Path planning system and method for cleaning robot

Abstract

The invention relates to the technical field of equipment control, and provides a path planning system and method for a cleaning robot, a comprehensive and intelligent path planning control system is constructed, and the path planning control system accurately plans a cleaning path and regulates and controls conveying tension through real-time acquisition and analysis of multi-dimensional data, so that the cleaning efficiency is improved. The cleaning efficiency and quality of the cleaning robot in a complex environment are remarkably improved, the stability and safety of operation of the cleaning robot are enhanced, energy consumption is effectively reduced, the service life of the cleaning robot is prolonged, and compared with a control system of a traditional cleaning robot, the control system has the advantages that the cost is reduced. The adaptability and the automation level of the path planning control system are greatly improved, manual intervention is reduced, and the labor cost is reduced.

Description

Technical Field

[0001] The present application relates to the field of equipment control technology, and in particular to a path planning system and method for a cleaning robot. Background Art

[0002] With the advancement of intelligent manufacturing and industrial automation technologies, path planning control systems have been widely used in logistics warehousing, intelligent robots, and automated production lines. Existing technologies usually obtain environmental information through sensors and combine them with traditional algorithms to implement path planning. Some systems introduce priority division or simple constraints to improve planning efficiency.

[0003] Existing systems often capture only single-dimensional data, lacking analysis of the correlation between environmental information and equipment status, fail to dynamically link path planning with equipment delivery tension, and fail to consider the impact of equipment posture on path curvature. They often use fixed rules to divide work areas. They fail to address the technical challenge of correlating multidimensional data in complex environments to divide areas and coordinate them with paths. Summary of the Invention

[0004] In response to the deficiencies of the prior art, the present application provides a path planning system and method for a cleaning robot.

[0005] In a first aspect, the present application provides a path planning system for a cleaning robot, the system comprising: acquiring strain data and stain distribution images during a conveying process in real time;

[0006] Extract the strain peak value and strain change rate from the strain data, and extract the stain area and stain distribution position from the stain distribution image;

[0007] Determining a conveying tension value of the cleaning robot according to a strain peak value and a strain change rate in the strain data, and generating a tension constraint condition to control the conveying tension of the cleaning robot;

[0008] The cleaning area is divided according to the stain area and stain distribution position in the stain distribution image, and the obstacle information is monitored in real time to optimize the boundary of the cleaning area to obtain the cleaning area. Within the cleaning area, based on the position of the cleaning robot and the tension constraint conditions, the cleaning path of the cleaning robot is planned by the A* algorithm, and the path curvature is synchronously adjusted in combination with the posture of the cleaning robot, and the cleaning path of the cleaning robot is dynamically adjusted.

[0009] As an optional implementation, the logic for generating the cleaning area includes:

[0010] Preliminarily divide the cleaning areas according to the stain area and stain distribution position in the stain distribution image, and simultaneously assign a cleaning priority to each cleaning area according to the stain area and stain distribution position;

[0011] Monitor obstacles in the cleaning environment in real time and superimpose the location and shape of obstacles on the cleaning area;

[0012] The obstacles are expanded by the expansion algorithm to optimize the boundaries of the cleaning area, and the final cleaning area and the cleaning priority of each cleaning area are obtained.

[0013] As an optional implementation, the planning logic of the cleaning path includes:

[0014] According to the position of the cleaning robot, the cleaning area and the tension constraint conditions, the A* algorithm is used to generate a path from the position of the cleaning robot to the cleaning area;

[0015] Filter paths based on the cleaning priority of the cleaning area and combine them to form a global path;

[0016] Based on the posture of the cleaning robot, the global path is locally optimized through spline curve fitting, and the path curvature is adjusted to obtain the cleaning path.

[0017] As an optional implementation manner, the global path formation sub-logic includes:

[0018] When the A* algorithm is used to generate the path from the cleaning robot's position to each cleaning area, the tension constraint condition is refined into a heuristic function with multiple dimensional parameters embedded in the A* algorithm;

[0019] Through multiple rounds of calculation, multiple paths from the cleaning robot's position to the cleaning area are generated, and the conveying tension value of each path is marked;

[0020] The paths are screened first based on the cleaning priority of each cleaning area, and then screened again based on the conveying tension value of each path;

[0021] The screened paths are connected and combined to form a global path, and the continuity and conflict of the global path are detected.

[0022] As an optional implementation, the path curvature adjustment sub-logic includes:

[0023] Acquire the posture of the cleaning robot in real time, perform correlation analysis on the posture of the cleaning robot and the global path, and obtain the correlation analysis results;

[0024] Based on the correlation analysis results, the global path is divided into multiple local path segments. For each local path segment, the boundary conditions of the spline curve fitting are determined in combination with the posture of the cleaning robot at the starting and ending points of the local path segment.

[0025] Each local path segment is fitted with a cubic spline curve, and the path curvature is adjusted by adjusting the control point positions of the cubic spline curve;

[0026] Perform smoothness and feasibility verification on the local path segment after adjusting the path curvature;

[0027] The local path segments that have been adjusted and verified for path curvature are recombined to update the global path.

[0028] As an optional implementation, the adjustment logic of the cleaning path includes:

[0029] Real-time monitoring of strain data, stain distribution images, obstacle information, and the position and posture of the cleaning robot to update the cleaning area, the cleaning robot's conveying tension value, and the cleaning path;

[0030] Determine whether to trigger the adjustment of the cleaning path based on changes in obstacle information in the cleaning area, abnormal changes in the conveying tension value, and deviation changes in the cleaning path;

[0031] When the cleaning path adjustment is triggered by abnormal changes in the conveying tension value, the avoidance mechanism is executed;

[0032] When the cleaning path adjustment is triggered by changes in obstacle information in the cleaning area and deviations of the cleaning path, local replanning is performed to obtain a local path, which is then merged with the original global path to update the global path.

[0033] As an optional implementation, the avoidance mechanism includes:

[0034] When the conveying tension is high as determined by comparing the conveying tension threshold, the avoidance area is determined based on the current position of the cleaning robot and the degree and direction of the abnormal conveying tension.

[0035] Outside the avoidance area, the A* algorithm is used to search for multiple candidate avoidance paths based on the updated cleaning area, and the feasibility of the candidate avoidance paths is verified;

[0036] Evaluate each candidate avoidance path based on cleaning efficiency, tension stability, and path safety, and determine the avoidance path;

[0037] The avoidance path is smoothed by spline curve fitting. When the cleaning robot completes the avoidance according to the avoidance path, the cleaning path is replanned according to the position and posture of the cleaning robot.

[0038] As an optional implementation, the generation logic of the tension constraint condition includes:

[0039] Determine the conveying tension threshold of the cleaning robot based on the mechanical safety of the cleaning robot;

[0040] Dynamically adjust the conveying tension threshold of the cleaning robot according to the stain area and stain distribution location;

[0041] The conveying tension threshold of the cleaning robot is corrected based on the energy consumption of the cleaning robot to generate a tension constraint condition.

[0042] As an optional implementation, the logic for acquiring the stain distribution image includes:

[0043] Combining a multispectral imaging camera and a visible light camera to obtain a camera group to capture the cleaning environment;

[0044] Deploy the lighting array, adjust the light source parameters of the lighting array, and use the camera group to obtain multi-angle images;

[0045] Multi-angle images are stitched and corrected through image stitching and distortion correction to generate a stain distribution image.

[0046] As an optional implementation manner, the extraction logic of the stain area and stain distribution position includes:

[0047] Performing image segmentation on the stain distribution image to determine the stain area, and performing morphological processing on the stain area to obtain a stain area image;

[0048] Counting the number of pixels in the stained area in the stained area image by a pixel statistics method, and converting the number of pixels in the stained area into the stained area of ​​the stained area;

[0049] The stained regions in the stained region image are marked by a region marking algorithm, an identifier is assigned to each stained region, and the centroid coordinates of each stained region are calculated to determine the stain distribution position of the stained region.

[0050] In a second aspect, the present application provides a path planning method for a cleaning robot, the method comprising: acquiring strain data and stain distribution images during a conveying process in real time;

[0051] Extract the strain peak value and strain change rate from the strain data, and extract the stain area and stain distribution position from the stain distribution image;

[0052] Determining a conveying tension value of the cleaning robot according to a strain peak value and a strain change rate in the strain data, and generating a tension constraint condition to control the conveying tension of the cleaning robot;

[0053] Divide the cleaning area according to the stain area and the stain distribution position in the stain distribution image;

[0054] Monitor obstacle information in real time to optimize the boundaries of the cleaning area and obtain the cleaning area;

[0055] Based on the position of the cleaning robot and the tension constraint conditions in the cleaning area, the cleaning robot's cleaning path is planned using the A* algorithm;

[0056] Synchronously adjust the path curvature in conjunction with the cleaning robot’s posture;

[0057] Dynamically adjust the cleaning path of the cleaning robot.

[0058] Compared with the existing technology, the beneficial effects of the present application are: by constructing a comprehensive and intelligent path planning control system, the path planning control system accurately plans the cleaning path and regulates the conveying tension through real-time acquisition and analysis of multi-dimensional data, which not only significantly improves the cleaning efficiency and quality of the cleaning robot in complex environments, but also enhances the stability and safety of the cleaning robot's operation, effectively reduces energy consumption, and extends the service life of the cleaning robot. Compared with the control system of traditional cleaning robots, the present application greatly improves the adaptability and automation level of the path planning control system, reduces manual intervention, and reduces labor costs.

[0059] Real-time acquisition of strain data and stain distribution images during the conveying process. The strain data can reflect the stress conditions of the cleaning robot during the conveying process in real time, provide early warning of potential mechanical failures, and ensure the safe operation of the equipment. The stain distribution images provide an intuitive basis for formulating targeted cleaning strategies, ensuring the accuracy of cleaning and avoiding ineffective cleaning.

[0060] The strain peak value and strain change rate are extracted from the strain data, and the stain area and stain distribution position are extracted from the stain distribution image. The extraction of the strain peak value and strain change rate helps to deeply analyze the force change trend during the transportation process of the cleaning robot and provide key parameters for tension adjustment. The extraction of the stain area and stain distribution position can accurately identify the key cleaning areas. Combined with the setting of cleaning priority, the cleaning operation can be more targeted, the cleaning efficiency can be improved, and resource waste can be avoided.

[0061] The conveying tension value of the cleaning robot is determined based on the strain peak value and strain change rate in the strain data, and tension constraints are generated to control the conveying tension of the cleaning robot. This not only ensures that the cleaning robot operates under appropriate tension, avoiding the impact of excessive or insufficient tension on the cleaning effect and equipment life, but also reduces equipment loss and improves operational stability.

[0062] The cleaning area is divided according to the stain area and stain distribution position in the stain distribution image, and the obstacle information is monitored in real time to optimize the boundary of the cleaning area to obtain the cleaning area, ensuring that the cleaning area planning not only covers all areas to be cleaned, but also can effectively avoid obstacles, avoid collisions between the cleaning robot and obstacles, and ensure the safety of equipment and personnel; within the cleaning area, based on the position of the cleaning robot and the tension constraint conditions, the cleaning path of the cleaning robot is planned by the A* algorithm, and the path curvature is adjusted synchronously with the posture of the cleaning robot, and the cleaning path of the cleaning robot is dynamically adjusted to ensure that the cleaning robot can adapt to complex and changeable cleaning environments and efficiently complete cleaning tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be derived from these drawings without inventive effort. Among them:

[0064] Figure 1 A system flow chart of a path planning system for a cleaning robot provided in an embodiment of the present application;

[0065] Figure 2 A global path formation logic diagram of a path planning system for a cleaning robot provided in an embodiment of the present application;

[0066] Figure 3 A cleaning path adjustment logic diagram of a path planning system for a cleaning robot provided in an embodiment of the present application;

[0067] Figure 4 A flow chart of a path planning method for a cleaning robot provided in an embodiment of the present application. DETAILED DESCRIPTION

[0068] In order to make the objectives, technical solutions and advantages of the embodiments of the present application more obvious and easy to understand, the technical solutions in the embodiments of the present application are clearly and completely described below in conjunction with the drawings in the specification. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments.

[0069] Example 1

[0070] like Figure 1 As shown, a system flow chart of a path planning system for a cleaning robot is provided for an embodiment of the present application. The system includes a data acquisition module, a data processing module, a tension adjustment module and a path planning module.

[0071] In this embodiment, the cleaning environment of the cleaning robot is taken as an air cooling island scenario.

[0072] The data acquisition module is used to obtain strain data and stain distribution images during the conveying process in real time.

[0073] Based on the concept of differentiated layout, in addition to arranging strain gauges at conventional stress points such as the conveying arms and connecting parts of the cleaning robot, fiber Bragg grating strain sensors are additionally deployed at the joints of special-shaped structures where stress concentration is prone to occur, as well as at locations where stress changes dynamically due to cleaning operations. The strain gauges can quickly respond to local stress changes, and the fiber Bragg grating strain sensors can perform distributed measurement of tiny strains, thereby acquiring a wide range of continuous strain signals. The strain signals of the strain gauges and fiber Bragg grating strain sensors are then acquired synchronously at a fixed sampling frequency to reduce data errors caused by acquisition time differences. To remove electromagnetic interference during the acquisition process, a low-pass filter is used to suppress high-frequency noise. At the same time, with the help of a gain amplifier, the amplification factor is dynamically adjusted according to the amplitude of the strain signal to ensure that weak strain signals can be accurately captured and to avoid overload of strong signals. The strain signals are converted into strain data and transmitted to the data processing module.

[0074] Specifically, the acquisition logic of the stain distribution image includes:

[0075] Combining a multispectral imaging camera and a visible light camera to obtain a camera group to capture the cleaning environment;

[0076] Deploy the lighting array, adjust the light source parameters of the lighting array, and use the camera group to obtain multi-angle images;

[0077] Multi-angle images are stitched and corrected through image stitching and distortion correction to generate a stain distribution image.

[0078] In the air-cooled island cleaning scenario, different types of stains have different reflection and absorption characteristics in the multispectral and visible light bands. It is difficult for a single camera to obtain comprehensive and accurate stain information. Therefore, a multispectral imaging camera and a visible light camera are combined into a camera group. The multispectral imaging camera can capture light in a specific band and effectively identify the type and composition of stains by analyzing the spectral information of different bands, such as distinguishing between oil, dust or chemical deposits. The visible light camera provides high-resolution image details, clearly showing the shape, size and distribution of the stains. The two complement each other, so that the captured images can more comprehensively reflect the stain conditions in the cleaning environment, laying the foundation for the subsequent accurate extraction of stain information.

[0079] The air-cooling island is large, and the lighting conditions in different areas vary significantly, resulting in uneven lighting and shadows. This can seriously affect the quality of the images taken by the camera group, resulting in loss or misjudgment of stain information. To solve this problem, a lighting array is deployed. The lighting array consists of multiple independently adjustable light sources. The brightness, color and illumination angle of the light source are dynamically adjusted according to the light intensity of the cleaning environment and the surface characteristics of the object. The brightness of the light source is increased in darker areas, and the angle of the light source is adjusted to reduce reflections on surfaces with strong reflectivity. This not only ensures that the camera group can clearly capture multi-angle images under various complex lighting conditions, but also improves the consistency and stability of the image, facilitating subsequent image stitching and processing.

[0080] Due to the complex structure of the air-cooling island equipment, there are many obstructions and blind spots. Single-angle shooting cannot obtain complete stain distribution information. Therefore, a camera group is used to shoot the cleaning environment from multiple angles. The camera is installed on a rotatable bracket to shoot at different positions and angles to ensure that all parts of the cleaning environment can be photographed. This greatly increases the amount of information in the image, reduces the information loss caused by obstruction, and provides sufficient data for generating a complete stain distribution image.

[0081] There are overlapping parts in images taken at different angles, but due to the different positions and angles of the cameras, there are geometric distortions and position deviations between the images, and they cannot be directly stitched together. Through the image stitching algorithm based on feature matching, the feature points in the image, such as corners and edges, are first extracted. By matching the feature points in different images, the transformation relationship between the images is determined, and multiple images are stitched into a panoramic image. At the same time, the intrinsic and extrinsic parameters of the camera are obtained through camera calibration technology. Based on these parameters, the image distortion is corrected to eliminate the barrel or pincushion distortion caused by the camera lens, and generate an accurate and distortion-free stain distribution image. This enables the stitched stain distribution image to truly reflect the actual situation of the cleaning environment, and provides accurate data for the subsequent extraction of stain area and stain distribution position.

[0082] The strain data and stain distribution images come from different sensors, and their coordinate systems are inconsistent, so they cannot be directly correlated and analyzed. In order to achieve consistency in the spatiotemporal mapping of the two, a spatiotemporal calibration matrix is ​​established. By setting multiple calibration objects with known positions in the cleaning environment, the calibration object images are captured by the camera, and the strain data of the calibration objects when they are subjected to force is measured by the strain sensor. According to the principles of three-dimensional reconstruction and coordinate transformation, the conversion relationship between the coordinate system of the strain gauge and fiber Bragg grating strain sensor and the coordinate system of the multispectral imaging camera and visible light camera is calculated, thereby obtaining the spatiotemporal calibration matrix. This enables the strain data and stain distribution images to be analyzed in the same spatiotemporal framework, providing a basis for the subsequent comprehensive use of multi-source data to control the cleaning robot.

[0083] In actual applications, sensors may be affected by factors such as environmental interference and hardware failure, resulting in unreliable data. To ensure the accuracy and reliability of the acquired data, the data credibility of strain gauges, fiber Bragg grating strain sensors, multispectral imaging cameras and visible light cameras is monitored in real time. Through data range verification and data change trend verification, it is judged whether the sensor data is abnormal. When the sensor data is detected to be abnormal, the sensor self-test program is triggered. Through hardware self-test and software algorithm diagnosis, the fault type and location are determined, and corresponding repairs are performed. This ensures that the sensor is always in normal working condition, provides accurate data for the conveying control system, and ensures the stable operation of the entire conveying control system.

[0084] The data processing module is used to extract the strain peak value and the strain change rate from the strain data, and to extract the stain area and the stain distribution position from the stain distribution image.

[0085] The acquired strain data is denoised by wavelet transform. Wavelet transform can analyze strain signals at different scales, decompose strain data into different frequency sub-bands, remove high-frequency sub-bands where noise is located through threshold processing, and effectively retain useful information of the strain signal. At the same time, the features of the denoised strain data are enhanced by empirical mode decomposition. Empirical mode decomposition can decompose complex strain signals into a series of intrinsic mode functions. Each intrinsic mode function characterizes the characteristics of the strain signal at different time scales. By analyzing and screening the intrinsic mode functions, characteristic information related to strain peak value and strain change rate is highlighted.

[0086] Based on the strain data after noise reduction and feature enhancement, a dynamic threshold is determined according to the mean and standard deviation of the strain data to adapt to the amplitude changes of the strain data under different working conditions. Then, the sequence of strain data is traversed, and data points greater than the dynamic threshold are marked as potential peak points. In order to eliminate pseudo peaks caused by noise or local fluctuations, the potential peak points are compared in the neighborhood. Only the data point with the largest amplitude within a certain neighborhood range is confirmed as the strain peak point, and the data value corresponding to the strain peak point is the strain peak value.

[0087] In order to accurately calculate the strain change rate, a moving window of fixed length is set on the strain data sequence by combining the central difference method with the moving window technology. For each data point in the moving window, the strain change rate is calculated by the central difference formula. In order to make the calculation results smoother, the calculated strain change rate sequence is subjected to weighted moving average filtering to remove high-frequency noise caused by calculation errors or data fluctuations, so as to obtain a strain change rate that accurately reflects the strain change trend.

[0088] When the strain change rate is greater than the set change rate threshold, the fixed sampling frequency is switched to a high-frequency sampling frequency to balance data accuracy and power consumption. When the strain change rate is less than or equal to the set change rate threshold, the fixed sampling frequency is switched to a low-frequency sampling frequency to reduce power consumption.

[0089] Specifically, the extraction logic of stain area and stain distribution location includes:

[0090] Performing image segmentation on the stain distribution image to determine the stain area, and performing morphological processing on the stain area to obtain a stain area image;

[0091] Counting the number of pixels in the stained area in the stained area image by a pixel statistics method, and converting the number of pixels in the stained area into the stained area of ​​the stained area;

[0092] The stained regions in the stained region image are marked by a region marking algorithm, an identifier is assigned to each stained region, and the centroid coordinates of each stained region are calculated to determine the stain distribution position of the stained region.

[0093] In the air-cooling island cleaning scenario, the stain distribution image contains a large amount of background information. In order to extract stain-related information, the first task is to separate the stain area from the background. Image segmentation is performed using the U-Net model based on deep learning. The U-Net model has an encoder-decoder structure. The encoder part gradually extracts the features of the stain distribution image through convolution operations, and the decoder part restores the resolution of the stain distribution image through upsampling operations and fuses the features of the encoder part to achieve accurate segmentation of the stain area. When training the U-Net model, a large number of stain images in the air-cooling island cleaning scenario are collected and labeled to distinguish the stain area from the background area. This is used to train the U-Net model to enable it to accurately identify stains. Through this U-Net model, the stain area in the stain distribution image can be accurately segmented, providing a basis for subsequent further analysis of the stain area. The accuracy of the segmented stain area directly affects the accuracy of subsequent calculations of the stain area and stain distribution position.

[0094] The stain area image obtained after image segmentation still has noise points, holes or discontinuous parts, which will affect the subsequent accurate calculation of the stain area and stain distribution position. Therefore, the segmented stain area image is subjected to morphological processing. Through the two basic morphological operations of corrosion and expansion, the corrosion operation can remove isolated noise points in the stain distribution image, and the expansion operation can fill small holes in the stain distribution image, making the stain area more continuous and complete. By performing the corrosion operation first and then the expansion operation, the stain area image can be effectively optimized. This step improves the quality of the stain area image, removes noise and defects that will interfere with the calculation, and makes the subsequent calculation of the stain area and stain distribution position based on the stain area image more accurate and reliable, providing high-quality data for calculating the accurate stain area and stain distribution position.

[0095] In order to calculate the stain area, we first need to count the number of pixels in the stain area in the stain area image. For each pixel point, determine whether it belongs to the stain area. If it belongs to the stain area, the counter is incremented by one to obtain the number of pixels in the stain area. This method is simple and direct, and can quickly and accurately count the number of pixels in the stain area, providing data support for the subsequent conversion of the number of pixels into the actual stain area. The accuracy of the statistical results directly determines the accuracy of the final calculated stain area. Since the image has a certain resolution, each pixel corresponds to a certain area in the actual scene. By pre-acquiring the resolution information of the stain area image and the ratio between the stain area image and the actual scene, the counted number of pixels in the stain area is converted into the actual stain area. For example, if the image resolution is 300 pixels per inch and the ratio of the image to the actual scene is 1:10, then the area corresponding to each pixel in the actual scene is (1 / 300×1 / 10). 2 Square inch, multiply the number of pixels in the stain area by the actual area corresponding to each pixel to get the actual stain area. This step realizes the conversion from image pixel information to actual stain area, provides intuitive stain area data for the conveying control system, and provides a key basis for subsequent operations such as tension adjustment and path planning based on the stain area.

[0096] In order to determine the distribution position of each stain area, it is necessary to mark the stain area in the stain area image. The region marking algorithm based on connected domain analysis is used. This algorithm scans the pixel matrix of the stain area image and divides the adjacent pixels belonging to the stain area into the same connected domain according to the connectivity between the pixels. A unique identifier is assigned to each connected domain. In this way, each stain area is given a unique identifier, which facilitates the subsequent separate analysis of each stain area and provides the prerequisite for calculating the centroid coordinates of each stain area, so that the positions of different stain areas can be distinguished and located.

[0097] For each marked stain area, its centroid coordinates are calculated to determine the stain distribution location. The calculation method of the centroid coordinates is based on the coordinate information of the pixel points in the stain area, and is obtained by weighted averaging the horizontal and vertical coordinates of all pixels in the stain area. The calculated centroid coordinates can accurately determine the center position of each stain area in the stain area image, thereby achieving accurate determination of the stain distribution location. This result provides the path planning module with stain location information, helping the cleaning robot to accurately plan the cleaning path and improve cleaning efficiency and effectiveness.

[0098] The tension adjustment module is used to determine the conveying tension value of the cleaning robot according to the strain peak value and strain change rate in the strain data, and generate tension constraint conditions to control the conveying tension of the cleaning robot.

[0099] Based on the strain data obtained above, combined with the mechanical structure parameters of the cleaning robot, the material properties of the cleaning object and the cleaning process requirements, a tension prediction model is constructed using the random forest algorithm in machine learning. During the training stage of the tension prediction model, the strain peak value, strain change rate, operating speed and load weight of the cleaning robot are used as input features, and the actual required conveying tension value is used as the output label. By learning from a large amount of data, the tension prediction model can accurately capture the complex relationship between each input feature and the conveying tension.

[0100] During the operation of the cleaning robot, strain data and other relevant working condition parameters are obtained in real time, including the operating speed and load weight of the cleaning robot. Based on these real-time data, the current cleaning working conditions are evaluated to determine whether they are within the typical working conditions covered during the training of the tension prediction model. If there is a large difference between the current working conditions and the training working conditions, the constructed tension prediction model is corrected in real time through the online learning algorithm. For example, through incremental learning, the newly acquired valid data is gradually incorporated into the tension prediction model training, so that the tension prediction model can adapt to different cleaning environment scenarios and improve the effectiveness of tension prediction.

[0101] Based on the revised tension prediction model, the conveying tension value required by the cleaning robot under the current working conditions is calculated. In order to ensure the smoothness and reliability of the tension adjustment, the calculated conveying tension value is optimized. Through the dynamic programming algorithm, an optimal tension adjustment trajectory is found under the premise of meeting the cleaning task requirements and the safety restrictions of the cleaning robot to avoid the impact of sudden changes in the conveying tension value on the cleaning robot, thereby achieving precise control of the conveying tension of the cleaning robot.

[0102] Specifically, the generation logic of tension constraints includes:

[0103] Determine the conveying tension threshold of the cleaning robot based on the mechanical safety of the cleaning robot;

[0104] Dynamically adjust the conveying tension threshold of the cleaning robot according to the stain area and stain distribution location;

[0105] The conveying tension threshold of the cleaning robot is corrected based on the energy consumption of the cleaning robot to generate a tension constraint condition.

[0106] The mechanical structure of the cleaning robot has a bearing limit. If the conveying tension exceeds this limit, it will cause damage to mechanical components, such as broken conveying arms and loose connection components, which will seriously affect the normal operation and service life of the cleaning robot. Therefore, in order to ensure the mechanical safety of the cleaning robot, it is necessary to determine a basic conveying tension threshold; by conducting a mechanical analysis of the cleaning robot and combining the design drawings of its mechanical structure, the material properties, size specifications and connection methods of each key component are determined, and the finite element analysis software is used to simulate the stress conditions of the cleaning robot under different working conditions, and the maximum tension that each component can withstand is calculated. Taking into account the bearing capacity of each component, the minimum and maximum values ​​are taken as the conveying tension threshold of the robot as a whole; thereby setting a safety boundary for the tension constraint condition, effectively avoiding mechanical damage to the cleaning robot due to excessive tension, and ensuring the stable operation of the cleaning robot. At the same time, the conveying tension threshold provides a benchmark for the subsequent dynamic adjustment of the conveying tension threshold according to other factors, and all subsequent adjustments are made within this safe range.

[0107] The stain area and stain distribution location in different areas are different. When the cleaning robot cleans these areas, the required cleaning force and conveying tension should also be different. If a fixed conveying tension threshold is used, it will lead to poor cleaning effect or excessive energy consumption. Therefore, after the data processing module extracts the information of the stain area and stain distribution location, a mapping relationship model between the stain and the required tension is established, and the stain area is divided into multiple levels, such as small area stains, medium area stains, and large area stains. The division of small area stains, medium area stains and large area stains is obtained based on the comparison with the area threshold, and the area threshold includes two area thresholds: minimum area and maximum area. For different levels of stains, combined with the cleaning process requirements, the corresponding optimal conveying tension is determined through experimental calculation. When stains are detected in a cleaning area, the corresponding tension adjustment value is obtained from the mapping relationship model according to the size of the stain area and the distribution position of the stain, and the conveying tension threshold is dynamically adjusted. For example, experiments have found that when cleaning a small area of ​​dense stains, appropriately increasing the conveying tension by 10% can achieve a better cleaning effect, while when cleaning a large area of ​​sparse stains, reducing the conveying tension by 5% can both ensure the cleaning effect and save energy; thus, the cleaning operation of the cleaning robot is more targeted, the cleaning effect is significantly improved, and the problem of incomplete cleaning or energy waste caused by improper conveying tension is avoided. At the same time, this adjustment result provides intermediate data for the conveying tension threshold based on energy consumption correction, so that the energy consumption correction can further optimize the tension constraint conditions on this basis.

[0108] During the operation of the cleaning robot, energy consumption is an important consideration. Excessive conveying tension will increase the energy consumption of the cleaning robot and reduce its operating efficiency. Therefore, during the operation of the cleaning robot, energy consumption monitoring sensors installed in key parts such as the motor and conveying device are used to obtain the energy consumption data of the cleaning robot in real time, analyze the energy consumption change law of the cleaning robot under different conveying tension values, and set an energy consumption optimization target. When the real-time monitored energy consumption is greater than the energy consumption value corresponding to the optimization target, the conveying tension value that needs to be reduced is determined, and the conveying tension threshold adjusted after the first two steps is corrected. At the same time, during the correction process, it is necessary to ensure that the adjusted conveying tension can still meet the cleaning effect and mechanical safety requirements, and obtain the final tension constraint condition; thereby effectively reducing the energy consumption of the cleaning robot and improving its operating economy. The optimized tension constraint condition provides a more reasonable tension limit for the path planning module, so that the path planning can fully consider the tension constraint and energy consumption requirements while meeting the cleaning task, further improving the performance of the entire conveying control system.

[0109] The path planning module is used to divide the cleaning area according to the stain area and stain distribution position in the stain distribution image, and monitor the obstacle information in real time to optimize the boundary of the cleaning area to obtain the cleaning area. Within the cleaning area, based on the position of the cleaning robot and the tension constraint conditions, the cleaning path of the cleaning robot is planned through the A* algorithm, and the path curvature is adjusted synchronously with the posture of the cleaning robot, and the cleaning path of the cleaning robot is dynamically adjusted.

[0110] Specifically, the generation logic of the cleaning area includes:

[0111] Preliminarily divide the cleaning areas according to the stain area and stain distribution position in the stain distribution image, and simultaneously assign a cleaning priority to each cleaning area according to the stain area and stain distribution position;

[0112] Monitor obstacles in the cleaning environment in real time and superimpose the location and shape of obstacles on the cleaning area;

[0113] The obstacles are expanded by the expansion algorithm to optimize the boundaries of the cleaning area, and the final cleaning area and the cleaning priority of each cleaning area are obtained.

[0114] The air-cooling island equipment has a large surface area, and the stains are scattered and unevenly distributed. To improve cleaning efficiency, it is necessary to reasonably divide the cleaning area according to the actual distribution of stains and assign a cleaning priority to each cleaning area so that the cleaning robot can prioritize areas with severe stains. The stain area and stain distribution location information of the stain distribution image are obtained from the data processing module. The DBSCAN algorithm is used to cluster the stains according to the spatial density of the stain points, and similar stains are divided into the same area. The cleaning priority is set for each cleaning area according to the size and concentration of the stains. The stain areas with large and concentrated stains have a higher cleaning priority. This achieves preliminary structured processing of the cleaning area and clarifies the focus and order of cleaning. This provides a basic framework for the subsequent superposition of obstacle information and optimization of the cleaning area boundaries, making the planning of the entire cleaning area more targeted.

[0115] There are a large number of obstacles such as pipes and brackets on the air-cooling island site. These obstacles will limit the movement range of the cleaning robot. Therefore, obstacle information needs to be integrated into the planning of the cleaning area to prevent the robot from colliding with obstacles. Through the use of equipment such as lidar and ultrasonic sensors, the obstacle information in the cleaning environment is monitored in real time to obtain the location and shape data of the obstacles. This data is superimposed on the preliminary divided cleaning area in the form of polygons or grids, marking the areas that the cleaning robot cannot reach. This ensures the safety and feasibility of the cleaning area planning, provides accurate obstacle information for the subsequent optimization of the cleaning area boundary through the expansion algorithm, and avoids the subsequent planned path from colliding with obstacles.

[0116] Taking into account the size and motion characteristics of the cleaning robot, directly planning the cleaning area according to the actual boundary of the obstacle will cause the robot to be unable to operate normally due to lack of space when approaching the obstacle. By processing the obstacles through the expansion algorithm, sufficient safety space can be reserved; through the morphological expansion algorithm, the expansion radius is determined according to the external dimensions of the cleaning robot and the safety margin required for movement, and the boundary of the obstacle is expanded so that the boundary of the cleaning area avoids the obstacle by a certain distance. At the same time, the boundary of the cleaning area is smoothed to ensure that the cleaning robot can move smoothly along the boundary; the cleaning area boundary thus obtained can not only ensure the safe operation of the cleaning robot, but also improve the cleaning coverage rate, which provides accurate cleaning area information for the subsequent cleaning path planning and reduces the risk of collision.

[0117] Specifically, the planning logic of the cleaning path includes:

[0118] According to the position of the cleaning robot, the cleaning area and the tension constraint conditions, the A* algorithm is used to generate a path from the position of the cleaning robot to the cleaning area;

[0119] Filter paths based on the cleaning priority of the cleaning area and combine them to form a global path;

[0120] Based on the posture of the cleaning robot, the global path is locally optimized through spline curve fitting, and the path curvature is adjusted to obtain the cleaning path.

[0121] The A* algorithm is a commonly used path search algorithm that can find the optimal path from the starting point to the end point in a given map. Combined with the position of the cleaning robot, the cleaning area, and the tension constraints, the A* algorithm can quickly generate local paths to each cleaning area. Taking the current position of the cleaning robot as the starting point and the center or entrance point of the cleaning area as the end point, a path search map is constructed, and the tension constraints are refined into multiple dimensional parameters, such as the maximum allowable conveying tension and tension change rate on the path, which are embedded in the heuristic function of the A* algorithm. During the search process, factors such as the length of the path, the conveying tension when passing through the path, and whether obstacles are avoided are considered. Multiple paths from the cleaning robot position to the cleaning area are generated, and the conveying tension value of each path is marked. Through this improved A* algorithm, the generated local paths not only take into account the shortest path, but also fully consider the tension constraints and obstacle avoidance, providing a variety of options for subsequent screening and combination of paths based on cleaning priority.

[0122] Since cleaning areas have different priorities, in order to improve cleaning efficiency, it is necessary to give priority to cleaning areas with high priorities, and at the same time, reasonably combine the local paths to form a coherent global path; according to the cleaning priority of each cleaning area, the paths generated by the A* algorithm are screened for the first round, and the paths leading to high-priority cleaning areas are given priority. When the priorities are the same, a secondary screening is performed based on the conveying tension value of each path, and the paths with more stable conveying tension and more in line with the working requirements of the cleaning robot are selected. The screened paths are connected and combined in the order of cleaning priority to form a global path; through this screening and combination method, it is ensured that the cleaning robot cleans in a reasonable order, which improves cleaning efficiency. At the same time, the obtained global path has good performance in terms of tension stability and path rationality, which provides a good foundation for subsequent local optimization based on the robot's posture.

[0123] During the movement of the cleaning robot, its posture will affect the cleaning effect and the feasibility of the path. Local optimization of the global path based on the posture of the cleaning robot can enable the cleaning robot to clean with a more appropriate posture and improve the cleaning quality. The posture information of the cleaning robot is obtained in real time, including position, direction and tilt angle, and the posture of the cleaning robot is correlated with the global path and analyzed. The global path is divided into multiple local path segments. For each local path segment, the boundary conditions of the spline curve fitting are determined based on the posture of the robot at the starting and end points of the segment. Each local path segment is fitted with the help of a cubic spline curve. By adjusting the control point position of the cubic spline curve, the path curvature is optimized so that the cleaning robot can maintain a suitable posture when passing through the path segment. The curvature of the cleaning path is thereby optimized, making the movement of the cleaning robot more stable and improving the cleaning quality. At the same time, the optimized path provides a benchmark for the dynamic adjustment of the subsequent path curvature, which is convenient for further optimization according to actual conditions.

[0124] Further, if Figure 2 As shown, the formation logic of the global path includes:

[0125] When the A* algorithm is used to generate the path from the cleaning robot's position to each cleaning area, the tension constraint condition is refined into a heuristic function with multiple dimensional parameters embedded in the A* algorithm;

[0126] Through multiple rounds of calculation, multiple paths from the cleaning robot's position to the cleaning area are generated, and the conveying tension value of each path is marked;

[0127] The paths are screened first based on the cleaning priority of each cleaning area, and then screened again based on the conveying tension value of each path;

[0128] The screened paths are connected and combined to form a global path, and the continuity and conflict of the global path are detected.

[0129] During the operation of the cleaning robot, the size of the conveying tension will affect the robot's stability and cleaning effect. Refining the tension constraint conditions into multiple dimensional parameters and embedding them into the heuristic function of the A* algorithm can guide the A* algorithm to generate a path that better meets the conveying tension requirements; analyzing the mechanical structure and cleaning process requirements of the cleaning robot, determining the parameters related to conveying tension, such as the robot's movement speed, load weight, and path curvature, and their influence on tension, converting these parameters into a mathematical model and integrating them into the heuristic function of the A* algorithm. When calculating the cost of the path, the path length, tension consumption and other related factors are comprehensively considered, so that the A* algorithm can take tension constraints into account when searching for the path; in this way, the generated path can better meet the tension constraint conditions while meeting the cleaning task requirements, providing candidate paths that meet the tension requirements for subsequent screening and combination of paths.

[0130] In order to obtain more path options and improve the flexibility and adaptability of path planning, multiple paths from the robot position to the cleaning area are generated through multiple rounds of calculations, and the conveying tension value of each path is marked for subsequent screening; each time the A* algorithm is run, some search parameters, such as the weight of the heuristic function and the search range, are randomly adjusted to generate different paths. The conveying tension is calculated for each generated path, and its conveying tension value is marked. Through multiple rounds of calculations, a set of diverse path sets is obtained; this increases the diversity of the paths, provides rich materials for subsequent screening based on cleaning priority and conveying tension value, and helps to select the optimal path combination.

[0131] According to the cleaning priority of the cleaning area and the conveying tension value of the path, the paths generated by multiple rounds of calculation are screened, and the optimal path is selected for connection and combination to form a global path; first, the paths are screened in the first round according to the priority of the cleaning area, and the paths leading to the high-priority cleaning area are retained. Then, a second round of screening is carried out according to the conveying tension value of each path, and paths with small tension fluctuations and that meet the working requirements of the cleaning robot are selected. The screened paths are connected according to the cleaning order, and the continuity and conflict of the paths are checked to ensure the rationality of the global path; thus, a global path is obtained that meets both the cleaning priority requirements and the tension constraint conditions. This global path provides a basic framework for the subsequent adjustment of the path curvature and the dynamic adjustment of the cleaning path.

[0132] Ensuring the continuity and conflict-free nature of the global path is the key to ensuring that the cleaning robot can successfully perform cleaning tasks. After the paths are connected and combined, the path needs to be inspected to promptly discover and resolve problems in the path. Through the collision detection algorithm, each path segment on the global path is checked to ensure that the path will not collide with obstacles. At the same time, the continuity of the path connection is checked to avoid path interruptions or mutations. If conflicts or discontinuities are found in the path, the path screening step is traced back and the path is reselected for combination. The global path after inspection and correction can ensure the safe operation of the cleaning robot and provide a reliable basis for the subsequent adjustment of the path curvature and the dynamic adjustment of the cleaning path.

[0133] Furthermore, the path curvature adjustment sub-logic includes:

[0134] Acquire the posture of the cleaning robot in real time, perform correlation analysis on the posture of the cleaning robot and the global path, and obtain the correlation analysis results;

[0135] Based on the correlation analysis results, the global path is divided into multiple local path segments. For each local path segment, the boundary conditions of the spline curve fitting are determined in combination with the posture of the cleaning robot at the starting and ending points of the local path segment.

[0136] Each local path segment is fitted with a cubic spline curve, and the path curvature is adjusted by adjusting the control point positions of the cubic spline curve;

[0137] Perform smoothness and feasibility verification on the local path segment after adjusting the path curvature;

[0138] The local path segments that have been adjusted and verified for path curvature are recombined to update the global path.

[0139] The posture of the cleaning robot will affect its motion stability and cleaning effect on the path. By correlating the posture of the cleaning robot with the global path, the posture that the robot should maintain at different positions on the path can be determined, providing a basis for adjusting the path curvature. The posture information of the cleaning robot, including position, direction, and tilt angle, is obtained in real time. The posture of the robot is matched with the coordinates of the global path, and the relationship between the posture of the cleaning robot at each position on the path and the path direction is analyzed. According to the analysis results, the position and direction of the curvature on the path that need to be adjusted are determined. This clarifies the goal and direction of the path curvature adjustment, and provides an important basis for the subsequent division of the global path into local path segments and the determination of the boundary conditions for spline curve fitting.

[0140] In order to adjust the path curvature more accurately, the global path needs to be divided into multiple local path segments, and the boundary conditions of spline curve fitting need to be determined for each local path segment in combination with the robot's posture; according to the results of the posture and path correlation analysis, the global path is divided into multiple local path segments at positions where the curvature changes greatly on the path (obtained by comparing with the curvature threshold) or the posture of the cleaning robot needs to be greatly adjusted (obtained by comparing with the deviation from the standard posture), and for each local path segment, the boundary conditions of spline curve fitting are determined in combination with the posture of the cleaning robot at the starting and end points of the local path segment, including the position of the starting and end points, the tangent direction and curvature, etc.; thereby, the curvature adjustment problem of the global path is transformed into a curvature optimization problem of multiple local path segments, reducing the complexity of the problem. The determined spline curve fitting boundary conditions provide the necessary parameters for the subsequent use of cubic spline curves for path curvature adjustment.

[0141] The cubic spline curve has good smoothness and flexibility. It can generate a smooth curve according to the given boundary conditions and is suitable for adjusting the path curvature. Each local path segment is fitted with the help of the cubic spline curve. The curvature of the curve is changed by adjusting the control point position of the cubic spline curve. During the adjustment process, according to the tension constraint conditions and the motion characteristics of the robot, it is ensured that the adjusted path curvature will not cause tension abnormalities or unstable robot motion. Thus, the precise adjustment of the path curvature is achieved, so that the cleaning robot can maintain a stable posture during movement and reduce tension fluctuations. The adjusted local path segment provides a basis for subsequent smoothness verification and feasibility verification.

[0142] In order to ensure that the adjusted path can meet the movement requirements of the cleaning robot, the path needs to be verified for smoothness and feasibility to avoid sudden changes or infeasibility of the path; the smoothness detection algorithm is used to check whether the curvature changes of the adjusted local path segments are continuous and smooth to avoid sudden curvature changes. At the same time, the kinematic and dynamic models of the cleaning robot are combined to verify the feasibility of the path to ensure that the cleaning robot can move according to the adjusted path and will not exceed its movement capacity; the verified local path segments have good smoothness and feasibility, which provides a reliable guarantee for the subsequent recombination of local path segments and updating of the global path.

[0143] The local path segments that have been adjusted and verified for path curvature are recombined to update the global path, so that the robot can perform cleaning operations according to the optimized path; the verified local path segments are recombined in the order of the original global path to form an updated global path. During the combination process, the connection between adjacent local path segments is ensured to be smooth to avoid path discontinuity; the updated global path has been optimized in terms of path curvature, tension constraint and robot posture, which provides a guarantee for the efficient and stable operation of the cleaning robot, and at the same time provides a new reference path for the dynamic adjustment of the cleaning path, which facilitates further optimization according to actual conditions.

[0144] Specifically, if Figure 3 As shown, the adjustment logic of the cleaning path includes:

[0145] Real-time monitoring of strain data, stain distribution images, obstacle information, and the position and posture of the cleaning robot to update the cleaning area, the cleaning robot's conveying tension value, and the cleaning path;

[0146] Determine whether to trigger the adjustment of the cleaning path based on changes in obstacle information in the cleaning area, abnormal changes in the conveying tension value, and deviation changes in the cleaning path;

[0147] When the cleaning path adjustment is triggered by abnormal changes in the conveying tension value, the avoidance mechanism is executed;

[0148] When the cleaning path adjustment is triggered by changes in obstacle information in the cleaning area and deviations of the cleaning path, local replanning is performed to obtain a local path, which is then merged with the original global path to update the global path.

[0149] The cleaning environment of the air-cooled island is complex and changeable. Strain data, stain distribution, obstacle information and the robot's own status may change at any time. To ensure the effectiveness and adaptability of the cleaning path, this information must be obtained in real time, and the cleaning area, conveying tension value and cleaning path must be updated in time so that the cleaning robot can respond to environmental changes and continuously and efficiently complete the cleaning task. The data acquisition module obtains strain data and stain distribution images in real time, and scans obstacles in the cleaning environment through lidar and ultrasonic sensors to obtain information on the position and shape of the obstacles. At the same time, the inertial measurement unit monitors the position and posture of the cleaning robot in real time, and transmits this data to the path planning module. Based on multi-threading technology, parallel processing and updating of data are realized. This ensures that the path planning module can obtain the latest environment and robot status information, providing accurate data support for subsequent judgments on whether the cleaning path needs to be adjusted. Timely updated data makes subsequent trigger condition judgments more timely and accurate.

[0150] Not all data changes require adjustment of the cleaning path. Frequent adjustments will increase the burden on the conveying control system and reduce cleaning efficiency. Therefore, clear trigger conditions need to be set. Path adjustment is triggered only when there is a significant change in the obstacle information, conveying tension value or cleaning path in the cleaning area to balance system performance and cleaning effect. Set the obstacle change scenario. When the lidar detects that the volume of the new obstacle is greater than the volume threshold, or the obstacle position is greater than the displacement threshold, the obstacle information is determined to have changed. For the conveying tension value, when the actual conveying tension value is greater than the maximum value of the conveying tension threshold or less than the minimum value of the conveying tension threshold, it is considered that the conveying tension value has changed abnormally. Through the path tracking algorithm, the actual position of the cleaning robot is compared with the preset cleaning path. When the deviation is greater than the deviation threshold, it is judged that the cleaning path has deviated. Once any of the above conditions is met, the cleaning path adjustment mechanism is triggered. Accurate trigger condition judgment avoids unnecessary path adjustments and improves the stability of the conveying control system and cleaning efficiency.

[0151] Excessive conveyor tension can damage the robot's conveyor system and affect the normal cleaning process. When the conveyor tension value increases abnormally, an avoidance mechanism is implemented to adjust the cleaning path to avoid the area causing the tension anomaly, protecting the cleaning robot and ensuring safe and stable cleaning. Using the cleaning robot's current position as the coordinate origin, a vector analysis method is used to determine the avoidance area based on the degree and direction of the conveyor tension anomaly. For example, when the conveyor tension anomaly points in a specific direction, a fan-shaped avoidance area with a fixed radius is defined around the cleaning robot based on that direction. Outside the avoidance area, an improved A* algorithm is used to search for candidate avoidance paths based on the updated cleaning area. The A* algorithm incorporates path tension constraints to ensure that new paths will not cause tension anomalies again. Candidate avoidance paths are then collided and verified for feasibility, eliminating paths that conflict with obstacles or exceed the robot's motion capabilities. This effectively avoids damage to the cleaning robot caused by tension anomalies. By rationally planning avoidance paths, a safe and feasible motion path is found for the cleaning robot. The selected avoidance paths are then smoothed to ensure smooth motion.

[0152] The change of obstacle information in the cleaning area and the deviation of the cleaning path will cause the original cleaning path to be unable to be executed smoothly, affecting the cleaning effect. By performing local replanning, the path can be adjusted quickly, so that the cleaning robot can adapt to the new environmental changes and continue to complete the cleaning task efficiently. When the obstacle information in the cleaning area changes or the cleaning path deviation is detected, the current position of the cleaning robot is used as the starting point and the target point on the original cleaning path that has not been cleaned is used as the end point. The path is replanned in the local range. The Dijkstra algorithm is used to combine the latest obstacle information and the cleaning area to generate a local path. The generated local path is merged with the original global path. The path splicing algorithm is used to ensure that the new path is smoothly connected with the original path. Local replanning quickly responds to environmental changes. The generated new path can bypass obstacles and correct path deviations. The fused global path provides accurate motion guidance for the subsequent cleaning operations of the cleaning robot, ensuring the continuity of the cleaning task.

[0153] Furthermore, the avoidance mechanism includes:

[0154] When the conveying tension is high, the avoidance area is determined based on the current position of the cleaning robot and the degree and direction of the abnormal conveying tension.

[0155] Outside the avoidance area, the A* algorithm is used to search for multiple candidate avoidance paths based on the updated cleaning area, and the feasibility of the candidate avoidance paths is verified;

[0156] Evaluate each candidate avoidance path based on cleaning efficiency, tension stability, and path safety, and determine the avoidance path;

[0157] The avoidance path is smoothed by spline curve fitting. When the cleaning robot completes the avoidance according to the avoidance path, the cleaning path is replanned according to the position and posture of the cleaning robot.

[0158] In the case of high conveying tension (greater than the maximum value of the conveying tension threshold), the avoidance area is determined to guide the robot to avoid dangerous areas that will cause further tension increase, protect the conveying control system of the cleaning robot, and provide a spatial range for the subsequent search for safe and feasible avoidance paths; the real-time position of the cleaning robot is obtained, and the avoidance area is determined by vector analysis in combination with the degree and direction of the abnormal conveying tension. For example, when the abnormal direction of the conveying tension points to a certain direction, a fan-shaped avoidance area with a fixed radius is delineated around the cleaning robot based on this direction. To ensure the avoidance effect, the boundaries of the avoidance area are smoothed to avoid sharp boundaries that make it difficult to plan the motion of the cleaning robot; the clear avoidance area defines a feasible range for the subsequent search of candidate avoidance paths, reduces the computational complexity of the path search, and improves the efficiency of path planning.

[0159] Multiple candidate avoidance paths are searched outside the avoidance area to provide diverse options for the subsequent selection of the optimal path. The feasibility of the candidate paths is verified to ensure that the paths meet the cleaning robot's motion capabilities and cleaning task requirements, avoiding the selection of infeasible paths that may cause the cleaning robot to be unable to execute. Based on the updated cleaning area, candidate avoidance paths are searched using the A* algorithm. In the A* algorithm, the cleaning area is divided into a grid map. The path search is performed with the current position of the cleaning robot as the starting point and the reachable point on the original cleaning path as the end point. During the search process, factors such as the length of the path, tension constraints, and the distance to obstacles are considered to generate multiple candidate avoidance paths. Collision detection is performed on each candidate avoidance path. The ray detection algorithm is used to check whether the path collides with obstacles. At the same time, based on the kinematic model of the cleaning robot, it is verified whether the path is within the range of the cleaning robot's motion capabilities, such as whether it is greater than the cleaning robot's turning radius and moving speed limit. After searching and verification, a set of feasible candidate avoidance paths is obtained, which provides a basis for the subsequent selection of the optimal path based on multi-index evaluation.

[0160] The optimal path is selected from multiple candidate avoidance paths, and cleaning efficiency, tension stability and path safety are comprehensively considered to ensure that the cleaning robot can complete the cleaning task efficiently and safely while avoiding areas with abnormal tension. A multi-index evaluation system is established to assign weights to cleaning efficiency, tension stability and path safety respectively. The weight values ​​are determined by the hierarchical analysis method. For cleaning efficiency, the time it takes for the path to reach the cleaning target point is calculated. For tension stability, the tension fluctuation on the path is analyzed. For path safety, the minimum distance between the path and the obstacle is evaluated. According to the evaluation system, each candidate avoidance path is scored, and the path with the highest score is selected as the avoidance path. The determined optimal avoidance path not only meets the requirements of the cleaning task but also ensures the safe and stable operation of the cleaning robot. The selected avoidance path will be smoothed to optimize the motion trajectory of the cleaning robot.

[0161] The original avoidance path contains sharp corners and discontinuous sections, which can cause large impacts during the movement of the cleaning robot, affecting the cleaning effect and the robot's stability. Smoothing makes the path more continuous and smooth, facilitating the robot's execution. After the robot completes the avoidance, the cleaning path is replanned based on its new position and posture to ensure the continuation of the cleaning task. The avoidance path is smoothed using a cubic spline curve fitting method. Key points on the avoidance path are used as control points of the spline curve. The positions and tangent directions of the control points are adjusted to generate a smooth curve. During the fitting process, the curve is ensured to meet the kinematic constraints of the cleaning robot. After the robot completes the avoidance according to the avoidance path, the robot's current position and posture are obtained. Using this position as the starting point and the remaining uncleaned area as the target, the cleaning path is replanned based on the latest cleaning area, obstacle information, and tension constraints. The smoothed avoidance path makes the robot's movement smoother and reduces equipment wear. The replanned cleaning path can continue to guide the robot to complete the cleaning task based on the robot's new position and environmental changes, ensuring the consistency and efficiency of the cleaning operation.

[0162] Example 2

[0163] like Figure 4 As shown, a flow chart of a path planning method for a cleaning robot is provided in an embodiment of the present application, and the method includes:

[0164] Real-time acquisition of strain data and stain distribution images during the conveying process;

[0165] Extract the strain peak value and strain change rate from the strain data, and extract the stain area and stain distribution position from the stain distribution image;

[0166] Determining a conveying tension value of the cleaning robot according to a strain peak value and a strain change rate in the strain data, and generating a tension constraint condition to control the conveying tension of the cleaning robot;

[0167] Divide the cleaning area according to the stain area and the stain distribution position in the stain distribution image;

[0168] Monitor obstacle information in real time to optimize the boundaries of the cleaning area and obtain the cleaning area;

[0169] Based on the position of the cleaning robot and the tension constraint conditions in the cleaning area, the cleaning robot's cleaning path is planned using the A* algorithm;

[0170] Synchronously adjust the path curvature in conjunction with the cleaning robot’s posture;

[0171] Dynamically adjust the cleaning path of the cleaning robot.

[0172] Since the principle of solving the problem by the method in the embodiment of the present application is similar to that of the system described above in the embodiment of the present application, the implementation of the method refers to the implementation of the system, and the repeated parts will not be repeated.

Claims

1. A path planning system for a cleaning robot, characterized in that: include: Real-time acquisition of strain data and stain distribution images during the conveying process; Extract the strain peak value and strain change rate from the strain data, and extract the stain area and stain distribution position from the stain distribution image; Determining a conveying tension value of the cleaning robot according to a strain peak value and a strain change rate in the strain data, and generating a tension constraint condition to control the conveying tension of the cleaning robot; The cleaning area is divided according to the stain area and stain distribution position in the stain distribution image, and the obstacle information is monitored in real time to optimize the boundary of the cleaning area to obtain the cleaning area. Within the cleaning area, based on the position of the cleaning robot and the tension constraint conditions, the cleaning path of the cleaning robot is planned by the A* algorithm, and the path curvature is synchronously adjusted in combination with the posture of the cleaning robot, and the cleaning path of the cleaning robot is dynamically adjusted.

2. The path planning system for a cleaning robot according to claim 1, wherein: The generation logic of the cleaning area includes: Preliminarily divide the cleaning areas according to the stain area and stain distribution position in the stain distribution image, and simultaneously assign a cleaning priority to each cleaning area according to the stain area and stain distribution position; Monitor obstacles in the cleaning environment in real time and superimpose the location and shape of obstacles on the cleaning area; The obstacles are expanded by the expansion algorithm to optimize the boundaries of the cleaning area, and the final cleaning area and the cleaning priority of each cleaning area are obtained.

3. The path planning system for a cleaning robot according to claim 2, wherein: The planning logic of the cleaning path includes: According to the position of the cleaning robot, the cleaning area and the tension constraint conditions, the A* algorithm is used to generate a path from the position of the cleaning robot to the cleaning area; Filter paths based on the cleaning priority of the cleaning area and combine them to form a global path; Based on the posture of the cleaning robot, the global path is locally optimized through spline curve fitting, and the path curvature is adjusted to obtain the cleaning path.

4. The path planning system for a cleaning robot according to claim 3, wherein: The formation sub-logic of the global path includes: When the A* algorithm is used to generate the path from the cleaning robot's position to each cleaning area, the tension constraint condition is refined into a heuristic function with multiple dimensional parameters embedded in the A* algorithm; Through multiple rounds of calculation, multiple paths from the cleaning robot's position to the cleaning area are generated, and the conveying tension value of each path is marked; The paths are screened first based on the cleaning priority of each cleaning area, and then screened again based on the conveying tension value of each path; The screened paths are connected and combined to form a global path, and the continuity and conflict of the global path are detected.

5. The path planning system for a cleaning robot according to claim 4, characterized in that: The path curvature adjustment sub-logic includes: Acquire the posture of the cleaning robot in real time, perform correlation analysis on the posture of the cleaning robot and the global path, and obtain the correlation analysis results; Based on the correlation analysis results, the global path is divided into multiple local path segments. For each local path segment, the boundary conditions of the spline curve fitting are determined in combination with the posture of the cleaning robot at the starting and ending points of the local path segment. Each local path segment is fitted with a cubic spline curve, and the path curvature is adjusted by adjusting the control point positions of the cubic spline curve; Perform smoothness and feasibility verification on the local path segment after adjusting the path curvature; The local path segments that have been adjusted and verified for path curvature are recombined to update the global path.

6. The path planning system for a cleaning robot according to claim 5, characterized in that: The adjustment logic of the cleaning path includes: Real-time monitoring of strain data, stain distribution images, obstacle information, and the position and posture of the cleaning robot to update the cleaning area, the cleaning robot's conveying tension value, and the cleaning path; Determine whether to trigger the adjustment of the cleaning path based on changes in obstacle information in the cleaning area, abnormal changes in the conveying tension value, and deviation changes in the cleaning path; When the cleaning path adjustment is triggered by abnormal changes in the conveying tension value, the avoidance mechanism is executed; When the cleaning path adjustment is triggered by changes in obstacle information in the cleaning area and deviations of the cleaning path, local replanning is performed to obtain a local path, which is then merged with the original global path to update the global path.

7. The path planning system for a cleaning robot according to claim 6, wherein: The avoidance mechanism includes: When the conveying tension is high as determined by comparing the conveying tension threshold, the avoidance area is determined based on the current position of the cleaning robot and the degree and direction of the abnormal conveying tension. Outside the avoidance area, the A* algorithm is used to search for multiple candidate avoidance paths based on the updated cleaning area, and the feasibility of the candidate avoidance paths is verified; Evaluate each candidate avoidance path based on cleaning efficiency, tension stability, and path safety, and determine the avoidance path; The avoidance path is smoothed by spline curve fitting. When the cleaning robot completes the avoidance according to the avoidance path, the cleaning path is replanned according to the position and posture of the cleaning robot.

8. The path planning system for a cleaning robot according to claim 7, wherein: The generation logic of the tension constraint condition includes: Determine the conveying tension threshold of the cleaning robot based on the mechanical safety of the cleaning robot; Dynamically adjust the conveying tension threshold of the cleaning robot according to the stain area and stain distribution location; The conveying tension threshold of the cleaning robot is corrected based on the energy consumption of the cleaning robot to generate a tension constraint condition.

9. The path planning system for a cleaning robot according to claim 8, wherein: The acquisition logic of the stain distribution image includes: Combining a multispectral imaging camera and a visible light camera to obtain a camera group to capture the cleaning environment; Deploy the lighting array, adjust the light source parameters of the lighting array, and use the camera group to obtain multi-angle images; Multi-angle images are stitched and corrected through image stitching and distortion correction to generate a stain distribution image.

10. The path planning system for a cleaning robot according to claim 9, characterized in that: The extraction logic of the stain area and stain distribution position includes: Performing image segmentation on the stain distribution image to determine the stain area, and performing morphological processing on the stain area to obtain a stain area image; The number of pixels in the stained area of ​​the stained area image is counted by a pixel statistics method, and the number of pixels in the stained area is converted into the stained area of ​​the stained area; The stained regions in the stained region image are marked by a region marking algorithm, an identifier is assigned to each stained region, and the centroid coordinates of each stained region are calculated to determine the stain distribution position of the stained region.

11. A path planning method for a cleaning robot, implemented based on the path planning system for a cleaning robot according to any one of claims 1 to 10, characterized in that: include: Real-time acquisition of strain data and stain distribution images during the conveying process; Extract the strain peak value and strain change rate from the strain data, and extract the stain area and stain distribution position from the stain distribution image; Determining a conveying tension value of the cleaning robot according to a strain peak value and a strain change rate in the strain data, and generating a tension constraint condition to control the conveying tension of the cleaning robot; Divide the cleaning area according to the stain area and the stain distribution position in the stain distribution image; Monitor obstacle information in real time to optimize the boundaries of the cleaning area and obtain the cleaning area; Based on the position of the cleaning robot and the tension constraint conditions in the cleaning area, the cleaning robot's cleaning path is planned using the A* algorithm; Synchronously adjust the path curvature in conjunction with the cleaning robot’s posture; Dynamically adjust the cleaning path of the cleaning robot.

Images