Obstacle avoidance method and system for unmanned sweeper

Abstract

The invention relates to the technical field of unmanned control, in particular to an obstacle avoidance method and system for an unmanned sweeper, and the method comprises the steps: obtaining a map of a region, needing to be cleaned, of a target floor in a shopping mall; dividing a plurality of cleaning areas according to the turnout nodes on the basis of an area map; acquiring a plurality of parking points set in the regional map, and generating a cleaning path based on a connecting line of two adjacent parking points to obtain a cleaning path network; obtaining the people flow density in a preset range around each parking point, and obtaining a people density distribution diagram; predicting a people flow density change condition based on the people density distribution diagram and the data of people entering and exiting the current floor; and when the total people flow density of the current floor is greater than a first preset value, performing cleaning based on the cleaning path network, and when the total people flow density is less than the first preset value, performing centralized cleaning in the cleaning area. The cleaning efficiency can be improved, the operation cost is reduced, and meanwhile the environment quality and the user experience of a shopping mall are improved.

Description

Technical Field

[0001] The present invention relates to the field of unmanned driving control technology, and in particular to an obstacle avoidance method and system for an unmanned driving sweeper. Background Art

[0002] Unmanned sweepers are mechanical devices that use modern information technology, sensor technology, and automation technology to autonomously complete sweeping tasks without direct human control. Such sweepers are usually equipped with advanced navigation systems, environmental perception devices (such as laser radar, cameras, ultrasonic sensors, etc.), and computer systems for data analysis and processing, allowing them to accurately identify garbage and clean it in a predetermined area.

[0003] The unmanned sweeper in the mall receives cleaning instructions from the central system and uses preset maps and path planning to navigate autonomously in the mall. It uses sensors such as lidar and cameras to sense the surrounding environment, avoid obstacles, and sweep the ground along the planned route. When encountering pedestrians, it will automatically avoid them to ensure safety. The garbage collected during the cleaning process will be stored in the dust box. When the dust box is close to full, the unmanned sweeper will automatically return to the designated location to dump the garbage and return to the charging station to replenish energy so that it can continue to complete the cleaning task.

[0004] However, when executing the above cleaning process, if there are too many people in the mall, the cleaning efficiency will be reduced and the travel of people will be affected. If there are too few people, the cleaning process needs to be started manually, which reduces work efficiency. Summary of the invention

[0005] The purpose of the present invention is to provide an obstacle avoidance method and system for an unmanned sweeper, which is intended to improve cleaning efficiency, reduce operating costs, and also improve the environmental quality and user experience of a shopping mall.

[0006] To achieve the above object, the present invention provides an obstacle avoidance method for an unmanned sweeping vehicle, comprising obtaining a map of an area to be cleaned on a target floor in a shopping mall;

[0007] Divide multiple cleaning areas according to fork nodes based on the regional map;

[0008] Acquire multiple parking spots set in the area map, and generate a cleaning path based on a line connecting two adjacent parking spots to obtain a cleaning path network;

[0009] Obtain the density of people flow within a preset range around each parking spot to obtain a population density distribution map;

[0010] Predict changes in crowd density based on the crowd density distribution map and data on people entering and leaving the current floor;

[0011] When the overall crowd density on the current floor is greater than a first preset value, cleaning is performed based on the cleaning path network; when it is less than the first preset value, centralized cleaning is performed in the cleaning area.

[0012] The specific steps of obtaining the map of the area to be cleaned on the target floor in the shopping mall include: obtaining a basic floor plan of the target floor, including the functional division of each area, the location of the passage and the entrance and exit;

[0013] Scan the target floor and identify obstacles;

[0014] Mark all obstacles that need to be avoided on the basic plan to obtain an area map.

[0015] The specific steps of dividing the multiple cleaning areas according to the fork nodes based on the area map include:

[0016] Identify all fork nodes on the base map;

[0017] Divide the entire floor into several independent cleaning areas according to the fork nodes;

[0018] Set boundaries for each cleaning area;

[0019] Plan the paths connecting the cleaning areas and obtain a cleaning area map.

[0020] The specific manner of setting the multiple parking spots in the area map includes:

[0021] Get the coordinates of all fork nodes;

[0022] Scan the obstacle coverage area within a preset radius based on the coordinates of the fork node;

[0023] When the obstacle coverage area is smaller than the preset area, the boundary of the cleaned area closest to the fork node is used as the parking point;

[0024] Mark the selected parking spot locations on the area map.

[0025] The specific steps of generating a cleaning path based on the connection line between two adjacent parking points to obtain a cleaning path network include:

[0026] Obtain the location coordinates of all temporary parking spots determined on the regional map;

[0027] Use curves to connect adjacent parking spots to form a path network;

[0028] Apply Dijkstra algorithm to optimize the path to avoid repeated sweeping and obtain the basic path;

[0029] On the basis of the basic path, the path is further optimized based on the obstacle data to obtain a clearing path network.

[0030] The specific steps of obtaining the density of people within a preset range around each parking spot and obtaining a population density distribution map include:

[0031] Install cameras in key areas of the mall, including entrances, exits, elevators, escalators, and rest areas;

[0032] Preprocess the video frames captured by the camera;

[0033] Use the YOLO detection model to detect people in the video frame;

[0034] Use multi-target tracking algorithms to track each individual in the video and estimate its trajectory;

[0035] Assign tracked individuals to cleaning areas;

[0036] Calculate the average number of people in each cleaning area to get the crowd density in the area, and finally combine them to get the crowd density on each floor.

[0037] The specific steps of performing cleaning based on the cleaning path network when the overall crowd density of the current floor is greater than a first preset value and performing centralized cleaning in the cleaning area when the overall crowd density of the current floor is less than the first preset value include:

[0038] Set crowd density thresholds based on historical data;

[0039] Compare the floor crowd density with the crowd density threshold, if the crowd density is greater than the crowd density threshold, enter the path cleaning mode; if the crowd density is less than the first preset value, enter the area centralized cleaning mode;

[0040] In the area concentrated cleaning mode, the crowd density in the cleaning area is continuously cleaned. When the crowd density is greater than the preset value, the cleaning vehicle is driven to the parking point and stays there until the crowd density is less than the preset value before cleaning.

[0041] In a second aspect, the present invention further provides an obstacle avoidance system for an unmanned sweeping vehicle, including a map acquisition module, an area division module, a cleaning path generation module, a personnel density calculation module, a prediction module, and a cleaning strategy generation module;

[0042] The map acquisition module is used to acquire a map of the area that needs to be cleaned on the target floor in the shopping mall;

[0043] The area division module is used to divide a plurality of cleaning areas according to the fork nodes based on the area map;

[0044] The cleaning path generation module is used to obtain multiple parking points set in the area map, and generate a cleaning path based on the connection line between two adjacent parking points to obtain a cleaning path network;

[0045] The personnel density calculation module is used to obtain the density of people within a preset range around each parking spot to obtain a personnel density distribution map;

[0046] The prediction module is used to predict changes in crowd density based on the crowd density distribution map and the data of people entering and leaving the current floor;

[0047] The cleaning strategy generation module is used to perform cleaning based on the cleaning path network when the overall crowd density on the current floor is greater than a first preset value, and to perform concentrated cleaning in the cleaning area when the overall crowd density on the current floor is less than the first preset value.

[0048] The obstacle avoidance method and system of an unmanned sweeper of the present invention scans and constructs a three-dimensional map of the interior of a shopping mall by integrating high-precision sensors. The map contains specific information of all areas that need to be cleaned, such as the location and spatial dimensions of obstacles, and provides basic data support for subsequent cleaning path planning. The key nodes in the map information are used as reference points to divide the entire floor into a number of independent cleaning units. Such a partitioning method is conducive to reducing the time waste caused by path overlap during the cleaning process, and is also convenient for adopting differentiated cleaning strategies for different areas. The selection of parking spots should take into account the convenience of sweepers entering and exiting without affecting the passage of customers. Based on this, one or more optimal cleaning paths are planned in combination with the shortest path algorithm to form a cleaning path network covering the entire floor. This design can effectively shorten the cleaning time and improve the overall work efficiency. The flow of people near each parking spot is monitored using environmental perception technology, and a population density distribution map is drawn accordingly. This helps to identify which areas are more crowded and which areas are relatively open, providing a basis for dynamically adjusting the cleaning plan. Combining historical data with real-time monitoring results, big data analysis and artificial intelligence technologies (such as deep learning models) are used to predict the changing trend of pedestrian flow in the future. The cleaning strategy is adjusted dynamically according to the flow of people. During periods of dense traffic, the sweeper drives slowly along the cleaning path network and cleans to minimize the impact on pedestrians; while during periods of sparse traffic, more detailed cleaning work can be carried out in designated areas. Through the application of the above steps and technologies, this method realizes the intelligent operation of unmanned sweepers, which not only improves cleaning efficiency and reduces operating costs, but also improves the environmental quality and user experience of the mall. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0050] Figure 1 It is a flow chart of an obstacle avoidance method for an unmanned sweeping vehicle of the present invention.

[0051] Figure 2 It is a flow chart of the present invention for obtaining a map of an area that needs to be cleaned on a target floor in a shopping mall.

[0052] Figure 3 It is a flow chart of the present invention for dividing a plurality of cleaning areas according to fork nodes based on an area map.

[0053] Figure 4 It is a flow chart of multiple parking spots set in the area map of the present invention.

[0054] Figure 5 It is a flow chart of the present invention for generating a cleaning path based on the connection line between two adjacent parking points to obtain a cleaning path network.

[0055] Figure 6 It is a flow chart of the present invention for obtaining the density of human traffic within a preset range around each parking spot and obtaining a human density distribution map.

[0056] Figure 7 It is a flowchart of the present invention that when the overall crowd density of the current floor is greater than a first preset value, cleaning is performed based on the cleaning path network, and when it is less than the first preset value, concentrated cleaning is performed in the cleaning area. DETAILED DESCRIPTION

[0057] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0058] First embodiment

[0059] See also Figure 1 to Figure 7 The present invention provides an obstacle avoidance method for an unmanned sweeping vehicle, comprising:

[0060] S101 obtains a map of the area that needs to be cleaned on the target floor of the shopping mall;

[0061] The specific steps include:

[0062] S201 obtains a basic floor plan of the target floor, including functional divisions of various areas, channel locations, and entrances and exits;

[0063] Obtain the original architectural drawings or CAD files of the target floor from the mall management or architectural design department. These files contain the basic structural information of the floor, such as functional area division, wall location, doors and windows, etc. If the original data is a paper document, it needs to be digitized and converted into an electronic format (such as PDF or vector graphics) for further processing and editing.

[0064] S202 scans the target floor and identifies obstacles;

[0065] Scan the target floor using LiDAR, 3D cameras or other advanced scanning equipment. These devices can capture detailed three-dimensional spatial information. Scanning should be done during uninterrupted time periods (such as after the mall is closed) to ensure data accuracy. During the scanning process, it is necessary to cover all areas, including corners and narrow spaces. The scanned point cloud data is processed and computer vision algorithms are used to identify the location of static obstacles such as furniture, display cabinets, vending machines, etc.

[0066] S203 marks all obstacles that need to be avoided on the basic plan to obtain an area map.

[0067] Overlay the scan results with the base floor plan to ensure that the coordinate systems are consistent. Mark all identified obstacles on the floor plan, using different colors or icons to distinguish different types of objects (for example, red for fixed, immovable obstacles and blue for temporarily placed items). Redefine the cleaning area based on the location of obstacles, marking which areas are open areas that can be cleaned and which areas are obstacles that need to be avoided.

[0068] S102 divides a plurality of cleaning areas according to fork nodes based on the area map;

[0069] The specific steps include:

[0070] S301 identifies all fork nodes on the basic map;

[0071] A fork in the road usually refers to the intersection of corridors or passages, that is, the point where at least three paths intersect. In practical applications, it may also include elevator entrances, staircase entrances and other locations. These nodes are automatically identified from the floor plan using image processing algorithms, such as edge detection and feature point extraction. Since automatic identification may have errors, manual review of the identification results is required to ensure that each fork in the road is correctly marked.

[0072] S302 divides the entire floor into a number of independent cleaning areas according to the fork nodes;

[0073] Based on the location of the fork nodes, the floor is divided into several relatively independent areas using the connectivity analysis method. These areas should be continuous and non-intersecting, and each area should have one or more clear entrances / exits.

[0074] S303 sets a boundary for each cleaning area;

[0075] Define clear boundaries for each cleaning area, which can be walls, railings, or identified fork nodes. Use lines or colors on the map to mark the boundaries of each area to ensure that the cleaning robot or staff can clearly distinguish between different areas.

[0076] S304 plans a path connecting each cleaning area to obtain a cleaning area map.

[0077] Design the shortest or optimal path so that the cleaning robot can move efficiently between different areas. This may involve using the shortest path algorithm in graph theory (such as Dijkstra's algorithm) or other path planning algorithms. Simulate the results of path planning to ensure that the planned path does not cause traffic congestion or cleaning dead ends. Through the above steps, the target floors in the mall can be effectively divided into multiple areas that are easy to manage and clean, and the connecting paths between areas can be planned to provide guidance for subsequent cleaning work.

[0078] S103: acquiring a plurality of parking spots set in the area map, and generating a cleaning path based on a line connecting two adjacent parking spots to obtain a cleaning path network;

[0079] The specific method of setting the multiple parking spots in the area map includes:

[0080] S401 obtains the coordinates of all fork nodes;

[0081] Extract the coordinate information of each node from the previously identified fork nodes. These coordinates should be in a unified coordinate system to ensure the consistency and accuracy of the coordinate data. Create a list containing the coordinates of all fork nodes for subsequent processing.

[0082] S402 scans the obstacle coverage area within a preset radius based on the coordinates of the fork node;

[0083] Set a preset radius to scan the area around the fork in the road. The radius should be chosen taking into account the size and operating range of the cleaning robot. Use the algorithm to scan obstacles within the preset radius around each fork in the road and record the area covered by these obstacles. Calculate the total area of ​​obstacles in each scanned area.

[0084] S403: when the obstacle coverage area is smaller than the preset area, the boundary of the cleaned area closest to the fork node is used as the parking point;

[0085] Set a threshold for the obstacle coverage area to determine whether it is suitable to set up a parking spot. If the obstacle coverage area around a fork in the road is lower than this threshold, it is considered that there is enough space to set up a parking spot near the node. For the fork in the road that meets the conditions, find the boundary of the cleaning area closest to the node. The boundary here should be a location that is not blocked by obstacles and is suitable for parking the cleaning robot. The found boundary positions are determined as parking spots, ensuring that these parking spots are close to the fork in the road and do not hinder traffic.

[0086] S404 marks the selected parking spot location on the area map.

[0087] Mark all selected parking spots on the cleaning area map with specific symbols or colors for easy identification and management. Create a list of parking spot locations, record the specific coordinates and related information of each parking spot. Review all selected parking spots to ensure that they are located reasonably and meet the requirements of the cleaning operation.

[0088] The specific steps of generating a cleaning path based on the connection line between two adjacent parking points to obtain a cleaning path network include:

[0089] S501 obtains the location coordinates of all temporary parking spots determined on the regional map;

[0090] Get the coordinates of all the parking spots that have been determined from the previous step. These coordinates should be in a unified coordinate system, and the location of each parking spot has been confirmed to be suitable for parking the cleaning robot. Organize the coordinate information of all parking spots into an ordered list or database for subsequent processing.

[0091] S502 uses curves to connect adjacent parking spots to form a path network;

[0092] For each pair of adjacent parking spots, define a path connecting them. "Adjacent" here means that the two parking spots can be connected by a direct path within the cleaning area. Use appropriate algorithms (such as Bezier curves, spline curves, etc.) to generate smooth paths to ensure that the cleaning robot can move smoothly from one parking spot to another.

[0093] All generated paths are combined to form a path network covering the entire cleaning area. The network should avoid repeated paths as much as possible and take into account the actual cleaning efficiency.

[0094] S503 applies Dijkstra algorithm to optimize the path to avoid repeated sweeping and obtain a basic path;

[0095] The cleaning area is considered as a graph, where parking points are nodes, paths connecting parking points are edges, and each edge is given a certain weight (such as path length, cleaning priority, etc.). Use the Dijkstra algorithm to find the shortest path on this graph. The algorithm can start from the starting point and gradually expand the path until all nodes are covered, while ensuring that the total length of the path is minimized. The Dijkstra algorithm is used to find the optimal set of paths, which should be able to cover all areas that need to be cleaned and minimize the possibility of repeated cleaning.

[0096] S504 further optimizes the path based on the obstacle data on the basis of the basic path to obtain a cleaning path network.

[0097] Based on the existing path, combined with the previously scanned obstacle data, the path is adjusted to ensure that the cleaning robot does not collide with obstacles during its movement. For those parts that need to be detoured due to obstacles, the local path is replanned to avoid obstacles while keeping the path as optimized as possible. Check the path network within the entire cleaning area to ensure that there are no missed areas and that all paths are optimal, that is, they can achieve the cleaning effect while shortening the cleaning time as much as possible. Perform a comprehensive test on the optimized path network to ensure that all paths are feasible and can effectively avoid obstacles.

[0098] Through the above steps, a complete and efficient cleaning path network can be generated to provide accurate path guidance for the automated operation of the cleaning robot and ensure the smooth completion of the cleaning task.

[0099] S104 obtains the density of people flow within a preset range around each parking spot to obtain a population density distribution map;

[0100] The specific steps include:

[0101] S601 installs cameras in key areas of the mall, including entrances, exits, elevators, escalators, and rest areas;

[0102] Install high-resolution cameras in key areas of the mall (such as entrances, exits, elevator halls, escalator areas, rest areas, etc.) to ensure that these areas can be effectively monitored. Adjust the angle and height of the camera to ensure that the video coverage is wide and the image is clear, avoiding dead angles or blind spots. Ensure that the installation process complies with local privacy protection laws, such as desensitizing video data and blocking sensitive information such as faces.

[0103] S602 pre-processes the video frames collected by the camera;

[0104] Connect the video stream collected by the camera to the central processing system or cloud server. Preprocess the video frames, including but not limited to adjusting brightness and contrast, removing noise, and ensuring image quality for subsequent analysis. Split the video stream into individual frame images for subsequent frame-by-frame analysis.

[0105] S603 uses the YOLO detection model to detect people in the video frame;

[0106] Choose YOLO (You Only Look Once) or other similar object detection models to detect people in the video frames. If necessary, the YOLO model can be fine-tuned or retrained according to the special needs of the shopping mall environment to improve the detection accuracy. Use the model to frame the detected people in each frame in preparation for subsequent tracking.

[0107] S604 tracks each individual in the video using a multi-target tracking algorithm and estimates its trajectory;

[0108] Use an algorithm suitable for multi-target tracking, such as DeepSORT (Deep-Learned Selective Object Tracking and Recognition), to track each individual detected in the video. Use the algorithm to track the movement trajectory of the individual and assign a unique identifier to each individual. Estimate the movement trajectory of each individual based on the tracking results to prepare for the next step of assigning individuals to cleaning areas.

[0109] S605 assigning the tracked individuals to cleaning areas;

[0110] According to the previously divided cleaning areas, each individual tracked in the video is assigned to the cleaning area where it is located. Ensure that the trajectory information of each individual corresponds to the cleaning area where it is located, which is convenient for the subsequent calculation of crowd density.

[0111] S606 calculates the average number of people in each cleaning area, thereby obtaining the crowd density of the area, and finally combining to obtain the crowd density of the floor.

[0112] In a certain period of time, count the number of people in each cleaning area, which can be instantaneous or cumulative. Calculate the crowd density of the area based on the area of ​​the cleaning area and the average number of people in the area. Summarize the crowd density information of all cleaning areas to generate a crowd density distribution map for the entire floor, using different colors or legends to represent areas with different density levels.

[0113] S105 predicts changes in crowd density based on the crowd density distribution map and data of people entering and leaving the current floor;

[0114] Integrate information from different data sources, such as cameras, access control systems, elevator usage records, etc., to ensure real-time data on people entering and leaving the building. Clean the collected raw data, remove invalid or duplicate data records, and ensure data accuracy and reliability. Format the data into a unified standard format for subsequent processing and analysis. Based on the latest data on people entering and leaving the building, update the population density distribution map in real time to reflect the density of people flow in each area of ​​the current floor. Analyze which areas are densely populated and which areas are relatively empty, providing basic information for predicting future changes in people flow.

[0115] By comparing with historical data, detect whether there is any abnormal crowd gathering or evacuation.

[0116] Select appropriate time series forecasting models, such as ARIMA (autoregressive integrated moving average model), LSTM (long short-term memory network), etc., to predict the changing trend of passenger flow density. Extract useful feature variables based on factors that affect passenger flow (such as time, weather, holidays, promotional activities, etc.) for model training. Use historical data to train the model and adjust model parameters to improve forecast accuracy. Predict changes in passenger flow density in the next few hours or a day to provide a basis for immediate management.

[0117] Through the above steps, the change of the density of people flow on the current floor of the mall can be effectively predicted, providing strong support for the operation management and resource allocation of the mall. Such prediction can not only help the mall better cope with sudden large passenger flow, but also improve the customer experience and ensure the smooth and orderly operation of the mall.

[0118] S106: When the overall crowd density on the current floor is greater than a first preset value, cleaning is performed based on the cleaning path network; when it is less than the first preset value, centralized cleaning is performed in the cleaning area.

[0119] The specific steps include:

[0120] S701 sets a crowd density threshold according to historical data;

[0121] Utilize the floor crowd density information in historical data to analyze the changing patterns of crowd flow in different time periods.

[0122] Based on these data, a reasonable first preset value is determined as the crowd density threshold, which should be the best balance point that ensures cleaning efficiency while not affecting pedestrian traffic. Considering that crowd density is affected by many factors (such as special events, holidays, etc.), the threshold setting should have a certain degree of flexibility and can be adjusted in time according to actual conditions.

[0123] S702 compares the floor crowd density with the crowd density threshold, and if the crowd density is greater than the crowd density threshold, enters the path cleaning mode; if the crowd density is less than the first preset value, enters the area centralized cleaning mode;

[0124] The crowd density of the current floor is monitored in real time through sensors or cameras installed in the floor. The real-time crowd density data monitored is compared with the first preset value set previously. If the current crowd density is higher than the first preset value, the system automatically switches to the cleaning path network mode; if it is lower than the first preset value, it switches to the regional centralized cleaning mode.

[0125] S703: In the area concentrated cleaning mode, the crowd density in the cleaning area is continuously cleaned. When the crowd density is greater than the preset value, the cleaning vehicle is driven to the parking point and stops until the crowd density is less than the preset value before cleaning.

[0126] When the density of people is low, the cleaning robots concentrate on certain designated areas for thorough cleaning. During the cleaning process, the changes in the density of people in the area are continuously monitored. Once the density of people in the area exceeds the preset value, the cleaning robot should stop cleaning immediately and move to a pre-set safe location or parking point to wait. When the density of people in the area drops below the preset value again, the cleaning robot restarts the cleaning task, and repeats this cycle to ensure that the cleaning work is completed without affecting the passage of pedestrians.

[0127] Second embodiment

[0128] On the basis of the first embodiment, the present invention also provides an obstacle avoidance system for an unmanned sweeper, including a map acquisition module, an area division module, a cleaning path generation module, a personnel density calculation module, a prediction module, and a cleaning strategy generation module; the map acquisition module is used to obtain an area map of a target floor in a shopping mall that needs to be cleaned; the area division module is used to divide a plurality of cleaning areas according to fork nodes based on the area map; the cleaning path generation module is used to obtain a plurality of parking points set in the area map, and generate a cleaning path based on the connection line between two adjacent parking points to obtain a cleaning path network; the personnel density calculation module is used to obtain the crowd density within a preset range around each parking point to obtain a crowd density distribution map; the prediction module is used to predict the change of crowd density based on the crowd density distribution map and the data of people entering and exiting the current floor; the cleaning strategy generation module is used to perform cleaning based on the cleaning path network when the overall crowd density of the current floor is greater than a first preset value, and to perform concentrated cleaning in the cleaning area when it is less than the first preset value.

[0129] In this embodiment, the map acquisition module collects environmental information of the target floor (such as a specific floor in a shopping mall) and generates an electronic map of the area. These maps are usually constructed from data collected by sensors such as laser radar and cameras, and can accurately depict the outline of the floor, the location of obstacles, and other important features.

[0130] Based on the data obtained from the map acquisition module, the region division module divides the entire cleaning area into several smaller sub-areas according to actual geographical features (for example, through detected corridor intersections or other natural dividing lines). This helps optimize the cleaning path and avoid repeated cleaning or missing cleaning of certain areas.

[0131] After understanding the specific location of each cleaning area, the cleaning path generation module will determine a series of parking points as references. It will plan the cleaning route based on these parking points, that is, connecting two adjacent parking points to form a cleaning path, and then build a complete cleaning path network.

[0132] The crowd density calculation module evaluates the flow of people near each parking spot in the cleaning area and forms a real-time or near-real-time crowd density map by counting the number of pedestrians within a certain range. Such information is crucial for formulating a reasonable cleaning plan.

[0133] The prediction module uses the data provided by the personnel density calculation module and the dynamics of personnel entering and leaving the current floor. The prediction module can analyze and infer the trend of personnel flow changes in the future. This can help the sweeper avoid peak hours and choose the appropriate time period for operation.

[0134] According to the results of the prediction module, when the overall flow of people exceeds the pre-set threshold, the cleaning strategy generation module will guide the sweeper to gradually clean along the cleaning path network. When the flow of people is less, it can concentrate on specific high-density areas for deep cleaning.

[0135] Through the coordinated work of the above modules, the driverless sweeper can autonomously navigate in complex environments and effectively perform cleaning tasks while ensuring the safety and efficiency of operations. This design not only improves the automation level of cleaning work, but also provides a new solution for maintaining the sanitation of public spaces.

[0136] What is disclosed above is only a preferred embodiment of the present invention, and it certainly cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiment and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. An obstacle avoidance method for an unmanned sweeping vehicle, characterized in that: include: Get a map of the area that needs to be cleaned on the target floor of the mall; Divide multiple cleaning areas according to fork nodes based on the regional map; Acquire multiple parking spots set in the area map, and generate a cleaning path based on a line connecting two adjacent parking spots to obtain a cleaning path network; Obtain the density of people flow within a preset range around each parking spot to obtain a population density distribution map; Predict changes in crowd density based on the crowd density distribution map and data on people entering and leaving the current floor; When the overall crowd density on the current floor is greater than a first preset value, cleaning is performed based on the cleaning path network; when it is less than the first preset value, centralized cleaning is performed in the cleaning area.

2. The obstacle avoidance method for an unmanned sweeping vehicle according to claim 1, characterized in that: The specific steps of obtaining the area map of the target floor in the shopping mall that needs to be cleaned include: Obtain the basic floor plan of the target floor, including the functional division of each area, the location of passages and entrances and exits; Scan the target floor and identify obstacles; Mark all obstacles that need to be avoided on the basic plan to obtain an area map.

3. The obstacle avoidance method for an unmanned sweeping vehicle as claimed in claim 2, characterized in that: The specific steps of dividing a plurality of cleaning areas according to the fork nodes based on the area map include: Identify all fork nodes on the base map; Divide the entire floor into several independent cleaning areas according to the fork nodes; Set boundaries for each cleaning area; Plan the paths connecting the cleaning areas and obtain a cleaning area map.

4. The obstacle avoidance method for an unmanned sweeping vehicle as claimed in claim 3, characterized in that: The specific method of setting the multiple parking spots in the area map includes: Get the coordinates of all fork nodes; Scan the obstacle coverage area within a preset radius based on the coordinates of the fork node; When the obstacle coverage area is smaller than the preset area, the boundary of the cleaned area closest to the fork node is used as the parking point; Mark the selected parking spot locations on the area map.

5. The obstacle avoidance method for an unmanned sweeping vehicle as claimed in claim 4, characterized in that: The specific steps of generating a cleaning path based on the connection line between two adjacent parking points to obtain a cleaning path network include: Obtain the location coordinates of all temporary parking spots determined on the regional map; Use curves to connect adjacent parking spots to form a path network; Apply Dijkstra algorithm to optimize the path to avoid repeated sweeping and obtain the basic path; On the basis of the basic path, the path is further optimized based on the obstacle data to obtain a clearing path network.

6. The obstacle avoidance method for an unmanned sweeping vehicle as claimed in claim 5, characterized in that: The specific steps of obtaining the density of people within a preset range around each parking spot and obtaining a population density distribution map include: Install cameras in key areas of the mall, including entrances, exits, elevators, escalators, and rest areas; Preprocess the video frames captured by the camera; Use the YOLO detection model to detect people in the video frame; Use multi-target tracking algorithms to track each individual in the video and estimate its trajectory; Assign tracked individuals to cleaning areas; Calculate the average number of people in each cleaning area to get the crowd density in the area, and finally combine them to get the crowd density on each floor.

7. The obstacle avoidance method for an unmanned sweeping vehicle as claimed in claim 6, characterized in that: The specific steps of performing cleaning based on the cleaning path network when the overall crowd density of the current floor is greater than the first preset value and performing centralized cleaning in the cleaning area when the overall crowd density of the current floor is less than the first preset value include: Set crowd density thresholds based on historical data; Compare the floor crowd density with the crowd density threshold, if the crowd density is greater than the crowd density threshold, enter the path cleaning mode; if the crowd density is less than the first preset value, enter the area centralized cleaning mode; In the area concentrated cleaning mode, the crowd density in the cleaning area is continuously cleaned. When the crowd density is greater than the preset value, the cleaning vehicle is driven to the parking point and stays there until the crowd density is less than the preset value before cleaning.

8. An obstacle avoidance system for an unmanned sweeping vehicle, using an obstacle avoidance method for an unmanned sweeping vehicle as claimed in any one of claims 1 to 7, characterized in that: It includes map acquisition module, area division module, cleaning path generation module, personnel density calculation module, prediction module, and cleaning strategy generation module; The map acquisition module is used to acquire a map of the area that needs to be cleaned on the target floor in the shopping mall; The area division module is used to divide a plurality of cleaning areas according to the fork nodes based on the area map; The cleaning path generation module is used to obtain multiple parking points set in the area map, and generate a cleaning path based on the connection line between two adjacent parking points to obtain a cleaning path network; The personnel density calculation module is used to obtain the density of people within a preset range around each parking spot to obtain a personnel density distribution map; The prediction module is used to predict changes in crowd density based on the crowd density distribution map and the data of people entering and leaving the current floor; The cleaning strategy generation module is used to perform cleaning based on the cleaning path network when the overall crowd density on the current floor is greater than a first preset value, and to perform concentrated cleaning in the cleaning area when the overall crowd density on the current floor is less than the first preset value.

Images