Path planning system and method for photovoltaic cleaning robot

Abstract

The invention discloses a path planning system and method for a photovoltaic cleaning robot, and relates to the technical field of path planning. An obstacle distribution acquisition module is used for detecting the distribution condition of obstacles on the surface of a photovoltaic panel in real time and determining a cleaning area; the area classification module is used for classifying the cleaning area into an area containing self-elimination type obstacles and a cleaning area containing fixed obstacles according to the types of the obstacles; the global path searching module is used for determining the distribution position of a cleaning area containing a fixed obstacle and generating a cleaning route with the shortest path, the obstacles on the surface of the photovoltaic panel are cleaned so as to determine the types of the obstacles, and therefore the cleaning robot can quickly lock the area really needing to be cleaned; therefore, invalid walking and repeated operation of the cleaning robot are avoided, the cleaning path length and the task time are greatly shortened, and energy consumption and equipment loss are reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of path planning, and in particular relates to a path planning system and method for a photovoltaic cleaning robot. Background Art

[0002] A fishery-photovoltaic complementary power station refers to a new power generation model that combines fishery farming with photovoltaic power generation. A photovoltaic panel array is set up above the water area, and aquaculture such as fish and shrimp is carried out in the water area below the photovoltaic panels, realizing the comprehensive development and utilization of "generating electricity above and raising fish below".

[0003] In the current field of path planning technology, advance planning and real-time planning are usually used to plan the route of the cleaning robot. However, in the actual process of cleaning path planning for dust or other impurities on photovoltaic panels, due to the variety of dust and impurities falling on photovoltaic panels, if only the two existing planning methods are used, many adverse effects may occur. For example, some obstacles can dissipate on their own. In this case, the cleaning robot does not need to clean them, so in the path planning link, you can choose to avoid these obstacles, thereby avoiding the waste of resources caused by unnecessary cleaning. On the contrary, if you only follow the conventional fixed mode from beginning to end for path planning, it will not only reduce the efficiency of the cleaning work, but also easily cause the cleaning robot to run out of power prematurely, thereby affecting its normal operation. Summary of the Invention

[0004] The purpose of the present invention is to provide a path planning system and method for a photovoltaic cleaning robot, which solves the technical problem in the prior art that the area that really needs to be cleaned cannot be quickly locked, thereby avoiding ineffective walking and repetitive operations of the cleaning robot.

[0005] A path planning system for a photovoltaic cleaning robot, comprising: Obstacle distribution acquisition module, which detects the distribution of obstacles on the surface of photovoltaic panels in real time and determines the clearing area; The area classification module classifies the clearing area into areas containing self-disappearing obstacles and areas containing fixed obstacles according to the type of obstacles; The global path search module determines the distribution of the cleaning area containing fixed obstacles and generates the shortest cleaning route.

[0006] As a further solution of the present invention: also include: S1: For areas containing self-disappearing obstacles, predict the dissipation time of the self-disappearing obstacles in each area. After determining the dissipation time of each obstacle, obtain the distribution position of each obstacle and determine whether the time taken by the cleaning robot to travel to the first target point along the shortest cleaning route is less than a preset threshold; S2: If the predicted self-disappearing obstacle dissipation time is less than the preset threshold, it is marked as a temporary avoidance zone and a waiting countdown is generated; if the predicted self-disappearing obstacle dissipation time is greater than or equal to the preset threshold, the cleaning robot is controlled to move normally to the next target point.

[0007] As a further solution of the present invention: also include: If the self-disappearing obstacle has not dissipated after the countdown ends, the current self-disappearing obstacle is changed to a fixed obstacle and its cleaning area is added to the new path planning.

[0008] As a further solution of the present invention, the marking as a temporary avoidance zone and generating a waiting countdown period further includes: During the waiting countdown, a waiting path is generated and the robot moves along the path at a preset speed until the obstacle disappears.

[0009] As a further solution of the present invention: also include: Obtain the obstacle removal status of the current cleaning area and confirm the remaining cleaning volume; Calculate the difference between the remaining task volume between returning to the previous self-disappearing obstacle dissipation area and heading to the next target point; When the difference in the remaining task amount is greater than or equal to the preset difference, the target point with the larger remaining task amount is selected as the next task point; When the difference in the remaining task amount is less than the preset difference, a target point close to the current position of the cleaning robot is selected as the next task point.

[0010] As a further solution of the present invention: the remaining task amount is calculated as follows: Remaining task volume = area to be cleaned * (1+0.3 * Dirt level), where the dirt level is determined by detecting the dirt coverage on the surface of the photovoltaic panel using a visual sensor.

[0011] As a further solution of the present invention: also include: Obtain the current power and acceleration of the cleaning robot to generate at least two candidate sub-paths; After generating several candidate sub-paths, the comprehensive cost of each sub-path is calculated, and the path with the minimum comprehensive cost is selected as the sub-path.

[0012] As a further solution of the present invention: after generating a plurality of candidate subpaths, the method further includes: Get the number of turns and acceleration direction of each sub-path; Generate a set of candidate paths with different numbers of turns and acceleration directions; Calculate the comprehensive cost of each path, which includes path length, energy consumption and number of turns; Select the path with the lowest comprehensive cost as the execution plan.

[0013] As a further solution of the present invention, it also includes an inter-path collaborative optimization module, specifically: Select the path segment with the least number of turns from the current candidate path as the straight segment, and select the path segment with the best acceleration direction as the energy-saving segment; Perform smooth curve transitions at the intersection of straight segments and energy-saving segments to ensure that the path curvature changes smoothly and does not exceed the set threshold; The total length of the fused path shall not exceed 1.1 times the average length of the original path, and the energy efficiency shall not be less than 90% of the original optimal path.

[0014] On the other hand, the present invention also proposes a path planning method for a photovoltaic cleaning robot, which is applicable to the path planning system of the photovoltaic cleaning robot mentioned above, and the method comprises the following steps: A1: Real-time detection of the distribution of obstacles on the surface of photovoltaic panels and determination of the cleaning area; A2: Based on the type of obstacles, the clearing area is classified into areas containing self-disappearing obstacles and areas containing fixed obstacles. A3: Determine the distribution of the cleanup areas containing fixed obstacles and generate the shortest cleaning route.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention clears obstacles on the surface of photovoltaic panels to determine the type of obstacles, so that the cleaning robot can quickly lock onto the area that really needs to be cleaned, which helps to avoid ineffective walking and repeated operations of the cleaning robot, greatly shortens the length of the cleaning path and task time, reduces energy consumption and equipment loss, and rationally allocates resources, giving priority to stubborn obstacle areas. Compared with traditional methods, it significantly improves cleaning efficiency, reduces labor costs and operation and maintenance expenses, and can also assist in the comprehensive management of fish-light complementary photovoltaic power stations by accumulating data, and further optimizes power generation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of the system framework structure of the present invention. DETAILED DESCRIPTION

[0017] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0018] First aspect: please refer to Figure 1, this application provides a path planning system for a photovoltaic cleaning robot, comprising: Obstacle distribution acquisition module, which detects the distribution of obstacles on the surface of photovoltaic panels in real time and determines the clearing area; The area classification module classifies the clearing area into areas containing self-disappearing obstacles and areas containing fixed obstacles according to the type of obstacles; The global path search module determines the distribution of the cleaning area containing fixed obstacles and generates the shortest cleaning route.

[0019] Among them, by installing wide-angle high-definition cameras at the diagonal ends or edges of each photovoltaic panel, the cameras collect image data on the surface of the photovoltaic panels at a frequency of 5-10 frames per second to ensure that dynamic obstacle changes are captured.

[0020] Multi-line LiDAR (e.g., 16- or 32-line) sensors are deployed on top of or around the photovoltaic array supports. They scan the panels' surfaces with laser beams to obtain three-dimensional spatial information about obstacles. The LiDAR's scanning angle covers no less than 180°, and its detection accuracy can reach the centimeter level.

[0021] Among them, computer vision technology can be used to process images collected by the camera using deep learning target detection algorithms such as YOLO and MaskR-CNN.

[0022] Furthermore, the algorithm is pre-trained with a dataset containing common obstacles such as bird droppings, fallen leaves, dust, and snow, to identify the type, location, and coverage area of ​​obstacles in the image.

[0023] For the point cloud data acquired by the LiDAR, filtering algorithms (such as voxel filtering and straight-through filtering) are used to remove noise points. Then, clustering algorithms (such as DBSCAN and Euclidean clustering) are used to aggregate the point cloud data into different obstacle regions. The spatial distribution and accumulation of obstacles are determined by calculating the volume and height of each cluster region.

[0024] The results of image recognition and point cloud analysis are integrated, and geographic information system (GIS) technology is used to mark all cleanup areas on the electronic map of the photovoltaic panel array, and each area is assigned a unique identification number to facilitate subsequent module calls and processing.

[0025] Obstacles such as dust, light mist, and small amounts of dew that can dissipate or decompose on their own under natural conditions (such as wind, light, and temperature changes) are classified as self-dissipating obstacle areas. The criteria for determining such obstacles are: the thickness of the obstacle does not exceed 0.5 mm, and the area has decreased by more than 30% in the past two hours (determined through historical image comparison and analysis); Obstacles such as bird droppings, fallen leaves, branches, snow, ice, etc. that cannot dissipate naturally and must be removed by physical cleaning means are classified as fixed obstacle areas.

[0026] It should be noted that when dividing the area, try to classify the same obstacles into the same area. If two types of obstacles are classified together due to various external factors, the obstacle area will be marked as a fixed obstacle area. This is because the area containing fixed obstacles can only be cleared by physical means. Even if the proportion is relatively small, physical clearance is required. Therefore, it is divided into a fixed obstacle area to facilitate the generation of a clearing path and reduce the number of backtracking.

[0027] Among them, by adopting machine learning classification algorithms such as support vector machine (SVM) and random forest, the extracted features are used as input and matched with pre-trained classification models to determine the obstacle type. This is an existing mature technology and will not be elaborated on here.

[0028] After obtaining the clearing area of ​​fixed obstacles, it is mapped to a unified coordinate system through a geographic coordinate conversion algorithm (such as WGS84 to UTM).

[0029] The center position of each area is used as a node for path planning, and its precise latitude and longitude coordinates or plane coordinates (X, Y) are recorded. Based on the layout of the photovoltaic panel array, a topological map containing fixed obstacle clearing area nodes and channels between photovoltaic panels is constructed.

[0030] The connecting edges between nodes represent the paths that the cleaning robot can travel. The weights of the edges are assigned according to factors such as path length, slope, and turning difficulty. A path planning algorithm such as the Dijkstra algorithm, A* algorithm, or improved genetic algorithm is used. The starting position (such as the parking point of the cleaning robot) is used as the starting point, and all fixed obstacle cleaning area nodes are used as the end point to search for the cleaning route with the shortest total path length. In this process, the planning of routes that pass through the same area repeatedly is avoided to improve the cleaning efficiency.

[0031] Furthermore, the generated cleaning route is visualized on an electronic map of the photovoltaic array, with arrows and lines marking the robot's direction and path. Route data, including node sequence, path length, and estimated cleaning time, is output as a text file or JSON format and transmitted to the cleaning robot's control system to guide its cleaning task.

[0032] By clearing obstacles on the surface of photovoltaic panels to determine the type of obstacles, the cleaning robot can quickly lock onto the area that really needs cleaning, which helps to avoid ineffective walking and repeated operations of the cleaning robot, greatly shorten the length of the cleaning path and task time, reduce energy consumption and equipment loss, and rationally allocate resources, giving priority to stubborn obstacle areas. Compared with traditional methods, it significantly improves cleaning efficiency, reduces labor costs and operation and maintenance expenses, and can also assist in the comprehensive management of fish-light complementary photovoltaic power stations by accumulating data, further optimizing power generation efficiency.

[0033] As an optional embodiment, the present invention further includes: S1: For areas containing self-disappearing obstacles, predict the dissipation time of the self-disappearing obstacles in each area. After determining the dissipation time of each obstacle, obtain the distribution position of each obstacle and determine whether the time taken by the cleaning robot to travel to the first target point along the shortest cleaning route is less than a preset threshold; S2: If the predicted self-disappearing obstacle dissipation time is less than the preset threshold, it is marked as a temporary avoidance zone and a waiting countdown is generated; if the predicted self-disappearing obstacle dissipation time is greater than or equal to the preset threshold, the cleaning robot is controlled to move normally to the next target point.

[0034] The system uses a network of meteorological sensors deployed at the photovoltaic power station to collect environmental parameters in real time, including but not limited to wind speed and light intensity. Machine learning regression models (such as random forest regression and LSTM neural networks) trained based on historical data are then used to predict the dissipation time of self-disappearing obstacles. When cleaning along the shortest cleaning route and after determining the dissipation time of self-disappearing obstacles, calculate the time required for the robot to travel from its current position to the first target point (fixed obstacle cleaning area) along the shortest route based on the cleaning robot's preset travel speed, path length, and terrain slope correction factor (1.0 for flat areas and 1.2-1.5 for sloped areas). The specific calculation method is not elaborated in detail; Furthermore, the calculated travel time is compared with a preset threshold (e.g., 300 seconds) to determine whether the robot can reach the target point within a reasonable time, while also assessing the potential impact of self-disappearing obstacle areas on the route. If the predicted clearing time for a self-disappearing obstacle zone (e.g., 20 minutes) is less than a preset threshold (30 minutes), and the robot would pass through that area if traveling along its original route, it will be marked as a temporary avoidance zone. A waiting countdown (20 minutes) will also be initiated, pausing path planning for that area. This helps prevent the robot from waiting in vain or repeating cleaning.

[0035] If the predicted clearing time is greater than or equal to a preset threshold (e.g., 35 minutes), or the robot's travel time is significantly less than the preset threshold, the area is determined to have no impact on cleaning efficiency, and the cleaning robot is controlled to proceed normally along the original shortest route to the next target point (fixed obstacle clearing area). For example, if the robot only needs 180 seconds to reach the next target point, while the clearing time for a self-clearing area is 40 minutes, no route adjustment is required.

[0036] By adding a status assessment of self-disappearing obstacles on the basis of the original path planning, it is helpful to avoid the robot wasting time waiting for natural dissipation, or repeated work caused by cleaning too early. The countdown end event of the temporary avoidance zone is re-associated with the path planning. When the countdown returns to zero, the system automatically re-evaluates the status of the area. If the obstacle has dissipated, the area is skipped; if it still exists, it is included in the fixed obstacle clearing area, triggering a new round of path planning, forming a complete closed loop of "detection-prediction-decision-reevaluation", which is conducive to improving work efficiency.

[0037] As an optional embodiment, the present invention further includes: If the self-disappearing obstacle has not dissipated after the countdown ends, the current self-disappearing obstacle is changed to a fixed obstacle and its cleaning area is added to the new path planning.

[0038] Specifically, the obstacle distribution acquisition module uses high-definition visual sensors and laser radar to scan the temporary avoidance zone in real time, comparing the current obstacle coverage area and thickness with the data before the countdown began. If the obstacle area is reduced by less than 50% or the thickness has not changed significantly, it is determined to have not dissipated; The obstacle type in the area is changed from "self-eliminating" to "fixed obstacle", and the corresponding record is updated in the database. The cleaning task for this area is added to the pending task queue and marked as high priority. The global path search module regenerates the shortest cleaning route that includes this area based on the current robot position, the cleaned area, and the newly added fixed obstacle area. For example, if the robot is cleaning another area, the new route will be planned as "current position → newly added fixed obstacle area → originally planned target point", ensuring a seamless cleaning process. The data on the obstacles that did not dissipate (type, environmental conditions, and reasons for non-dissipation) is fed back into the dissipation time prediction model as training samples to optimize the algorithm and avoid similar misjudgments. For example, if dew does not dissipate as predicted in a high-humidity environment, the model will adjust the weighting parameter for humidity on dew dissipation.

[0039] The prediction of dissipation time and the temporary avoidance mechanism can help prevent the cleaning robot from performing ineffective operations on obstacles that can be dissipated naturally, saving equipment operating time and energy consumption. After the countdown, the type of undispersed obstacles can be converted and the path can be replanned to ensure that the cleaning task can still be completed in complex weather conditions (such as continuous high humidity resulting in dew not drying up), which helps avoid cleaning blind spots caused by fluctuations in natural conditions.

[0040] As an optional embodiment, marking the temporary avoidance zone and generating a waiting countdown period further includes: During the waiting countdown, a waiting path is generated and the robot moves along the path at a preset speed until the obstacle disappears.

[0041] It should be understood that fixed obstacle clearing areas or other pending tasks around the temporary avoidance zone (such as within a radius of 50 meters) should be retrieved and included in the waiting path as a priority. The path conflict detection algorithm should be used to ensure that the waiting path does not overlap with the planned cleaning route, which is conducive to preventing the robot from repeated cleaning. If there is no urgent cleaning task in the surrounding area, the robot is planned to move along the edge of the photovoltaic panel array or the main road to maintain the ability to patrol the entire area.

[0042] Furthermore, an improved A* algorithm or Dijkstra algorithm is used to generate a waiting path with the temporary avoidance area as the starting point and combined with the following data; A digital map constructed based on information such as photovoltaic panel array layout, channel location, and obstacle distribution provides a basic framework for path planning. The cleaning priority and progress of the surrounding area are obtained from the pending task queue to ensure that the waiting path prioritizes high-priority tasks. Taking into account the physical limitations of the cleaning robot, such as turning radius and climbing ability, it helps avoid planning unreachable paths. For example, when a photovoltaic panel area is marked as a temporary avoidance zone due to dew, the system detects a fixed obstacle cleaning area with accumulated bird droppings 20 meters to the east. It uses this area as the target point and generates a waiting path containing several intermediate nodes. During the movement process, moving at a lower preset speed can not only reduce energy consumption, but also facilitate rapid switching in response to the disappearance of obstacles at any time. If the obstacles in the temporary avoidance zone disappear ahead of time during the movement, the waiting path will be terminated immediately, the shortest route will be replanned, and the robot will be guided back to the original target point.

[0043] Converting waiting time into effective working time and reducing the idle state of the robot will help improve the overall cleaning efficiency in complex weather scenarios. Through dynamic path planning, the robot's ineffective movement distance will be reduced, battery life will be extended, equipment wear will be reduced, and dynamic monitoring and flexible response to self-eliminating obstacles will be achieved, which will help enhance adaptability in changing environments and ensure that the cleaning coverage rate of photovoltaic panels is maintained at a stable high level.

[0044] As an optional embodiment, the present invention further includes: Obtain the obstacle removal status of the current cleaning area and confirm the remaining cleaning volume; Calculate the difference between the remaining task volume between returning to the previous self-disappearing obstacle dissipation area and heading to the next target point; When the difference in the remaining task amount is greater than or equal to the preset difference, the target point with the larger remaining task amount is selected as the next task point; When the difference in the remaining task amount is less than the preset difference, a target point close to the current position of the cleaning robot is selected as the next task point.

[0045] As an optional embodiment, the remaining task amount is calculated as follows: Remaining task volume = area to be cleaned * (1+0.3 * Dirt level), where the dirt level is determined by detecting the dirt coverage on the surface of the photovoltaic panel using a visual sensor.

[0046] Specifically, the high-definition visual sensors and lidar in the obstacle distribution acquisition module collect data at a dynamic frequency: for areas suspected of being about to dissipate, the collection frequency is increased to 10 seconds per time to ensure timely detection of obstacle changes; for stable areas, the regular frequency of 30 seconds per time is maintained. For example, in areas covered by dew, if the light intensity is detected to be continuously increasing, the data collection frequency is automatically increased; Using deep learning image segmentation technology (such as the improved DeepLabv3+ model), not only can obstacle types and coverage be identified, but also "partially dissipated areas" can be distinguished from "stubborn residual areas." For example, in an area covered with fallen leaves, the wind-blown areas can be segmented and marked from the tightly packed areas. The residual thickness of obstacles can be determined through point cloud density analysis. For example, the snow-covered area can be divided into different layers according to thickness, providing three-dimensional data support for calculating the remaining clearing volume.

[0047] Furthermore, the remaining task volume = area to be cleaned * (1+0.3 * Soil Level) is calculated and recorded in real time: "Self-eliminating obstacle areas that have been cleaned but still have residual obstacles" and "Fixed obstacle areas to be cleaned." For example, after the robot completes the initial cleaning of a fallen leaf area, if it detects that there are still 10% residual areas, the area will automatically be added to the previous area queue. When the difference reaches a preset threshold (e.g., 3), priority is given to target points with a large amount of remaining tasks, and expedited processing mode is initiated. For example, the cleaning robot's operating speed is increased to 0.6 m / s (normal speed is 0.3 m / s) and more powerful cleaning components are used. For example, a high-pressure blowing and scraping combination mode is activated for heavily snowed areas. If the difference is small, the comprehensive distance evaluation method is used: by calculating the straight-line distance between the robot's current position and the target point and combining it with the path complexity coefficient (such as 1.2 for narrow channels and 1 for flat areas), the actual travel distance is obtained and the target point with the shortest actual travel distance is selected, while avoiding areas undergoing maintenance.

[0048] By evaluating the remaining cleanup volume and dynamic task selection, compared with the original mechanism, it is helpful to avoid invalid round-trip paths, thereby improving work efficiency.

[0049] As an optional embodiment, the present invention further includes: Obtain the current power and acceleration of the cleaning robot to generate at least two candidate sub-paths; After generating several candidate sub-paths, the comprehensive cost of each sub-path is calculated, and the path with the minimum comprehensive cost is selected as the sub-path.

[0050] As an optional embodiment, after generating several candidate sub-paths, the following steps are further included: Get the number of turns and acceleration direction of each sub-path; Generate a set of candidate paths with different numbers of turns and acceleration directions; Calculate the comprehensive cost of each path, which includes path length, energy consumption and number of turns; Select the path with the lowest comprehensive cost as the execution plan.

[0051] Specifically, a high-precision battery management system (BMS) is used to update the remaining power every 5 seconds with an error control within ±1%. The battery's state of health (SOH) is also recorded, and the energy consumption weight is automatically increased when the SOH is below 80%. Using a six-axis inertial measurement unit (IMU), the robot's three-axis acceleration and angular velocity data are collected at a frequency of 100Hz. The Kalman filter algorithm is used to eliminate noise and accurately capture dynamic processes such as the robot's start, stop, and turn.

[0052] Furthermore, in the path planning process of the cleaning robot, obtaining the current power and travel acceleration of the cleaning robot is an important basis for generating reasonable candidate sub-paths. The battery status of the robot is directly related to its endurance. If the battery is low, it is necessary to prioritize planning a path that can quickly return to the charging station; if the battery is sufficient, you can choose a path that covers more cleaning areas. Travel acceleration affects the robot's movement efficiency and energy consumption on different paths. Although paths with greater acceleration may be faster, energy consumption will also increase. Based on these factors, the system will generate at least two candidate sub-paths to meet different demand scenarios; After generating several candidate sub-paths, a comprehensive evaluation of each sub-path is required, that is, the comprehensive cost of each sub-path is calculated. The calculation of the comprehensive cost involves multiple dimensions. In addition to considering the length of the path and the energy consumption of the robot on the path (related to power and acceleration), factors such as obstacles and complex terrain that may be encountered on the path are also taken into account. For example, although a path is short, it is full of obstacles, and the robot needs to adjust its direction frequently. This will not only increase the travel time, but also lead to increased energy consumption, and its comprehensive cost may be higher. By accurately calculating the comprehensive cost of each candidate sub-path, the path with the lowest comprehensive cost is finally selected as the sub-path, ensuring that the cleaning robot can complete the cleaning task efficiently and energy-savingly.

[0053] After generating several candidate sub-paths, the system further analyzes the characteristics of each sub-path to determine its number of turns and acceleration direction. These two key parameters are crucial for optimizing path planning. The number of turns directly reflects the complexity of the path. Excessive turns mean the robot needs to frequently adjust its direction, which not only increases the difficulty of motion control but also may lead to reduced cleaning efficiency and increased energy consumption. The acceleration direction affects the smoothness of the robot's movement along the path. Inappropriate switching of acceleration direction can cause unnecessary acceleration and deceleration during the robot's movement, also consuming more power.

[0054] Based on the acquired information about the number of turns and acceleration direction, a set of candidate paths with varying numbers of turns and acceleration directions is generated. By combining and adjusting these parameters, a variety of path solutions can be created to suit different cleaning environments and robot operating states. For example, in large, open spaces, paths with fewer turns and smoother acceleration direction changes can be prioritized. In areas with numerous obstacles and confined spaces, flexible paths that can adapt to complex terrain need to be generated, even though these paths may have a high number of turns.

[0055] Next, each path in the candidate path set is comprehensively evaluated, calculating the comprehensive cost of each path. This comprehensive cost encompasses multiple dimensions. Besides the basic path length, it also prioritizes energy consumption and the number of turns. Path length determines the total distance the robot must travel to complete a cleaning task and is a fundamental factor influencing energy consumption. Energy consumption is closely related to path length, acceleration, and directional changes. Acceleration, deceleration, and frequent changes in direction consume more power. The number of turns is a key indicator of path complexity. Each turn requires speed adjustments and direction corrections, which increase energy consumption and time costs. By establishing a precise mathematical model, these factors are quantified and comprehensively calculated to derive the comprehensive cost of each path. This mathematical model can employ a weighted summation approach, assigning different weights to factors such as path length, energy consumption, and the number of turns. The values ​​of each factor are then multiplied by the corresponding weights and summed to produce the comprehensive cost. Alternatively, a hierarchical analysis method (AHP) model can be used. This is currently available and will not be elaborated upon here.

[0056] Finally, the combined costs of all candidate paths are compared, and the path with the lowest combined cost is selected as the execution plan. This helps ensure that the cleaning robot can operate optimally during actual cleaning tasks, while maintaining cleaning efficiency while minimizing energy consumption and improving overall work effectiveness, providing users with a smarter and more efficient cleaning experience.

[0057] As an optional embodiment, a path-to-path collaborative optimization module is further included, specifically: Select the path segment with the least number of turns from the current candidate path as the straight segment, and select the path segment with the best acceleration direction as the energy-saving segment; Perform smooth curve transitions at the intersection of straight segments and energy-saving segments to ensure that the path curvature changes smoothly and does not exceed the set threshold; The total length of the fused path shall not exceed 1.1 times the average length of the original path, and the energy efficiency shall not be less than 90% of the original optimal path.

[0058] The inter-path collaborative optimization module is a key step in deeply optimizing paths after comprehensive cost evaluation and preliminary screening of candidate paths. It aims to integrate the advantages of different path segments and generate a final execution path that balances efficiency and energy conservation. First, the module performs targeted screening from the current candidate paths. The path segment with the fewest turns is selected as the straight segment. This is because fewer turns means the robot can continue moving at a more stable speed, reducing the efficiency loss caused by frequent directional adjustments. The path segment with the optimal acceleration direction is selected as the energy-saving segment. The optimal here refers to a path segment with smooth acceleration direction changes and fewer acceleration and deceleration operations, which reduces the robot's energy consumption. In this way, the sections of the candidate paths with different advantages are extracted, laying the foundation for subsequent path fusion. After determining the straight and energy-saving segments, a smooth transition is created at their intersection. This step is crucial because drastic changes in curvature during path switching can not only complicate motion control but also increase wear on mechanical components and even create the risk of collision. The system uses a specific algorithm to ensure that path curvature changes are gentle and do not exceed a set threshold, allowing the robot to safely and smoothly navigate path transitions.

[0059] Finally, the fused path undergoes constraint verification. The total length of the fused path must not exceed 1.1 times the average length of the original path, preventing a significant drop in cleaning efficiency due to over-optimization. Furthermore, the energy efficiency must be no less than 90% of the original optimal path, ensuring that the optimized path maintains good energy-saving performance. Only fused paths that meet these two conditions are selected as the final execution path. This module is closely linked to the comprehensive cost calculation model. By quantifying factors such as path length, energy consumption, and the number of turns, the comprehensive cost calculation model selects relatively optimal candidate paths, providing an optimization foundation for the collaborative optimization module. The inter-path collaborative optimization module, on the other hand, further explores the potential of each path segment based on the candidate paths, integrating advantages and verifying constraints to generate paths with better overall performance. The two complement each other: the comprehensive cost calculation model ensures the basic quality of the path, while the collaborative optimization module improves the overall performance of the path, jointly achieving the goal of efficient and energy-saving path planning for the cleaning robot.

[0060] In a second aspect, the present invention further proposes a path planning method for a photovoltaic cleaning robot, the method comprising the following steps: A1: Real-time detection of the distribution of obstacles on the surface of photovoltaic panels and determination of the cleaning area; A2: Based on the type of obstacles, the clearing area is classified into areas containing self-disappearing obstacles and areas containing fixed obstacles. A3: Determine the distribution of the cleanup areas containing fixed obstacles and generate the shortest cleaning route.

[0061] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A path planning system for a photovoltaic cleaning robot, characterized in that: include: Obstacle distribution acquisition module, which detects the distribution of obstacles on the surface of photovoltaic panels in real time and determines the clearing area; The area classification module classifies the clearing area into areas containing self-disappearing obstacles and areas containing fixed obstacles according to the type of obstacles; The global path search module determines the distribution of the cleaning area containing fixed obstacles and generates the shortest cleaning route.

2. A path planning system for a photovoltaic cleaning robot according to claim 1, characterized in that: Also includes: S1: For areas containing self-disappearing obstacles, predict the dissipation time of the self-disappearing obstacles in each area. After determining the dissipation time of each obstacle, obtain the distribution position of each obstacle and determine whether the time taken by the cleaning robot to travel to the first target point along the shortest cleaning route is less than a preset threshold; S2: If the predicted self-disappearing obstacle dissipation time is less than the preset threshold, it is marked as a temporary avoidance zone and a waiting countdown is generated; if the predicted self-disappearing obstacle dissipation time is greater than or equal to the preset threshold, the cleaning robot is controlled to move normally to the next target point.

3. A path planning system for a photovoltaic cleaning robot according to claim 2, characterized in that: Also includes: If the self-disappearing obstacle has not dissipated after the countdown ends, the current self-disappearing obstacle is changed to a fixed obstacle and its cleaning area is added to the new path planning.

4. A path planning system for a photovoltaic cleaning robot according to claim 3, characterized in that: The marking as a temporary avoidance zone and generating a waiting countdown period also include: During the waiting countdown, a waiting path is generated and the robot moves along the path at a preset speed until the obstacle disappears.

5. A path planning system for a photovoltaic cleaning robot according to claim 4, characterized in that: Also includes: Obtain the obstacle removal status of the current cleaning area and confirm the remaining cleaning volume; Calculate the difference between the remaining task volume between returning to the previous self-disappearing obstacle dissipation area and heading to the next target point; When the difference in the remaining task amount is greater than or equal to the preset difference, the target point with the larger remaining task amount is selected as the next task point; When the difference in the remaining task amount is less than the preset difference, a target point close to the current position of the cleaning robot is selected as the next task point.

6. A path planning system for a photovoltaic cleaning robot according to claim 5, characterized in that: The remaining task volume is calculated as follows: Remaining task volume = area to be cleaned * (1+0.3 * Dirt level), where the dirt level is determined by detecting the dirt coverage on the surface of the photovoltaic panel using a visual sensor.

7. A path planning system for a photovoltaic cleaning robot according to claim 1, characterized in that: Also includes: Obtain the current power and acceleration of the cleaning robot to generate at least two candidate sub-paths; After generating several candidate sub-paths, the comprehensive cost of each sub-path is calculated, and the path with the minimum comprehensive cost is selected as the sub-path.

8. The path planning system for a photovoltaic cleaning robot according to claim 1, characterized in that: After generating several candidate subpaths, it also includes: Get the number of turns and acceleration direction of each sub-path; Generate a set of candidate paths with different numbers of turns and acceleration directions; Calculate the comprehensive cost of each path, which includes path length, energy consumption and number of turns; Select the path with the lowest comprehensive cost as the execution plan.

9. The path planning system for a photovoltaic cleaning robot according to claim 1, characterized in that: It also includes inter-path collaborative optimization modules, specifically: Select the path segment with the least number of turns from the current candidate path as the straight segment, and select the path segment with the best acceleration direction as the energy-saving segment; Perform smooth curve transitions at the intersection of straight segments and energy-saving segments to ensure that the path curvature changes smoothly and does not exceed the set threshold; The total length of the fused path shall not exceed 1.1 times the average length of the original path, and the energy efficiency shall not be less than 90% of the original optimal path.

10. A path planning method for a photovoltaic cleaning robot, applicable to a path planning system for a photovoltaic cleaning robot according to any one of claims 1 to 9, characterized in that: The method comprises the following steps: A1: Real-time detection of the distribution of obstacles on the surface of photovoltaic panels and determination of the cleaning area; A2: Based on the type of obstacles, the clearing area is classified into areas containing self-disappearing obstacles and areas containing fixed obstacles. A3: Determine the distribution of the cleanup areas containing fixed obstacles and generate the shortest cleaning route.

Images