Self-adaptive intelligent irrigation robot for hilly terrain

Abstract

The invention discloses a self-adaptive intelligent irrigation robot for hilly terrains. The self-adaptive intelligent irrigation robot comprises a crawler-type mobile chassis, a multi-mode irrigation device, a visual navigation module and an energy management system. The robot does not depend on a high-cost GPS, but utilizes an OpenCV visual module to carry out field path identification and hydrant positioning. Through a deep learning model, the system can automatically identify the types of crops such as tea trees and fruit trees, and automatically switch between two irrigation modes of drip irrigation and micro-spraying according to an identification result, so that precise operation is realized. The crawler-type chassis is integrated with a dip angle self-adaptive system, the differential speed of the crawler belts on the two sides can be dynamically adjusted, and it is ensured that the crawler-type chassis stably advances on a 35-degree abrupt slope. After the irrigation task is completed, the pneumatic recovery device can efficiently wind the water pipe, and the automation efficiency is improved. The problems that a traditional irrigation robot is high in positioning cost, poor in terrain adaptability and serious in water resource waste are effectively solved.

Description

Technical Field

[0001] This invention relates to the field of agricultural automation technology, and in particular to an adaptive intelligent irrigation robot for hilly terrain. Background Technology

[0002] Traditional agricultural irrigation methods, whether manual or fixed sprinkler / drip irrigation systems, have significant limitations. Manual irrigation is labor-intensive, inefficient, and struggles to ensure uniform irrigation. While fixed irrigation systems improve efficiency, they are costly and difficult to deploy in unstructured terrains such as hills and mountains, and lack flexibility, often resulting in significant water waste as they cannot dynamically adjust to different crop growth stages or water requirements. Summary of the Invention

[0003] The purpose of this invention is to provide a hilly terrain adaptive intelligent irrigation robot to achieve low-cost, high-precision, highly adaptable to terrain and capable of intelligent decision-making irrigation functions.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is: to provide a hilly terrain adaptive intelligent irrigation robot, the hilly terrain adaptive intelligent irrigation robot comprising: Tracked mobile chassis; A multimodal irrigation device, wherein the multimodal irrigation device is a water tap that can change the irrigation mode and is fixed to the ground, the multimodal irrigation device being used to perform at least two different irrigation modes; The visual navigation module is used to acquire and process environmental image information for navigation and target recognition. The data processing and decision-making unit is electrically connected to the visual navigation module, the multimodal irrigation device, and the tracked mobile chassis. The data processing and decision-making unit controls the movement of the tracked mobile chassis and the operation mode of the multimodal irrigation device based on the information processed by the visual navigation module.

[0005] In one embodiment, the tracked mobile chassis integrates a tilt adaptive system that can dynamically adjust the speed difference between the tracks on both sides of the vehicle based on the real-time monitored vehicle slope, so as to maintain vehicle stability when traveling on slopes.

[0006] In one embodiment, the multimodal irrigation device includes a drip irrigation mode and a micro-spraying mode, which are controlled and switched by a solenoid valve.

[0007] In one embodiment, the data processing and decision-making unit includes an irrigation decision engine that can automatically select the drip irrigation mode or the micro-sprinkler mode based on the crop type identified by the visual navigation module.

[0008] In one embodiment, the visual navigation module uses a deep learning model to identify crop types. When the identification result is a drought-resistant crop such as tea tree or strawberry, the irrigation decision engine selects drip irrigation mode; when the identification result is a moisture-loving crop such as fruit tree, micro-sprinkler mode is selected.

[0009] In one embodiment, the visual navigation module is based on the OpenCV vision library and uses image processing algorithms to identify field paths and visually locate field water hydrants.

[0010] In one embodiment, the data processing and decision-making unit employs a Simultaneous Localization and Mapping (SLAM) algorithm, which integrates the water hydrant location information collected by the visual navigation module to achieve high-precision autonomous positioning and optimal path planning for the robot.

[0011] In one embodiment, a pneumatic irrigation hose retraction device is also included for automatically retracting the irrigation hose after the irrigation operation is completed.

[0012] In one embodiment, the pneumatic irrigation hose recovery device uses compressed air to drive rollers for high-speed winding and is equipped with a tension sensor that automatically reduces the winding speed to prevent the irrigation hose from breaking when abnormal winding tension is detected.

[0013] In one embodiment, the image processing algorithm of the visual navigation module for field paths includes the following steps: converting the image to grayscale, performing Gaussian filtering for noise reduction, using the Canny operator for edge detection, and fitting a straight path using the Hough transform.

[0014] The above-described technical solutions in the embodiments of the present invention have at least the following technical effects or advantages: The hilly terrain adaptive intelligent irrigation robot provided in this invention uses a pure visual navigation solution to replace the expensive GPS / RTK module, significantly reducing hardware costs by approximately 40%. By combining SLAM algorithm with visual markers (water taps), millimeter-level positioning accuracy can be achieved even in areas with signal obstruction, such as hilly terrain, thereby reducing costs while improving navigation accuracy.

[0015] In addition, the tracked chassis, combined with the differential speed adjustment system, enables the robot to operate stably on steep slopes of up to 35°, significantly expanding the robot's application scenarios and enabling it to undertake irrigation tasks in complex terrains such as hills and terraced fields, thus enhancing the terrain adaptability of the irrigation robot.

[0016] Furthermore, by combining deep learning-based crop identification (with an accuracy rate of up to 95%) with automatic switching between drip irrigation and micro-sprinkler modes, it enables differentiated and precise irrigation for different crops. Compared to traditional single-mode irrigation, water saving rates can be increased by more than 35%, significantly improving water resource utilization efficiency, achieving precision irrigation, and conserving water resources.

[0017] Finally, the entire process, from path planning, crop identification, mode switching, and irrigation belt retrieval, is fully automated. In particular, the pneumatic retrieval device can complete hose retrieval within seconds, significantly improving retrieval efficiency, reducing manual intervention, supporting continuous operation around the clock, and offering advantages of high automation and high irrigation efficiency. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating the workflow of the hilly terrain adaptive intelligent irrigation robot provided in this embodiment of the invention; Figure 2 This is a schematic diagram of the structure of the hilly terrain adaptive intelligent irrigation robot provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of the pneumatic irrigation belt recovery device provided in an embodiment of the present invention.

[0020] The labels for the various figures are as follows: 1. Tracked mobile chassis; 2. Robotic arm; 3. Vision navigation module; 4. Pneumatic irrigation belt; 5. Irrigation hose; 6. Multimodal irrigation device. Detailed Implementation

[0021] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0022] In the description of this invention, it should be understood that the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0023] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0024] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0025] Please see Figures 1 to 3 This application provides an adaptive intelligent irrigation robot for hilly terrain, including a tracked mobile chassis 1, a multimodal irrigation device 6, a visual navigation module 3, and a data processing and decision-making unit. The multimodal irrigation device 6 is a water tap with a changeable irrigation mode, fixed to the ground, and is used to execute at least two different irrigation modes. The visual navigation module 3 is used to collect and process environmental image information for navigation and target recognition. The data processing and decision-making unit is electrically connected to the visual navigation module 3, the multimodal irrigation device 6, and the tracked mobile chassis 1. Based on the information processed by the visual navigation module 3, the data processing and decision-making unit controls the movement of the tracked mobile chassis 1 and the operating mode of the multimodal irrigation device 6.

[0026] Specifically, the tracked mobile chassis 1 serves as the system's mobile platform, and its structure is as follows: Figure 1 and Figure 2 As shown, it is used to support the upper functional modules and moves through complex environments such as hills and slopes with its track structure.

[0027] The multimodal irrigation device 6 is the core of precision irrigation, capable of performing both drip irrigation and micro-sprinkler irrigation modes. Specifically, the multimodal irrigation device is a water tap that can change the irrigation mode and is fixed to the ground.

[0028] The visual navigation module 3 is installed on the upper part of the robot and is usually a camera (such as...). Figure 2 As shown in the image, it replaces the traditional GPS / RTK positioning module and is responsible for collecting image information in the field for path recognition, crop type recognition, and positioning of key landmarks (such as water hydrants).

[0029] The data processing and decision-making unit is the robot's "brain," receiving and analyzing data from the vision module. Internally, this unit runs SLAM algorithms and an irrigation decision engine. Based on the analysis results, it sends movement commands to the tracked chassis for autonomous navigation, and simultaneously sends commands to the irrigation device to control it to switch to the appropriate irrigation mode at the right time. These four core components work together, enabling the robot to autonomously complete the entire process from environmental perception, path planning, intelligent decision-making to precise irrigation.

[0030] In one embodiment, the tracked mobile chassis 1 integrates a tilt adaptive system, which can dynamically adjust the speed difference between the two tracks on both sides of the vehicle according to the real-time monitored vehicle slope, so as to maintain vehicle stability when traveling on slopes.

[0031] Optionally, one or more tilt sensors (part of the IMU) can be installed inside the robot's chassis. These sensors monitor the robot's tilt angle in the X and Y axes in real time. When the robot is traveling on, for example, a steep slope of 35°, the sensor data is sent to the chassis controller. The controller executes a differential speed adjustment algorithm based on a preset stability threshold. For example, when the robot turns left to climb a slope, the vehicle will tilt to the right. At this time, the controller will reduce the motor speed of the inner (left) track to increase its grip and adhesion, while appropriately increasing the speed of the outer (right) track to generate a torque to resist sideslip, thereby ensuring stable vehicle posture and effectively preventing sideslip or rollover accidents when operating on slopes.

[0032] In one embodiment, the multimodal irrigation device 6 includes a drip irrigation mode and a micro-spraying mode, which are controlled and switched by a solenoid valve.

[0033] The multimodal irrigation device 6 physically integrates two sets of water outlet pipelines: one for drip irrigation, characterized by a small water flow that directly targets crop roots; and the other for micro-sprinkler irrigation, characterized by forming a mist-like water curtain with wider coverage. Switching between these two pipelines is precisely controlled by a solenoid valve. When the data processing and decision-making unit issues a switching command, the controller applies an electrical signal to the solenoid valve, causing its valve core to actuate and connecting the main water path to either the drip irrigation or micro-sprinkler pipeline. The entire switching process is rapid, with a switching time of only 0.2 seconds, ensuring seamless transitions between irrigation modes when the robot moves between different crop areas. Simultaneously, the system can precisely control the water pump output, achieving a water flow control accuracy of 0.5 liters per minute.

[0034] In one embodiment, the data processing and decision-making unit includes an irrigation decision engine that can automatically select drip irrigation mode or micro-sprinkler mode based on the crop type identified by the visual navigation module 3.

[0035] Specifically, the engine internally uses a pre-defined rule base or decision tree, which maps the crop types identified by the vision module to the optimal irrigation mode. For example, the rule base defines: "If crop type = tea tree, then irrigation mode = drip irrigation"; "If crop type = fruit tree, then irrigation mode = micro-sprinkler". When the visual navigation module 3 uses a deep learning model to identify the crop ahead as "tea tree", the identification result is passed to the irrigation decision engine. The engine queries the rule base, matches the corresponding rule, and immediately generates a "execute drip irrigation" instruction, which is then sent to the irrigation execution mechanism in the control execution layer. The entire process from identification to decision to instruction issuance has a response latency of less than 0.5 seconds.

[0036] In one embodiment, the visual navigation module 3 uses a deep learning model to identify crop types. When the identification result is a drought-resistant crop such as tea tree or strawberry, the irrigation decision engine selects the drip irrigation mode; when the identification result is a moisture-loving crop such as fruit tree, the micro-sprinkler mode is selected.

[0037] The crop type identification function is primarily implemented using deep learning. In practice, a pre-trained lightweight convolutional neural network model, such as ResNet-18, is deployed on an embedded computing platform (such as Jetson Nano).

[0038] 1. Model loading: When the system starts, the ONNX format model file stored locally is loaded through the OpenCV DNN module function cv2.dnn.readNetFromONNX().

[0039] 2. Image preprocessing: The real-time image frames captured by the camera are first preprocessed using the blobFromImage() function, including scaling the image size to the required 224×224 pixels for model input and normalization.

[0040] 3. Model Inference: The preprocessed image data (blob) is fed into the ResNet-18 network for forward propagation calculation, and outputs a vector containing the confidence scores of various crops.

[0041] 4. Output Results: The system selects the category with the highest confidence level as the final recognition result. This scheme supports the recognition of at least 7 types of crops, including tea trees and fruit trees, with an accuracy rate of up to 95% in actual tests. To improve robustness under complex lighting conditions such as rain and under forest cover, the system also calls the cv2.equalizeHist() function in the preprocessing stage to perform histogram equalization on the image, thereby enhancing image contrast.

[0042] In one embodiment, the visual navigation module 3 is based on the OpenCV vision library and uses image processing algorithms to identify field paths and visually locate field water hydrants.

[0043] The visual navigation module 3 completely replaces the high-cost GPS / RTK system. It achieves environmental perception through one or more cameras and by running image processing algorithms based on the OpenCV library.

[0044] Field path recognition: The edge lines of the field ridges or work paths are identified through a series of image processing procedures (see the description in claim 10 for details), thereby guiding the robot to move along a predetermined route.

[0045] Visual localization of water hydrants: Water hydrants serve as fixed artificial markers in farmland. The system searches for and identifies the specific shape or pattern of the water hydrant within the camera's field of view using template matching or feature point matching algorithms (such as ORB). Once successfully identified, its pixel coordinates in the image, combined with camera intrinsic and extrinsic parameters, can be used to calculate the precise distance and azimuth angle of the robot relative to the water hydrant, achieving millimeter-level localization (error ≤ 5mm).

[0046] A robotic arm 2 is installed on the tracked mobile chassis 1. The robotic arm 2 grabs the irrigation hose 5 from the water supply hydrant and uses the tracked mobile chassis 1 to pull the irrigation hose 5 to the area that needs irrigation.

[0047] In one embodiment, the data processing and decision-making unit employs a Simultaneous Localization and Mapping (SLAM) algorithm, which integrates the water hydrant location information collected by the visual navigation module 3 to achieve high-precision autonomous positioning and optimal path planning for the robot.

[0048] Simultaneous Localization and Mapping (SLAM) is the core of the data processing and decision-making unit's high-precision autonomous navigation. In practice, the SLAM algorithm continuously fuses data from multiple sensors. It uses the water hydrants identified by the visual navigation module 3 as stable and unique "landmarks." As the robot moves through the farmland, it continuously detects and locates water hydrants within its field of vision, using these landmarks to correct its position calculated by the wheel speed sensor or IMU, effectively solving the sensor drift problem and gradually building a farmland environment map containing the precise locations of the water hydrants. Based on this real-time updated map, the optimal path planning module (such as the A* algorithm) can calculate the shortest and safest path from the current position to the next target point (the next water hydrant or irrigated area), actively avoiding obstacles such as ditches and wetlands marked on the map.

[0049] In one embodiment, a pneumatic irrigation hose 4 recovery device is also included for automatically retracting the irrigation hose 5 after the irrigation operation is completed.

[0050] The pneumatic irrigation belt recovery device is an automated auxiliary device installed on the robot body. Its structural diagram is shown below. Figure 3 As shown. Its main function is to automatically and quickly retract the irrigation hose 5 (or water belt) dragged on the ground onto the rollers when the robot completes the irrigation task in one area and is ready to move to the next area, thereby avoiding the trouble of manual retrieval and preventing the hose from being dragged, damaged, or tripping up nighttime workers during the robot's movement.

[0051] In one embodiment, the pneumatic irrigation hose 4 recovery device uses compressed air to drive rollers for high-speed winding and is equipped with a tension sensor that automatically reduces the winding speed to prevent the irrigation hose 5 from breaking when abnormal winding tension is detected.

[0052] The working principle of this pneumatic retraction device is as follows: The core of the device is a pneumatic motor driven by compressed air, with a small air pump on the robot providing approximately 0.4 MPa of air pressure. When a retraction command is issued, the solenoid valve opens the air circuit, and the high-pressure gas drives the pneumatic motor to rotate the roller at high speed, thus achieving rapid retraction of the irrigation hose 5. Its efficiency can reach 5 seconds to retract a 10-meter-long hose. To ensure the safety and reliability of the retraction process, a tension sensor is integrated on the roller shaft. During the retraction process, this sensor monitors the hose tension in real time. If the hose is stuck by stones or crops, the tension will increase instantaneously. When the tension exceeds a preset safety threshold, the sensor immediately sends a signal to the controller, which then instantly reduces the speed of the pneumatic motor or pauses it. Operation resumes only after the obstacle is removed, effectively preventing hose breakage due to violent pulling.

[0053] In one embodiment, the image processing algorithm of the visual navigation module 3 for the field path includes the following steps: converting the image to grayscale, performing Gaussian filtering for noise reduction, using the Canny operator for edge detection, and fitting a straight path using the Hough transform.

[0054] The specific implementation steps of the field path recognition image processing algorithm in the OpenCV library are as follows: 1. Grayscale conversion: Call the cv2.cvtColor() function to convert the BGR three-channel color image captured by the camera into a single-channel grayscale image to reduce the complexity of subsequent calculations.

[0055] 2. Gaussian filtering: Call the cv2.GaussianBlur() function to smooth the grayscale image. The typical Gaussian kernel size parameter (ksize) is set to (5,5) to suppress the interference of random noise in the image on edge detection.

[0056] 3. Canny Edge Detection: Call the cv2.Canny() function, a classic edge detection algorithm. In this embodiment, its low and high thresholds can be set to 50 and 150 respectively to extract clear outlines of field ridges or paths.

[0057] 4. Hough Transform: After obtaining the binarized edge image, the cv2.HoughLinesP() function (probabilistic Hough Transform) is called to fit straight lines from the pixels. Its parameters can be set to: a voting threshold of 50 and a minimum line length of 100 pixels to filter out short and discontinuous lines, ultimately outputting a set of straight line segments representing the path direction. Through this combination of steps, the system can stably and accurately identify paths suitable for the robot to navigate.

[0058] This application provides an adaptive intelligent irrigation robot for hilly terrain. Its workflow is mainly divided into three levels: environmental perception, data processing and decision-making, and control execution.

[0059] I. Environmental Perception Layer The environment perception layer is the entry point for the robot to interact with its external environment. Its core is the OpenCV vision module, which is mounted on a gimbal on top of the robot (e.g., ...). Figure 1 As shown in the image), this module is responsible for collecting and initially processing field image information. This module primarily performs two functions: 1. Field path identification and navigation positioning: After the robot starts, the camera captures a real-time video stream of the path ahead. The image is first converted to grayscale using the cv2.cvtColor() function to reduce the computational load.

[0060] Then, a Gaussian filter (e.g., kernel size set to (5,5)) is performed using cv2.GaussianBlur() to remove image noise.

[0061] Use the cv2.Canny() edge detection algorithm (e.g., low threshold 50, high threshold 150) to extract the contours of ridges or paths.

[0062] To address situations where path lines break in an image, morphological operations (such as closing operations) using cv2.morphologyEx() are used to connect them.

[0063] Finally, the straight path is probabilistically detected and fitted using the cv2.HoughLinesP() Hough transform, and its angle error can be controlled within 1°.

[0064] In environments with severe weed interference, horizontal zonal projection combined with K-means clustering can effectively eliminate noisy interference points. Simultaneously, the ExG (Super Green) index is used to enhance the contrast between crop rows and the soil background, improving the robustness of path recognition.

[0065] 2. Crop type identification: This invention prioritizes a deep learning approach. The system loads a pre-trained ResNet-18 model (using the cv2.dnn.readNetFromONNX() function).

[0066] The acquired image frames are preprocessed using the blobFromImage() function, including resizing (e.g., adjusting to 224×224 pixels) and normalization.

[0067] The preprocessed image is fed into a neural network for inference, achieving a single-frame processing speed of less than 0.3 seconds on embedded platforms such as Jetson Nano.

[0068] The model can accurately identify at least 7 common crops (such as tea trees, fruit trees, strawberries, etc.) with a classification accuracy of 95%. The identification results (if the crop is a fruit tree) will directly trigger subsequent irrigation decisions (switching to micro-sprinkler mode).

[0069] To cope with different lighting conditions, the system also integrates the cv2.equalizeHist() histogram equalization function in the image preprocessing stage, which effectively compensates for the impact of insufficient lighting environments such as rainy days or under forests on recognition performance.

[0070] II. Data Processing and Decision-Making This layer is the robot's intelligent core, responsible for processing data from the perception layer and making optimal action decisions. Its workflow consists of two parallel logical lines: 1. Spatial localization and path planning (SLAM algorithm): The system runs the SLAM algorithm, which merges image data from the vision module with attitude data from the IMU (Inertial Measurement Unit) to construct a real-time map of the farmland.

[0071] A key positioning method is to identify and locate pre-installed water taps (water supply interfaces) in the field. Through visual feature matching, the robot can accurately calculate its relative position to the water tap, with a positioning accuracy of within 5 millimeters.

[0072] Based on the constructed map, the system plans an optimal navigation route from the current location to the target water tap or the next irrigation area. This route will actively avoid obstacles marked on the map, such as ditches or excessively muddy areas.

[0073] 2. Irrigation Decision Engine: The engine receives crop identification results from the environment perception layer.

[0074] An internally pre-defined rule base associates crop types with optimal irrigation patterns. For example: When the identification result is a drought-resistant crop such as tea tree or strawberry, or a crop whose root system requires precise watering, the decision engine selects the drip irrigation mode.

[0075] When the identification result is fruit trees, vegetables, or other crops that prefer moisture or require a larger coverage area, the decision engine selects the micro-spraying mode.

[0076] Once a decision is made, instructions are immediately sent to the control execution layer.

[0077] These two logic lines are linked in real time, ensuring efficient coordination between the robot's movement and irrigation actions. For example, when the robot is on its way to the orchard area, the SLAM algorithm has already located the location of the next water tap and planned the path, while the irrigation decision engine has already determined in advance that it will start the micro-sprinkler mode upon arrival, with the entire decision delay being less than 0.5 seconds.

[0078] III. Control Execution Layer This layer is the robot's physical execution end, responsible for translating the instructions from the decision-making layer into specific mechanical actions.

[0079] 1. Tracked chassis control: The core component is the differential speed control mechanism. Tilt sensors inside the robot chassis monitor the robot's posture in real time.

[0080] When the robot is traveling on a slope (for example, a slope of 35°), the control system will automatically reduce the speed of the inner track to enhance traction based on the tilt angle data, while appropriately increasing the speed of the outer track to counteract the tendency to sideslip, thereby ensuring the stability of the vehicle and preventing overturning.

[0081] 2. Irrigation implementing agency: Upon receiving instructions from the decision-making level, a solenoid valve is controlled to quickly switch between drip irrigation and micro-spraying modes.

[0082] The switching response time is extremely short, requiring only 0.2 seconds. For example, switching from drip irrigation for tea trees to micro-sprinkler irrigation for fruit trees is almost instantaneous, ensuring the continuity and precision of irrigation. The device can achieve water flow control accurate to 0.5 liters per minute.

[0083] 3. Pneumatic recovery device: like Figure 3 As shown, the device activates after the irrigation task is completed. It uses approximately 0.4 MPa of compressed air to drive internal rollers to rotate at high speed, automatically retracting the irrigation hose 5 (water hose) laid on the ground like a measuring tape. Retracting a 10-meter-long hose takes only 5 seconds.

[0084] The device has a built-in tension sensor. Once it detects that the hose is stuck and the tension is abnormally increased, it will immediately reduce the winding speed or stop to prevent the hose from being forcibly broken, thus improving the reliability and safety of the device.

[0085] Example 1: When the robot operates in a hilly area with mixed plantings of tea gardens and orchards, its workflow is as follows: 1. The robot starts up, identifies the field path through the OpenCV vision module, and uses the SLAM algorithm to locate the first water tap, then moves autonomously and approaches it.

[0086] 2. When traveling through a tea plantation area, the differential speed control system adjusts the speed of the tracks on both sides in real time according to the slope to keep the vehicle stable.

[0087] 3. At the same time, the crop identification system determines that the current crop is a tea tree, and the decision engine issues a "drip irrigation" command.

[0088] 4. Upon arrival at the work site, the irrigation actuator activates the solenoid valve to switch to drip irrigation mode, precisely replenishing water to the roots of the tea trees.

[0089] 5. After completing irrigation in this area, the robot plans its path to the next orchard area. During the movement, the pneumatic recovery device quickly retrieves the irrigation strip from the previous section.

[0090] 6. After entering the fruit tree area, the system identifies the fruit trees and seamlessly switches to micro-spraying mode to spray the fruit trees with mist.

[0091] 7. After all tasks are completed, the robot automatically returns to the charging station to replenish its energy, and the entire process requires no human intervention.

[0092] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A hilly terrain adaptive intelligent irrigation robot, characterized in that, The hilly terrain adaptive intelligent irrigation robot includes: Tracked mobile chassis; A multimodal irrigation device, wherein the multimodal irrigation device is a water tap that can change the irrigation mode and is fixed to the ground, the multimodal irrigation device being used to perform at least two different irrigation modes; The visual navigation module is used to acquire and process environmental image information for navigation and target recognition. The data processing and decision-making unit is electrically connected to the visual navigation module, the multimodal irrigation device, and the tracked mobile chassis. The data processing and decision-making unit controls the movement of the tracked mobile chassis and the operation mode of the multimodal irrigation device based on the information processed by the visual navigation module.

2. The hilly terrain adaptive intelligent irrigation robot according to claim 1, characterized in that: The tracked mobile chassis integrates a tilt adaptive system, which can dynamically adjust the speed difference between the two tracks based on the real-time monitored vehicle slope to maintain vehicle stability when traveling on slopes.

3. The hilly terrain adaptive intelligent irrigation robot according to claim 1, characterized in that: The multimodal irrigation device includes drip irrigation mode and micro-spraying mode, which are controlled and switched by a solenoid valve.

4. A hilly terrain adaptive intelligent irrigation robot according to claim 1 or 3, characterized in that: The data processing and decision-making unit includes an irrigation decision engine that can automatically select the drip irrigation mode or micro-sprinkler mode based on the crop type identified by the visual navigation module.

5. The hilly terrain adaptive intelligent irrigation robot according to claim 4, characterized in that: The visual navigation module uses a deep learning model to identify crop types. When the identification result is a drought-resistant crop such as tea tree or strawberry, the irrigation decision engine selects drip irrigation mode; when the identification result is a moisture-loving crop such as fruit tree, micro-sprinkler mode is selected.

6. The hilly terrain adaptive intelligent irrigation robot according to claim 1, characterized in that: The visual navigation module is based on the OpenCV vision library and uses image processing algorithms to identify field paths and visually locate field water hydrants.

7. The hilly terrain adaptive intelligent irrigation robot according to claim 6, characterized in that: The data processing and decision-making unit adopts the Simultaneous Localization and Mapping (SLAM) algorithm, which integrates the water hydrant location information collected by the visual navigation module to achieve high-precision autonomous positioning and optimal path planning for the robot.

8. The hilly terrain adaptive intelligent irrigation robot according to claim 1, characterized in that: It also includes a pneumatic irrigation hose reel device for automatically reeling in the irrigation hose after the irrigation operation is completed.

9. The hilly terrain adaptive intelligent irrigation robot according to claim 8, characterized in that: The pneumatic irrigation tape recovery device uses compressed air to drive rollers for high-speed winding and is equipped with a tension sensor that automatically reduces the winding speed to prevent the irrigation hose from breaking when abnormal winding tension is detected.

10. The hilly terrain adaptive intelligent irrigation robot according to claim 6, characterized in that: The visual navigation module's image processing algorithm for field paths includes the following steps: converting the image to grayscale, performing Gaussian filtering for noise reduction, using the Canny operator for edge detection, and fitting a straight path using the Hough transform.

Images