Control methods and systems for weeding robots in photovoltaic environments

Abstract

This invention discloses a control method and system for a weeding robot suitable for photovoltaic environments, belonging to the field of intelligent control technology. The method includes: when the weeding robot autonomously moves to a starting point, it scans the gaps between photovoltaic panels, identifies weed areas based on the scanned images, and generates a weed density map; simultaneously collects obstacle data to construct a three-dimensional obstacle map; performs weeding path analysis and dual-mode weeding unit weeding parameter analysis to generate weeding control decisions; based on the weeding control decisions, it analyzes the frequency of weeding mode switching, establishes unstable switching intervals caused by frequent switching, optimizes stable switching of the weeding mode, and generates optimized switching parameters; after adjusting the weeding control decisions with the optimized switching parameters, the results are input to the control terminal of the weeding robot for weeding control. This solves the technical problem of low weeding efficiency in existing weeding robots, achieving the technical effect of improving weeding efficiency.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, specifically to a control method and system for weeding robots applicable to photovoltaic environments. Background Technology

[0002] With the widespread application of photovoltaic (PV) power generation technology and the continuous expansion of PV power plant scale, the problem of weed growth under and between PV panels has become increasingly prominent. Weeds not only block sunlight, affecting the power generation efficiency of PV modules, but also pose safety hazards such as equipment overheating and fires. Therefore, timely and effective weed control in PV environments has become a crucial aspect of ensuring the operation of PV power plants. While existing weeding robots can replace manual weeding to some extent, they suffer from problems such as insufficient weed recognition accuracy, slow obstacle avoidance response, unreasonable weeding path planning, and frequent switching of weeding modes leading to operational instability in environments with densely arranged PV panels. These issues affect weeding efficiency and operational continuity, making it difficult to meet the weeding needs of actual PV scenarios. Summary of the Invention

[0003] This application provides a control method and system for weeding robots suitable for photovoltaic environments, which solves the technical problems of insufficient obstacle avoidance ability of weeding robots in densely populated photovoltaic areas and low operation efficiency and unstable weeding effect caused by frequent switching of weeding modes in the prior art.

[0004] The first aspect of this application provides a control method for a weeding robot suitable for a photovoltaic environment, the method comprising:

[0005] When the weeding robot autonomously moves to the starting point, it scans the gaps between the photovoltaic panels using its onboard multispectral recognition module. Based on the scanned images, it identifies weed areas and generates a weed density map. Simultaneously, it uses its onboard multi-sensor obstacle avoidance module to collect obstacle data and construct a three-dimensional obstacle map. Combining the weed density map and the three-dimensional obstacle map, it performs weeding path analysis and dual-mode weeding unit weeding parameter analysis to generate weeding control decisions. Based on these weeding control decisions, it analyzes the frequency of weeding mode switching, establishes unstable switching intervals caused by frequent switching, optimizes stable switching of weeding modes, and generates optimized switching parameters. After adjusting the weeding control decisions with these optimized switching parameters, the results are input to the weeding robot's control terminal for weeding control.

[0006] A second aspect of this application provides a weeding robot control system suitable for photovoltaic environments, the system comprising:

[0007] The system comprises the following components: a scanning component, used by the multispectral recognition module to scan the gaps between photovoltaic panels when the weeding robot autonomously moves to the starting point, identifies weed areas based on the scanned images, and generates a weed density map; a data acquisition component, used to simultaneously collect obstacle data using the multi-sensor obstacle avoidance module on the weeding robot, and construct a three-dimensional obstacle map; an analysis component, used to combine the weed density map and the three-dimensional obstacle map to perform weeding path analysis and weeding parameter analysis of the dual-mode weeding unit, generating weeding control decisions; an optimization component, used to analyze the frequency of weeding mode switching based on the weeding control decisions, establish unstable switching intervals caused by frequent switching, optimize stable switching of the weeding mode, and generate optimized switching parameters; and a control component, used to adjust the weeding control decisions with the optimized switching parameters and input them into the control terminal of the weeding robot for weeding control.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] When the weeding robot autonomously moves to the starting point, it scans the gaps between the photovoltaic panels using its onboard multispectral recognition module. Based on the scanned images, it identifies the weed areas and generates a weed density map. Simultaneously, the robot's multi-sensor obstacle avoidance module collects obstacle data and constructs a 3D obstacle map. Combining the weed density map and the 3D obstacle map, it performs weeding path analysis and dual-mode weeding unit weeding parameter analysis to generate weeding control decisions. Furthermore, based on the weeding control decisions, it analyzes the frequency of weeding mode switching, establishes unstable switching ranges caused by frequent switching, and optimizes stable weeding mode switching to generate optimized switching parameters. Finally, the optimized switching parameters are used to adjust the weeding control decisions before being input into the robot's control terminal for weeding control. This solves the technical problems of insufficient obstacle avoidance capability of weeding robots in densely populated photovoltaic panel areas and low operating efficiency and unstable weeding effects caused by frequent weeding mode switching in existing technologies, achieving the technical effect of improving weeding efficiency and operational reliability. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0011] Figure 1 This is a schematic flowchart of a weeding robot control method applicable to a photovoltaic environment, provided in an embodiment of this application.

[0012] Figure 2This is a schematic diagram of the control system structure for a weeding robot suitable for a photovoltaic environment, provided in an embodiment of this application.

[0013] Explanation of reference numerals in the attached diagram: Scanning component 11, Data acquisition component 12, Analysis component 13, Optimization component 14, Control component 15. Detailed Implementation

[0014] This application provides a control method and system for weeding robots suitable for photovoltaic environments, which solves the technical problems of insufficient obstacle avoidance ability of weeding robots in densely populated photovoltaic areas and low operating efficiency and unstable weeding effect caused by frequent switching of weeding modes in the prior art.

[0015] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0016] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0017] Example 1, as Figure 1 As shown, this application provides a control method for a weeding robot suitable for photovoltaic environments, wherein the method includes:

[0018] When the weeding robot moves autonomously to the starting point, it scans the gaps between the photovoltaic panels using its onboard multispectral recognition module, identifies the weed areas based on the scanned images, and generates a weed density map.

[0019] In this embodiment, after the weeding robot autonomously moves to the starting point, it scans the gaps between the photovoltaic panels using its onboard multispectral recognition module. The multispectral recognition module includes at least one RGB camera and one near-infrared camera to acquire image data in different spectral bands. By analyzing this image data and combining it with image processing algorithms, the robot identifies the weeds in the area below the photovoltaic panels. Specifically, the image data is processed using a deep learning model, particularly a weed recognition network trained on a YOLOv7-based improved neural network, to efficiently and accurately locate the presence and distribution of weeds. Based on these scanned images, a map containing weed density information for different areas (a weed density map) is generated. The weed density map displays the distribution and density of weeds by region.

[0020] Furthermore, the multispectral recognition module includes RGB and near-infrared cameras and a weed recognition network trained based on a YOLOv7 improved neural network; the multi-sensor obstacle avoidance module includes millimeter-wave radar, a ToF depth camera, and an ultrasonic sensor; the dual-mode weeding unit includes a first mode and a second mode, wherein the first mode uses a laser weeding head for spot weeding inactivation, and the second mode uses a titanium alloy rotary cutting blade assembly to cut the above-ground parts, supplemented by a laser weeding head for secondary inactivation of the cuts; the weeding robot also includes a tracked chassis, the mowing height of which can be adjusted according to the tilt angle of the photovoltaic panels.

[0021] The multispectral recognition module includes at least one RGB camera and one near-infrared camera to acquire image data in different wavelengths, thereby enhancing the ability to identify weeds in the area under the photovoltaic panels. The RGB camera is mainly responsible for capturing visible light images, while the near-infrared camera is mainly used to acquire images in the near-infrared band. This is particularly important for identifying different types of weeds, especially providing higher accuracy in distinguishing weeds from soil and vegetation.

[0022] The multispectral recognition module also includes a weed recognition network trained on an improved YOLOv7 neural network. This network can efficiently and accurately identify weed areas in scanned images in real time and generate precise weed location and density information, forming a weed density identification map that can be used for subsequent processing.

[0023] The multi-sensor obstacle avoidance module includes millimeter-wave radar, a ToF (Time-of-Flight) depth camera, and ultrasonic sensors. Combining data from these three sensors, the lawnmower robot can achieve precise obstacle recognition and avoidance in complex environments. Millimeter-wave radar can detect obstacles at long distances and provide distance and velocity information; the ToF depth camera calculates the three-dimensional spatial position of objects by analyzing laser reflection time, providing the robot with detailed spatial information; and the ultrasonic sensor is used for close-range obstacle detection, further enhancing the flexibility and stability of the obstacle avoidance module.

[0024] The dual-mode weeding unit includes a first mode and a second mode. The first mode uses a laser weeding head for targeted weed inactivation. The laser head precisely targets weeds with a high-energy-density laser beam, and the instantaneous high temperature effectively inactivates the weeds. The second mode uses a titanium alloy rotary cutting blade assembly to cut the above-ground parts of the weeds, uprooting them and avoiding the problem of only cutting the surface. Simultaneously, the laser weeding head also performs secondary inactivation on the roots of the cut weeds, ensuring that they do not regrow.

[0025] The weeding robot also includes a tracked chassis, which offers excellent adaptability and maneuverability, enabling stable movement on uneven terrain. The weeding height of the tracked chassis can be automatically adjusted according to the tilt angle of the solar panels, ensuring that the robot maintains the optimal weeding height at different panel tilt angles, thereby improving weeding efficiency and operational stability.

[0026] Simultaneously, the multi-sensor obstacle avoidance module on the weeding robot is used to collect obstacle data and construct a three-dimensional obstacle map.

[0027] In this embodiment, a multi-sensor obstacle avoidance module onboard the weeding robot is used to collect obstacle data and construct a 3D obstacle map. The multi-sensor obstacle avoidance module includes millimeter-wave radar, a ToF (Time-of-Flight) depth camera, and an ultrasonic sensor. Each of these sensors provides different types of environmental data. The millimeter-wave radar, by emitting and receiving electromagnetic waves, can detect the presence of distant obstacles and provide information such as the relative position and speed of the obstacles. The ToF depth camera, by measuring the time it takes for light to travel from the sensor to the object and back, accurately calculates the depth information of the surrounding environment, generating a high-resolution 3D point cloud map, thus providing more precise spatial positioning of obstacles. The ultrasonic sensor detects nearby obstacles by measuring the sound wave reflection time. These sensors work together to collect environmental data in real time. After data fusion and processing, a complete 3D obstacle map is formed. The 3D obstacle map describes the spatial layout of obstacles around the robot, including feature information such as the distance, shape, and size of obstacles. With the 3D obstacle map, the weeding robot can perform precise path planning in the complex environment of a photovoltaic power station, avoiding collisions with obstacles and ensuring the safety and efficiency of the operation.

[0028] By combining the weed density map and the three-dimensional obstacle map, weeding path analysis and weeding parameter analysis of the dual-mode weeding unit are performed to generate weeding control decisions.

[0029] Based on a weed density map, the distribution and density of weeds are identified. By analyzing the weed distribution characteristics of each area, key weed-removal areas are determined, typically those with high weed density or vigorous growth. The weeding robot will focus its efforts on these areas. Simultaneously, combined with a 3D obstacle map, the weeding robot can accurately understand the distribution of obstacles in its surroundings, including the location, size, and shape of objects such as solar panels, supports, and rocks. A safe and efficient operating path is planned based on the 3D obstacle map, avoiding collisions or interference with obstacles and ensuring the robot maintains the optimal movement path during weeding. After path planning, the weeding parameters of the dual-mode weeding unit are analyzed to determine which weeding mode is most efficient in different areas. For example, in areas with dense weeds, a laser weeding head is used for spot-shooting inactivation, quickly and accurately removing weeds; while in areas with more scattered weeds or requiring root removal, a titanium alloy rotary cutting blade assembly is used for cutting, supplemented by a laser weeding head for secondary inactivation, ensuring thorough weed removal.

[0030] Furthermore, by combining the weed density map with the three-dimensional obstacle map to perform weeding path analysis and dual-mode weeding unit weeding parameter analysis, weeding control decisions are generated, including:

[0031] The first weed density selection range and the relationship between the first control parameter based on weed density for the first mode and the second weed density selection range and the relationship between the second control parameter based on weed density for the second mode are determined in the dual-mode weeding unit. The matching area of ​​the first and second weed density selection ranges is located based on the weed density identification map, and the first weeding control parameter and the second weeding control parameter corresponding to the matching area of ​​the first mode are determined based on the first and second control parameter relationships. The weeding path is planned for the weeding robot based on the three-dimensional obstacle map, and the ground clearance parameter of the tracked chassis is determined based on the photovoltaic panel tilt angle. The weeding control decision is generated using the weeding path, ground clearance parameter, first weeding control parameter of the matching area of ​​the first mode, and second weeding control parameter corresponding to the matching area of ​​the second mode.

[0032] Based on a weed density map, the operating ranges of the first and second modes in the dual-mode weeding unit are determined. Specifically, for the first mode, a first weed density selection range is set based on the weed density map. This range is identified by analyzing areas with different weed densities to determine areas suitable for weeding using the first mode (e.g., laser weeding head spot inactivation). Simultaneously, based on the first weed density selection range, a first control parameter relationship based on weed density is determined. This first control parameter relationship includes parameters such as the power, irradiation time, and irradiation frequency of the weeding equipment (e.g., laser weeding head) in the first mode within this density range. This ensures that the laser can accurately target and effectively inactivate the weeds within this area, avoiding energy waste or incomplete inactivation.

[0033] For the second mode, similarly, a second weed density selection range is set. This range mainly includes areas with low or scattered weed density, where the laser weeding head is unsuitable. In these areas, the second mode (e.g., a titanium alloy rotary cutting blade assembly) will perform ground cutting operations. Based on the second weed density selection range, a second control parameter relationship is determined. This second control parameter relationship defines parameters such as the rotation speed, cutting depth, and working angle of the blade assembly in the second mode, ensuring effective cutting of weed roots on the ground. The laser weeding head then performs secondary inactivation on the cut surfaces to prevent weed regrowth.

[0034] Based on the weed density map, the system locates matching areas for the first and second weed density selection ranges. Specifically, the weed density map provides the spatial distribution of weeds within the photovoltaic power station area. By analyzing the weed density in different areas of the map, the system identifies which areas meet the operational requirements of both the first mode (laser weeding head) and the second mode (titanium alloy rotary cutting blade assembly). For the first mode matching area, the system identifies areas with high weed density on the weed density map; these areas are typically suitable for spot weeding with the laser weeding head. Based on the location and shape of these high-density areas, the system generates the first mode matching area. Similarly, for the second mode matching area, the system identifies areas with low or sparse weed density; these areas are suitable for cutting with the titanium alloy rotary cutting blade assembly. By analyzing these low-density areas, the system generates the second mode matching area.

[0035] Based on the first control parameter relationship and the second control parameter relationship, the weeding control parameters for each matching area are determined. Specifically, based on the weed density of the first mode matching area, the system determines the first weeding control parameters for that area according to the previously set first control parameter relationship, including parameters such as the power of the laser weeding head, irradiation time, and irradiation frequency; while based on the weed density of the second mode matching area, the system determines the second weeding control parameters for that area according to the second control parameter relationship, including parameters such as the rotation speed of the rotary blade assembly, cutting depth, and cutting angle.

[0036] Based on a 3D obstacle map, the weeding robot plans its weeding path. The 3D obstacle map provides detailed 3D information about obstacles in the robot's surrounding environment, including the location, shape, and size of obstacles such as solar panels, supports, and rocks. The optimal path is calculated based on the 3D obstacle map to ensure that the weeding robot can avoid obstacles during operation and perform the weeding task efficiently and smoothly.

[0037] Simultaneously, based on the tilt angle of the photovoltaic panels, the system automatically adjusts the ground clearance parameters of the tracked chassis. The system dynamically adjusts the ground clearance of the tracked chassis according to the tilt angle of the photovoltaic panels and the ground height below them, ensuring sufficient mobility and stability for the robot in complex terrain. Finally, by combining the generated weeding path, ground clearance parameters, first weeding control parameters for the first pattern matching area, and second weeding control parameters for the second pattern matching area, a comprehensive weeding control decision is generated. This decision serves as the input command for the weeding robot control system, ensuring that the robot efficiently and accurately performs weeding tasks within the photovoltaic power station, maximizing operational efficiency and ensuring safety.

[0038] Based on the weeding control decision, the frequency of weeding mode switching is analyzed, an unstable switching range caused by frequent switching is established, stable switching of weeding mode is optimized, and optimized switching parameters are generated.

[0039] Based on weed control decisions, the system analyzes the frequency of weeding mode switching and establishes unstable switching intervals caused by frequent switching. Specifically, weed control decisions may involve switching between different weeding modes (e.g., switching between a laser weeding head and a titanium alloy rotary cutting blade assembly). These switching typically depend on factors such as weed density, area characteristics, and path planning. Frequent mode switching can lead to system instability, increased energy consumption, and reduced weeding efficiency. By analyzing the frequency of each mode switch and identifying the intervals within which frequent switching occurs, the system determines which intervals constitute unstable switching intervals—those where frequent switching negatively impacts weeding operations.

[0040] Based on unstable switching intervals, the system optimizes stable switching of weeding modes, reducing unnecessary mode switching and avoiding frequent mode changes within these unstable intervals, thereby improving operational efficiency and robot stability. Through optimization algorithms, the system optimizes the weeding mode switching strategy based on multiple factors (such as weed density, work area characteristics, and path planning), enabling the robot to transition more smoothly to different modes during weeding, rather than engaging in frequent and ineffective mode switching. Based on the results of stable switching optimization, the system generates optimized switching parameters. These parameters include the appropriate timing, frequency, and post-switching working mode under different working conditions (such as weed density, work area characteristics, and path planning). By adjusting these parameters, the system ensures that the weeding robot operates more efficiently and stably, reducing energy consumption and improving overall weeding effectiveness.

[0041] Furthermore, based on the frequency analysis of weeding mode switching according to the aforementioned weeding control decision, an unstable switching interval caused by frequent switching is established, and stable switching of the weeding mode is optimized to generate optimized switching parameters, including:

[0042] Based on the weeding control decision, multiple mode switching nodes are identified, including different mode change nodes and same mode change nodes; based on the multiple mode switching nodes, M unstable switching intervals are identified where the continuous mode switching characteristics satisfy a preset frequent switching characteristic, where M is an integer greater than 1; in the M unstable switching intervals, weeding mode merging optimization of the switching nodes is performed to generate M merging optimization strategies; the optimized switching parameters are generated using the M merging optimization strategies.

[0043] Based on weeding control decisions, the system identifies multiple mode-switching nodes, including inter-mode change nodes and intra-mode change nodes. Inter-mode change nodes indicate that the weeding robot is switching between two different weeding modes (e.g., laser weeding head and titanium alloy rotary cutting blade assembly), while intra-mode change nodes indicate adjustments within the same weeding mode (e.g., different laser power settings or cutting depths). By identifying these mode-switching nodes, the system can track the robot's mode switching during operation.

[0044] Based on the identified multiple mode-switching nodes, the system analyzes the continuous mode-switching characteristics of these nodes and determines which continuous mode-switching patterns meet preset frequent switching characteristics. These preset frequent switching characteristics typically include conditions such as excessively frequent mode switching and switching frequencies exceeding a reasonable range. Based on these conditions, the system identifies M unstable switching intervals, where M is an integer greater than 1. Unstable switching intervals refer to intervals where mode switching is too frequent, leading to low weeding efficiency or excessive robot energy consumption.

[0045] After identifying M unstable switching intervals, the system performs a weeding and merging optimization process on the mode-switching nodes within these intervals. Through optimization algorithms, frequently switching modes are merged, making them more stable within the merged region. For example, for frequently switching heterogeneous mode nodes, the system may adjust their switching timing to reduce unnecessary switching, or fuse the operating parameters of the two modes to create a new mode, avoiding multiple switching. After merging, the system generates M merging optimization strategies, providing optimized switching strategies for each unstable switching interval.

[0046] Based on the generated M merging optimization strategies, the system calculates and generates optimized switching parameters. These optimized switching parameters include the optimal timing for switching weeding modes, the switching frequency, and the control parameters after mode adjustment (such as laser power and cutting depth) within each unstable switching interval.

[0047] Furthermore, the method for determining the preset frequent switching feature includes:

[0048] Collect historical switching process operation records of the weeding robot under different modes; extract mode switching response features and switching load from the switching process operation records, identify mode switching frequency features and operation state switching features that cause over-response, under-response, and load exceeding a preset safe load threshold, and generate the preset frequent switching features.

[0049] The system collects historical mode-switching operation records generated by the weeding robot during actual operation. These records include the time, frequency, and operating mode used during each mode switch (e.g., switching between the laser weeding head and the titanium alloy rotary cutting blade assembly). The system analyzes these records to extract mode-switching response characteristics and switching loads. Mode-switching response characteristics include the robot's reaction time, reaction speed, and stability after a mode switch, assessing the impact of each mode switch on the robot's operational performance and weeding effect. Switching load refers to the energy, computing resources, and control system load required by the robot during mode switching, which significantly affects the robot's efficiency and stability. The system further analyzes the extracted response characteristics and load data to identify mode-switching situations leading to over-response and under-response. Over-response refers to excessive energy consumption or over-operation due to large changes in control parameters caused by mode switching, while under-response refers to insufficient changes in control parameters after a mode switch, resulting in reduced weeding efficiency. Furthermore, the system also identifies switching situations where the load exceeds a preset safe load threshold, which typically leads to equipment overload or malfunction risks. Based on the above analysis, the system will generate preset frequent switching characteristics according to the mode switching frequency characteristics and the job status switching characteristics. These characteristics include the frequency of mode switching, the switching response characteristics, and whether there is over-response, under-response, or overload during the switching process.

[0050] Furthermore, over-response refers to a response state in which the energy consumption of any weeding mode in the dual-mode weeding unit exceeds the theoretical energy requirement due to response deviation, and the deviation value is greater than the preset error threshold. Under-response refers to a response state in which the energy consumption of any weeding mode in the dual-mode weeding unit is less than the theoretical energy requirement due to response deviation, and the deviation value is less than the preset error threshold.

[0051] Over-response refers to a situation where, due to response deviation, the energy consumption of any weeding mode in a dual-mode weeding unit exceeds the theoretical energy requirement, and the deviation value is greater than a preset error threshold. In this response state, the weeding robot over-responds when switching or adjusting modes during weeding operations, resulting in excessive energy consumption. For example, in areas with high-density weeds, if the laser power or cutting depth of the weeding mode is too high, exceeding the energy required for actual weeding, it will lead to energy waste by the robot and may affect the long-term stability of the equipment.

[0052] Under-response refers to a situation where, due to response deviation, the energy consumption of either weeding mode in a dual-mode weeding unit is lower than the theoretically required energy, and the deviation is less than a preset error threshold. In an under-response state, the robot fails to provide sufficient energy to perform the intended weeding task, resulting in poor weeding performance and incomplete weed removal. For example, if the power of the laser weeding head is too low, or the cutting depth of the rotary blade is insufficient, causing energy consumption to be lower than the actual requirement, the weeding effect will significantly decrease, or even fail to effectively remove weeds.

[0053] Furthermore, based on the multiple mode switching nodes, M unstable switching intervals are identified where the continuous mode switching characteristics satisfy a preset frequent switching characteristic, where M is an integer greater than 1, including:

[0054] Extract the first switching node and the adjacent second switching node from the plurality of mode switching nodes; determine whether the continuous mode switching characteristics of the first switching node and the second switching node satisfy the preset frequent switching characteristics; if so, continue to obtain the third switching node, and analyze whether the continuous mode switching characteristics of the first switching node, the second switching node and the third switching node satisfy the preset frequent switching characteristics; if the continuous mode switching characteristics of the first switching node, the second switching node and the third switching node do not satisfy the preset frequent switching characteristics, construct a first unstable switching interval with the first switching node and the second switching node, and use the third switching node as the initial analysis node to continue to analyze whether the continuous mode switching characteristics between the third switching node and the fourth switching node satisfy the preset frequent switching characteristics, and so on, traversing the plurality of mode switching nodes to construct the M unstable switching intervals.

[0055] The system extracts the first switching node and its adjacent second switching node from multiple mode switching nodes. The first and second switching nodes represent the initial switching between different modes by the weeding robot and their adjacent switching patterns. It then determines whether the continuous mode switching characteristics between the first and second switching nodes meet preset frequent switching characteristics. Preset frequent switching characteristics typically refer to excessively high mode switching frequency, unstable response characteristics during switching, or unexpected energy consumption. If the switching characteristics of these two nodes meet the preset frequent switching criteria, it indicates potential optimization space for the switching between these two nodes. If the continuous mode switching characteristics of the first and second switching nodes meet the frequent switching characteristics, the system continues to acquire a third switching node and analyzes whether the continuous mode switching characteristics between the first, second, and third switching nodes still meet the preset frequent switching characteristics. If they do, the analysis scope is further expanded to acquire more switching nodes.

[0056] If, during the analysis, the continuous mode switching characteristics of the first, second, and third switching nodes no longer satisfy the frequent switching characteristic, the system will construct a first unstable switching interval based on the first and second switching nodes. Simultaneously, the system will use the third switching node as the initial analysis node to continue analyzing the continuous mode switching characteristics between the third and fourth switching nodes to determine if the frequent switching characteristic is satisfied. If not, a new unstable switching interval will be constructed. This process will continue in this manner, traversing all mode switching nodes, ultimately constructing M unstable switching intervals.

[0057] Furthermore, in the M unstable switching intervals, the weeding mode merging optimization of the switching nodes is performed to generate M merging optimization strategies, including:

[0058] Extract the first unstable switching interval from the M unstable switching intervals; collect the switching interval features corresponding to the stable switching intervals other than the M unstable switching intervals; perform segmentation to maximize the uniformity index of weed density distribution in each mode duration within the first unstable switching interval using the switching interval features, generating N segmentation results, and marking them with the maximum weed density within the segmentation results; for the N segmentation results, perform single-mode weeding parameter configuration based on the first and second modes of the dual-mode weeding unit, and determine again whether the updated continuous mode switching features meet the preset frequent switching features. If so, generate a first merging optimization strategy using the single-mode weeding parameters corresponding to each of the N segmentation results; add the first merging optimization strategy to the M merging optimization strategies.

[0059] From M unstable switching intervals, one unstable switching interval is extracted as the first unstable switching interval for analysis. Switching interval characteristics are collected for all stable switching intervals other than the M unstable switching intervals. These characteristics typically include the time interval between each mode switch, the switching frequency, and changes in energy consumption during the switching process. Based on the switching interval characteristics of the stable switching intervals, the duration of each mode within the first unstable switching interval is segmented, and optimization is performed based on the uniformity of weed density distribution. Specifically, the system divides the first unstable switching interval into N segments by maximizing the uniformity of weed density distribution. The maximum weed density within each segment is used as a marker for that region for subsequent optimization.

[0060] For the N segmentation results generated, the system will perform single-mode weeding parameter configuration based on the first and second modes of the dual-mode weeding unit. Specifically, it will determine the most suitable working mode (first mode or second mode) for each segmentation result and configure the optimal weeding parameters for that mode, such as laser power, cutting depth, and rotation speed. After configuration, the system will re-check the updated continuous mode switching characteristics to determine if they meet the preset frequent switching characteristics. If the updated continuous mode switching characteristics meet the preset frequent switching characteristics, the system will generate a first merging optimization strategy based on the single-mode weeding parameters of each segmentation result. This strategy will include optimized weeding modes and parameters applied within specific unstable switching intervals. Finally, the system will add the first merging optimization strategy to the generated M merging optimization strategies to form the final optimization strategy set.

[0061] Furthermore, if the updated continuous mode switching feature does not meet the preset frequent switching feature, the N segmentation results are merged again according to twice the switching interval feature, until the merged continuous mode switching feature meets the preset frequent switching feature.

[0062] After configuring single-mode weeding parameters and generating a merging optimization strategy, the system checks the updated continuous mode switching characteristics. If these characteristics still do not meet the preset frequent switching criteria (i.e., the switching frequency is too high or the response during switching is unstable), the merging optimization strategy is considered unsatisfactory and may lead to unnecessary energy consumption or inefficiency. In this case, the system merges the N segmentation results again based on the switching interval characteristics (i.e., the time interval between mode switching events). Specifically, the system merges the segmentation results into a larger area, with the merging standard being twice the original switching interval characteristics, to ensure that the merged area is more stable and reduces excessively frequent mode switching. The system continuously repeats the merging process, merging each time with twice the switching interval characteristics, until the merged continuous mode switching characteristics meet the preset frequent switching characteristics, thus achieving the predetermined mode switching stability and efficiency requirements.

[0063] After adjusting the weeding control decision with the optimized switching parameters, the decision is input to the control terminal of the weeding robot for weeding control.

[0064] After generating optimized switching parameters, the system adjusts the weeding control decisions based on these parameters. By optimizing the switching timing, frequency, and related control parameters (such as laser power, cutting depth, and rotation speed), the system ensures that the weeding robot can smoothly transition to the most suitable operating mode in different work areas. Through these optimized switching parameters, the robot can perform weeding tasks more efficiently, avoiding excessive mode switching and energy waste, thus improving operational efficiency.

[0065] The adjusted weeding control decisions are input as instructions to the weeding robot's control unit. The robot's control unit, comprising a central processing unit and various execution modules, is responsible for automatically adjusting the robot's operating mode and parameters based on the received control instructions. According to the optimized weeding control decisions, the control unit adjusts the working status of equipment such as the laser weeding head and titanium alloy rotary cutting blade assembly to achieve precise weeding.

[0066] In summary, the embodiments of this application have at least the following technical effects:

[0067] When the weeding robot autonomously moves to the starting point, it scans the gaps between the photovoltaic panels using its onboard multispectral recognition module. Based on the scanned images, it identifies the weed areas and generates a weed density map. Simultaneously, the robot's multi-sensor obstacle avoidance module collects obstacle data and constructs a 3D obstacle map. Combining the weed density map and the 3D obstacle map, it performs weeding path analysis and dual-mode weeding unit weeding parameter analysis to generate weeding control decisions. Furthermore, based on the weeding control decisions, it analyzes the frequency of weeding mode switching, establishes unstable switching ranges caused by frequent switching, and optimizes stable weeding mode switching to generate optimized switching parameters. Finally, the optimized switching parameters are used to adjust the weeding control decisions before being input into the robot's control terminal for weeding control. This solves the technical problems of insufficient obstacle avoidance capability of weeding robots in densely populated photovoltaic panel areas and low operating efficiency and unstable weeding effects caused by frequent weeding mode switching in existing technologies, achieving the technical effect of improving weeding efficiency and operational reliability.

[0068] Example 2, based on the same inventive concept as the weeding robot control method applicable to photovoltaic environments in the foregoing examples, such as... Figure 2 As shown, this application provides a weeding robot control system suitable for photovoltaic environments, wherein the system includes:

[0069] The scanning component 11 is used to scan the gaps between photovoltaic panels using a multispectral recognition module when the weeding robot autonomously moves to the starting point, identify weed areas based on the scanned images, and generate a weed density map. The data acquisition component 12 is used to simultaneously collect obstacle data using the multi-sensor obstacle avoidance module on the weeding robot to construct a three-dimensional obstacle map. The analysis component 13 is used to perform weeding path analysis and dual-mode weeding unit weeding parameter analysis by combining the weed density map and the three-dimensional obstacle map to generate weeding control decisions. The optimization component 14 is used to analyze the frequency of weeding mode switching based on the weeding control decisions, establish unstable switching intervals caused by frequent switching, optimize stable switching of weeding modes, and generate optimized switching parameters. The control component 15 is used to adjust the weeding control decisions with the optimized switching parameters and input them into the control terminal of the weeding robot for weeding control.

[0070] Furthermore, the scanning component 11 is used to perform the following methods:

[0071] The multispectral recognition module includes RGB and near-infrared cameras and a weed recognition network trained based on a YOLOv7 improved neural network; the multi-sensor obstacle avoidance module includes millimeter-wave radar, a ToF depth camera, and an ultrasonic sensor; the dual-mode weeding unit includes a first mode and a second mode, wherein the first mode uses a laser weeding head for spot weeding inactivation, and the second mode uses a titanium alloy rotary cutting blade assembly to cut the above-ground parts, supplemented by a laser weeding head for secondary inactivation of the cuts; the weeding robot also includes a tracked chassis, the mowing height of which can be adjusted according to the tilt angle of the photovoltaic panels.

[0072] Furthermore, the optimization component 14 is used to perform the following method:

[0073] Based on the weeding control decision, multiple mode switching nodes are identified, including different mode change nodes and same mode change nodes; based on the multiple mode switching nodes, M unstable switching intervals are identified where the continuous mode switching characteristics satisfy a preset frequent switching characteristic, where M is an integer greater than 1; in the M unstable switching intervals, weeding mode merging optimization of the switching nodes is performed to generate M merging optimization strategies; the optimized switching parameters are generated using the M merging optimization strategies.

[0074] Furthermore, the optimization component 14 is used to perform the following method:

[0075] Collect historical switching process operation records of the weeding robot under different modes; extract mode switching response features and switching load from the switching process operation records, identify mode switching frequency features and operation state switching features that cause over-response, under-response, and load exceeding a preset safe load threshold, and generate the preset frequent switching features.

[0076] Furthermore, the optimization component 14 is used to perform the following method:

[0077] Over-response refers to a response state in which the energy consumption of any weeding mode in the dual-mode weeding unit exceeds the theoretical energy requirement due to response deviation, and the deviation value is greater than the preset error threshold. Under-response refers to a response state in which the energy consumption of any weeding mode in the dual-mode weeding unit is less than the theoretical energy requirement due to response deviation, and the deviation value is less than the preset error threshold.

[0078] Furthermore, the optimization component 14 is used to perform the following method:

[0079] Extract the first switching node and the adjacent second switching node from the plurality of mode switching nodes; determine whether the continuous mode switching characteristics of the first switching node and the second switching node satisfy the preset frequent switching characteristics; if so, continue to obtain the third switching node, and analyze whether the continuous mode switching characteristics of the first switching node, the second switching node and the third switching node satisfy the preset frequent switching characteristics; if the continuous mode switching characteristics of the first switching node, the second switching node and the third switching node do not satisfy the preset frequent switching characteristics, construct a first unstable switching interval with the first switching node and the second switching node, and use the third switching node as the initial analysis node to continue to analyze whether the continuous mode switching characteristics between the third switching node and the fourth switching node satisfy the preset frequent switching characteristics, and so on, traversing the plurality of mode switching nodes to construct the M unstable switching intervals.

[0080] Furthermore, the optimization component 14 is used to perform the following method:

[0081] Extract the first unstable switching interval from the M unstable switching intervals; collect the switching interval features corresponding to the stable switching intervals other than the M unstable switching intervals; perform segmentation to maximize the uniformity index of weed density distribution in each mode duration within the first unstable switching interval using the switching interval features, generating N segmentation results, and marking them with the maximum weed density within the segmentation results; for the N segmentation results, perform single-mode weeding parameter configuration based on the first and second modes of the dual-mode weeding unit, and determine again whether the updated continuous mode switching features meet the preset frequent switching features. If so, generate a first merging optimization strategy using the single-mode weeding parameters corresponding to each of the N segmentation results; add the first merging optimization strategy to the M merging optimization strategies.

[0082] Furthermore, the optimization component 14 is used to perform the following method:

[0083] If the updated continuous mode switching feature does not meet the preset frequent switching feature, the N segmentation results are merged again according to twice the switching interval feature, until the merged continuous mode switching feature meets the preset frequent switching feature.

[0084] Furthermore, the analysis component 13 is used to perform the following methods:

[0085] The first weed density selection range and the relationship between the first control parameter based on weed density for the first mode and the second weed density selection range and the relationship between the second control parameter based on weed density for the second mode are determined in the dual-mode weeding unit. The matching area of ​​the first and second weed density selection ranges is located based on the weed density identification map, and the first weeding control parameter and the second weeding control parameter corresponding to the matching area of ​​the first mode are determined based on the first and second control parameter relationships. The weeding path is planned for the weeding robot based on the three-dimensional obstacle map, and the ground clearance parameter of the tracked chassis is determined based on the photovoltaic panel tilt angle. The weeding control decision is generated using the weeding path, ground clearance parameter, first weeding control parameter of the matching area of ​​the first mode, and second weeding control parameter corresponding to the matching area of ​​the second mode.

[0086] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0087] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0088] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A control method for a weeding robot suitable for photovoltaic environments, characterized in that, The method includes: When the weeding robot moves autonomously to the starting point, it scans the gaps between the photovoltaic panels using its multispectral recognition module, identifies the weed areas based on the scanned images, and generates a weed density map. Simultaneously, the multi-sensor obstacle avoidance module on the weeding robot is used to collect obstacle data and construct a three-dimensional obstacle map; By combining the weed density map and the three-dimensional obstacle map, weeding path analysis and dual-mode weeding unit weeding parameter analysis are performed to generate weeding control decisions. Based on the weeding control decision, the frequency of weeding mode switching is analyzed, an unstable switching interval caused by frequent switching is established, stable switching of weeding mode is optimized, and optimized switching parameters are generated. After adjusting the weeding control decision with the optimized switching parameters, the data is input to the control terminal of the weeding robot for weeding control; based on the weeding control decision, the frequency of weeding mode switching is analyzed, an unstable switching range caused by frequent switching is established, stable switching of the weeding mode is optimized, and optimized switching parameters are generated, including: Based on the weeding control decision, multiple mode switching nodes are identified, including different mode change nodes and same mode change nodes. Based on the multiple mode switching nodes, M unstable switching intervals are identified that satisfy the preset frequent switching characteristics of continuous mode switching, where M is an integer greater than 1. In the M unstable switching intervals, weeding mode merging optimization is performed at the mode switching node to generate M merging optimization strategies; The optimized switching parameters are generated using the M merging optimization strategies. Different mode change nodes indicate that the weeding robot is switching between two different weeding modes; while same mode change nodes indicate adjustments within the same weeding mode.

2. The control method for a weeding robot suitable for a photovoltaic environment as described in claim 1, characterized in that, The multispectral recognition module includes RGB and near-infrared cameras and a weed recognition network trained based on a YOLOv7 improved neural network; the multi-sensor obstacle avoidance module includes millimeter-wave radar, a ToF depth camera, and an ultrasonic sensor. The dual-mode weeding unit includes a first mode and a second mode. In the first mode, a laser weeding head is used for spot weeding and inactivation. In the second mode, a titanium alloy rotary cutting blade assembly is used to cut the above-ground parts, and the laser weeding head is used to perform secondary inactivation on the cut surfaces. The weeding robot also includes a tracked chassis, the mowing height of which can be adjusted according to the tilt angle of the photovoltaic panels.

3. The control method for a weeding robot suitable for a photovoltaic environment as described in claim 1, characterized in that, The method for determining the preset frequent switching characteristics includes: Collect operation records of the switching process during the historical mode switching of the weeding robot; Historical mode switching response features and switching loads are extracted from the switching process operation records. Historical mode switching frequency features and operation state switching features that cause over-response, under-response, and load exceeding the preset safe load threshold are identified, and the preset frequent switching features are generated. The system collects historical mode switching records generated during actual operation of the weeding robot. These records include the time, frequency, and operating mode used during each mode switch. The collected switching records are analyzed to extract historical mode switching response characteristics and switching loads. The historical mode switching response characteristics include the robot's reaction time, reaction speed, and stability after a mode switch, assessing the impact of each mode switch on the robot's operational performance and weeding effect. Further analysis of the extracted historical mode switching response characteristics and switching loads identifies historical mode switching situations leading to over-response and under-response. Additionally, it identifies switching situations where the load exceeds a preset safe load threshold, which typically leads to equipment overload or malfunction risks. Based on the above analysis and the historical mode switching frequency characteristics and operational status switching characteristics, preset frequent switching characteristics are generated. These characteristics include the frequency of historical mode switching, switching response characteristics, and whether there are over-response, under-response, or overload situations during the switching process.

4. The control method for a weeding robot suitable for a photovoltaic environment as described in claim 3, characterized in that, Over-response refers to a response state in which the energy consumption of any weeding mode in the dual-mode weeding unit exceeds the theoretical energy requirement due to response deviation, and the deviation value is greater than the preset error threshold. Under-response refers to a response state in which the energy consumption of any weeding mode in the dual-mode weeding unit is less than the theoretical energy requirement due to response deviation, and the deviation value is less than the preset error threshold.

5. The control method for a weeding robot suitable for a photovoltaic environment as described in claim 1, characterized in that, Based on the multiple mode switching nodes, M unstable switching intervals are identified where the continuous mode switching characteristics satisfy a preset frequent switching characteristic, and M is an integer greater than 1, including: Extract the first switching node at the top and the adjacent second switching node from the plurality of mode switching nodes; Determine whether the continuous mode switching characteristics of the first switching node and the second switching node satisfy the preset frequent switching characteristics; If so, continue to acquire the third switching node and analyze whether the continuous mode switching characteristics of the first, second, and third switching nodes meet the preset frequent switching characteristics. If the continuous mode switching characteristics of the first switching node, the second switching node, and the third switching node do not meet the preset frequent switching characteristics, a first unstable switching interval is constructed using the first switching node and the second switching node, and the third switching node is used as the initial analysis node to continue analyzing whether the continuous mode switching characteristics between the third switching node and the fourth switching node meet the preset frequent switching characteristics. This process is repeated to traverse the multiple mode switching nodes and construct the M unstable switching intervals.

6. The control method for a weeding robot suitable for a photovoltaic environment as described in claim 1, characterized in that, In the M unstable switching intervals, weeding mode merging optimization is performed at the mode switching nodes to generate M merging optimization strategies, including: Extract the first unstable switching interval from the M unstable switching intervals; Collect the switching interval characteristics corresponding to the stable switching intervals other than the M unstable switching intervals; Using the switching interval feature, each mode duration phase within the first unstable switching interval is segmented to maximize the uniformity index of weed density distribution, generating N segmentation results, and marking them with the maximum weed density within the segmentation results. For the N segmentation results, execute the single-mode weeding parameter configuration based on the first and second modes of the dual-mode weeding unit, and determine again whether the updated continuous mode switching feature meets the preset frequent switching feature. If so, generate the first merging optimization strategy with the single-mode weeding parameters corresponding to each of the N segmentation results. Add the first merging optimization strategy to the M merging optimization strategies.

7. The weeding robot control method applicable to photovoltaic environments as described in claim 6, characterized in that, If the updated continuous mode switching feature does not meet the preset frequent switching feature, the N segmentation results are merged again according to twice the switching interval feature, until the merged continuous mode switching feature meets the preset frequent switching feature.

8. The control method for a weeding robot suitable for a photovoltaic environment as described in claim 1, characterized in that, By combining the weed density map and the 3D obstacle map, weeding path analysis and dual-mode weeding unit weeding parameter analysis are performed to generate weeding control decisions, including: Determine the first weed density selection range and the relationship between the first control parameter based on the weed density in the first mode of the dual-mode weeding unit, and the second weed density selection range and the relationship between the second control parameter based on the weed density in the second mode; Based on the weed density identification map, the matching area of ​​the first weed density selection range and the second weed density selection range is located, and the first weeding control parameter of the first pattern matching area and the second weeding control parameter corresponding to the second pattern matching area are determined based on the first control parameter relationship and the second control parameter relationship. The weeding robot plans a weeding path based on the three-dimensional obstacle map, and determines the ground clearance parameter of the tracked chassis based on the tilt angle of the photovoltaic panel. The weeding control decision is generated using the weeding path, the ground clearance parameter, the first weeding control parameter of the first mode matching area, and the second weeding control parameter corresponding to the second mode matching area.

9. A weeding robot control system suitable for photovoltaic environments, characterized in that, For implementing the weeding robot control method applicable to photovoltaic environments as described in any one of claims 1-8, the system comprises: The scanning component is used to scan the gaps between photovoltaic panels using the onboard multispectral recognition module when the weeding robot autonomously moves to the starting point. Based on the scanned image, it identifies the weed area and generates a weed density map. The data acquisition component is used to synchronously collect obstacle data using the multi-sensor obstacle avoidance module mounted on the weeding robot and construct a three-dimensional obstacle map. An analysis component is used to combine the weed density marker map with the three-dimensional obstacle map to perform weeding path analysis and dual-mode weeding unit weeding parameter analysis, and generate weeding control decisions. The optimization component is used to perform frequency analysis of weeding mode switching based on the weeding control decision, establish unstable switching intervals caused by frequent switching, optimize stable switching of weeding mode, and generate optimized switching parameters. A control component is used to adjust the weeding control decision with the optimized switching parameters and then input it to the control terminal of the weeding robot for weeding control.

Images