a 540W rotational cutting force, it quickly cuts into and removes visible weeds above the water surface. A protective shell is installed on the outside of the front weed-cutting blade to isolate the blade from the water chestnut plants on both sides during cutting, limiting the effective radius of action and preventing damage to the crop stems and roots.
[0053] While the disc cutter's end effector physically removes weeds, the turbid water weed suppression device is simultaneously activated. The control system outputs operating commands to the motor drive component of the turbid water weed suppression device. The motor drive component establishes a transmission connection with the synchronous pulley via a synchronous belt. The synchronous pulley drives the drive shaft to rotate. The drive shaft is rotatably connected to the turbid water weed suppression device's mounting frame via bearings. The stirring device's mounting frame is interference-fitted with the drive shaft and rotates synchronously with it. The turbid water stirring component mounted on the stirring device's mounting frame continuously agitates the water layer. In this embodiment, the turbid water stirring component is composed of nylon rods. During rotation, the nylon rods generate uniform eddies and ripples, increasing suspended particles in the water, increasing water turbidity, and disrupting water surface light transmittance, thereby inhibiting the photosynthesis of underwater weeds and achieving ecological weed suppression.
[0054] During operation, the radar identification unit continuously scans the height information of the environment ahead. The control system dynamically adjusts the operation of the screw drive motor based on real-time height data, causing the connecting shaft of the weeding device to drive the front weed cutting blade to correct its vertical position in real time, ensuring that the cutter always operates only in the weed area during the movement. The physical cutting of the front weed cutting blade and the ecological weed suppression of the turbid water stirring component form a physical-ecological synergistic weeding mechanism. The former immediately removes visible weeds above the water surface, while the latter inhibits the photosynthesis of underwater weeds by continuously stirring the water, achieving long-term weed control while maintaining the ecological balance of the water body.
[0055] S5. After the task is completed, the control system judges the degree of completion of the current row based on the perception data, controls the robot to turn and enter the next work row.
[0056] After the task is completed, the control system makes a comprehensive judgment on the completion status of the current row based on multi-source data from the sensing system. Specifically, the camera processes the image of the field of view in front, and when it detects the end of the current crop row and there are no continuous row features in the field of view, it outputs a row end signal to the control system; the lidar module scans the environment in front to confirm that there are no obstacles within the preset safe distance; the GPS-RTK module compares the current absolute positioning coordinates with the coordinates of the current row end point of the preset work path to verify that the robot has reached the end point of the row. When the row end signal, the obstacle-free confirmation signal, and the positioning arrival signal are all satisfied simultaneously, the control system determines that the current row task is completed.
[0057] After determining that the current row is completed, the control system executes differential steering control to enter the next row. Based on the differential speed principle of the left and right helical propulsion mechanism, the control system outputs differential control commands to the left and right walking drive motors, causing a speed difference between the blades of the left and right helical devices, driving the robot to complete a turn in place in the soft mud. During the turn, the control system, according to a preset grid path planning strategy, adopts a "longitudinal-then-lateral" approach under field layout conditions of 60cm plant spacing and 100cm row spacing. That is, the robot moves longitudinally to the end of the current row, then laterally to the starting point of the adjacent row, and then aligns with the next row. The speed difference control during the turn is based on a modified turning radius model, mapping the target turning radius to the speed difference command of the left and right helical propulsion mechanisms.
[0058] After the turn is completed, the control system re-initializes the vision-assisted row control algorithm. Based on the water chestnut row information of the new work row, it calculates the initial lateral deviation, generates new control commands to drive the robot to move precisely along the next work row, and repeats steps S2 to S5 until the weeding work in all field work areas is completed.
[0059] Finally, the energy management system monitors the status of the battery and solar panel in real time, and realizes intelligent charging and discharging through the MPPT controller to ensure that the robot can operate stably for a long time.
[0060] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.
Claims
1. A smart weeding robot for water chestnut fields, characterized in that, include: The chassis support frame (11), the screw propulsion mechanism (1), the disc cutter height control device, the disc cutter weeding end actuator, the muddy water weed suppression device (3), and the control system (9); The spiral propulsion mechanism (1) is installed at the bottom of the chassis support frame (11), the disc cutter height control device is set at the top of the chassis support frame (11), the disc cutter weeding end actuator is installed below the disc cutter height control device and is connected to the disc cutter height control device in a transmission manner, and the chassis support frame (11) is provided with a muddy water weed suppression device (3) in the middle. The control system (9) is installed on one side of the chassis support frame (11) and is electrically connected to the screw propulsion mechanism (1), the disc cutter height control device, the muddy water weed suppression device (3) and the disc cutter weeding end actuator. The chassis support frame (11) is a left-right symmetrical double float structure, and the outer periphery of the chassis support frame (11) is wrapped with a foam buoyancy layer.
2. The intelligent weeding robot for water chestnut fields according to claim 1, characterized in that, The spiral propulsion mechanism (1) includes a spiral device drive shaft (101), a spiral device fixed cylinder (102), spiral device blades (103), and a travel drive motor; The drive shaft (101) of the spiral device is press-fitted with the fixed cylinder (102) of the spiral device, the blades (103) of the spiral device are disposed on the outer peripheral wall of the fixed cylinder (102) of the spiral device, and the drive shaft (101) of the spiral device is connected to the chassis support frame (11) through a bearing. The walking drive motor is fixed to the outside of the chassis support frame (11) by bolts. The output end of the walking drive motor is connected to the drive shaft (101) of the screw device. The motor fixing plate of the walking drive motor is welded to the chassis support frame (11) on the side.
3. The intelligent weeding robot for water chestnut fields according to claim 1, characterized in that, The disc cutter height control device includes a second synchronous pulley (10), a weeding device connecting shaft, a spiral screw (7), a spiral screw (7) fixing frame, a radar identification unit (8), a spiral screw (7) drive motor (2), and a drive motor fixing frame (4); The drive motor mounting bracket (4) is fixedly connected to the chassis support bracket (11). The screw (7) drive motor (2) is mounted on the drive motor mounting bracket (4). The screw (7) is connected to the output end of the screw (7) drive motor (2). The screw (7) is fixedly mounted above the chassis support bracket (11) through the screw (7) mounting bracket. The weeding device connecting shaft is connected to the screw (7) and reciprocates vertically under the rotation drive of the screw (7). The second synchronous pulley (10) is mounted on the connecting shaft of the weeding device and is used to drive the disc cutter weeding end actuator to rotate synchronously; The radar identification unit (8) is installed on the disc cutter height control device and is used to scan the height information of the environment in front.
4. The intelligent weeding robot for water chestnut fields according to claim 1, characterized in that, The disc cutter weeding end effector includes a front blade drive shaft, a blade connector (6), a front weed cutting blade (5), and a protective shell; The front weed cutting blade (5) is bolted to the blade connector (6), the front blade drive shaft is welded to the blade connector (6), and the front blade drive shaft is connected to the weeding device connecting shaft through a coupling so as to drive the front weed cutting blade (5) to rotate under the drive of the second synchronous pulley (10). The protective shell is set on the outside of the front weed cutting blade (5).
5. The intelligent weeding robot for water chestnut fields according to claim 1, characterized in that, The muddy water weed suppression device (3) includes a muddy water weed suppression device fixing frame (301), a motor drive component (306), a synchronous belt (305), a synchronous pulley (302), a drive shaft, a stirring device fixing frame (304), and a muddy water stirring component (303); The muddy water weed suppression device mounting bracket (301) is fixedly connected to the chassis support bracket (11). The motor drive component (306) is welded to the muddy water weed suppression device mounting bracket (301) and is connected to the synchronous pulley (302) via a synchronous belt (305). The drive shaft is driven by the synchronous pulley (302). The drive shaft is connected to the muddy water weed suppression device mounting bracket (301) via a bearing. The stirring device mounting bracket (304) is interference-fitted to the drive shaft. The muddy water stirring component (303) is mounted on the stirring device mounting bracket (304).
6. The intelligent weeding robot for water chestnut fields according to claim 1, characterized in that, It also includes a sensing system and an energy system, both of which are electrically connected to the control system (9); The sensing system includes a lidar module, an IMU module, a camera, and a GPS-RTK module; The lidar module is installed on the top of the chassis support frame (11) for real-time scanning of the surrounding environment; The IMU module is mounted on the chassis support frame (11) and is used to measure the robot's posture; The camera is installed at the front end of the chassis support frame (11) and is used to collect visual information; The GPS-RTK module is placed inside the chassis support frame (11) and is used to provide an absolute positioning reference.
7. The intelligent weeding robot for water chestnut fields according to claim 6, characterized in that, The energy system includes a lithium battery pack, flexible solar panels, and an MPPT controller; The lithium battery pack is installed inside the chassis support frame (11) to provide the main power storage; The flexible solar panel is installed on the top of the chassis support frame (11) for real-time energy replenishment; The MPPT controller is electrically connected to the lithium battery pack and the flexible solar panel respectively, and is used to optimize solar energy conversion efficiency and monitor battery status.
8. A control method for an intelligent weeding robot in a water chestnut field, based on the intelligent weeding robot in a water chestnut field as described in claim 1, characterized in that, include: S1. The control system (9) drives the spiral propulsion mechanism (1) to operate, so that the robot moves along the preset path; S2. The sensing system collects multi-source sensor data in real time, fuses the multi-source sensor data, and identifies the row lines of water chestnuts and the distribution information of weeds. S3. Based on the information of the water chestnut row, run the visual-assisted row control algorithm, calculate the lateral deviation and generate control commands, adjust the speed and direction of the left and right spiral propulsion mechanism (1) so that the robot can move accurately along the crop row; S4. After arriving at the work area, the execution system starts the disc cutter weeding end actuator to cut the weeds, and at the same time drives the turbid water weed suppression device (3) to stir the water layer to increase the turbidity of the water and suppress the photosynthesis of weeds. S5. After the work is completed, the control system (9) judges the degree of completion of the current row based on the perception data, controls the robot to turn and enter the next work row.
9. The control method for an intelligent weeding robot in a water chestnut field according to claim 8, characterized in that, Step S3 includes: Based on the robot posture data collected by the IMU module, the adaptive slip factor in the soft mud environment is estimated in real time using the recursive least squares method. Based on the Ackermann steering geometry model, the adaptive slip factor for the mud environment is introduced into the calculation of the equivalent steering radius, and the theoretical steering radius is corrected for mud slip. The corrected turning radius is mapped to the speed difference command of the left and right helical propulsion mechanism (1) to compensate for the pose drift caused by the lateral slippage of the soft mud, so that the robot can move accurately along the crop row. The corrected turning radius satisfies the following formula: , in, This is the corrected turning radius. The wheelbase is the projected distance in the horizontal plane of the line connecting the centers of the left and right helical propulsion mechanisms. For the equivalent front wheel steering angle, It is an adaptive slip factor for soft mud environments.
10. The control method for an intelligent weeding robot in a water chestnut field according to claim 9, characterized in that, Step S3 also includes: The state estimation of the robot is performed on the multi-source sensor data collected by the perception system based on the extended Kalman filter, and the current state of the robot is output through the prediction stage and the update stage. Based on the self-calibrating controller, the slip rate factor of the soft mud is estimated in real time using the recursive least squares method according to the current state of the robot. Based on the lateral alignment deviation and the slip ratio factor of the soft mud, an adaptive speed difference control command is constructed, which satisfies the following formula: , in, The difference in speed between the left and right helical propulsion motors. This is the proportional gain coefficient. This refers to the lateral alignment deviation. This is the slip compensation gain coefficient. It is the slip ratio factor of mud.
