An unmanned sweeper cleaning method based on target distribution situation analysis

Abstract

The present application relates to the technical field of cleaning mechanism control, and particularly relates to a sweeping method of unmanned sweeper based on target distribution situation analysis. The method comprises the following steps: establishing a corresponding cleaning operation area of the unmanned sweeper, dividing the cleaning operation area into multiple cleaning units; identifying the road boundary of the cleaning unit according to the road point cloud data collected by the laser radar; identifying the garbage target along the road boundary, determining the garbage cleaning mode for the garbage target; identifying the passing obstacle on the road boundary, and setting the cleaning path; controlling the unmanned sweeper to clean the garbage target in the garbage cleaning mode along the cleaning path, and detecting the target distribution situation; and calculating the garbage target residual degree according to the target distribution situation. The present application can realize edge cleaning and automatic dumping of full garbage bin based on the cleaning mechanism control technology, and improve the cleaning efficiency of the unmanned sweeper.

Description

Technical Field

[0001] This invention relates to the field of sweeping mechanism control technology, and in particular to a sweeping method for unmanned sweepers based on target distribution pattern analysis. Background Technology

[0002] Current unmanned sweeping vehicle cleaning methods still have shortcomings: On the one hand, existing methods mostly sweep according to preset paths, making it difficult to dynamically adjust the sweeping method based on the distribution and type of garbage near the road boundary; on the other hand, they lack differentiated and coordinated control of disc brushes, roller brushes, suction nozzles, and nozzles for different types of garbage such as lightweight waste, gravel, mud, and dust. Furthermore, existing methods lack sufficient analysis of garbage residue and distribution changes after sweeping, making it difficult to identify residual areas in a timely manner and perform targeted re-sweeping, thus affecting sweeping efficiency and cleaning effect. Summary of the Invention

[0003] Therefore, it is necessary for the present invention to provide a sweeping method for unmanned sweepers based on target distribution pattern analysis, so as to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, an unmanned sweeping vehicle cleaning method based on target distribution pattern analysis includes the following steps: Step S1: Establish the cleaning operation area corresponding to the unmanned sweeper, and divide the cleaning operation area into multiple cleaning units; identify the road boundaries of the cleaning units based on the road point cloud data collected by LiDAR; Step S2: Identify garbage targets along the road boundary and determine the garbage cleaning method for the garbage targets; identify obstacles at the road boundary and set the cleaning path; Step S3: Control the unmanned sweeper to sweep the garbage targets along the sweeping path using a garbage sweeping method, and detect the distribution pattern of the targets; calculate the garbage target residue based on the target distribution pattern; Step S4: Set the cleaning priority according to the target amount of garbage residue, and control the unmanned sweeper to perform supplementary cleaning operations according to the cleaning priority, and automatically dump the garbage when the bin is full.

[0005] The beneficial effects of this invention are as follows: 1. By dividing the cleaning operation area into cleaning units and using LiDAR to collect road point cloud data to identify road boundaries, combined with the identification of garbage targets and obstacles along the road boundaries, corresponding cleaning paths are set. This solves the problem that existing unmanned sweepers mainly clean according to preset routes and are difficult to adapt to changes in road boundaries and local traffic environment, thereby improving the unmanned sweeper's path adaptability and edge cleaning effect.

[0006] 2. By identifying lightweight waste, gravel waste, mud and sand-attached waste, and dry dust waste, and by using front disc brushes, central roller brushes, central suction nozzles, and front nozzles to work together for different types of waste, a suction and sweeping mode, a wet sweeping mode, or a water spraying and dust suppression mode is formed. This solves the problem that the existing sweeping methods are relatively fixed and it is difficult to differentiate the treatment of different physical forms of waste, thereby improving the sweeping efficiency and sweeping effect of different types of waste.

[0007] 3. By detecting the distribution pattern of garbage targets during the cleaning process and calculating the garbage target residue based on the target distribution pattern after cleaning, the cleaning priority is set and the unmanned sweeping vehicle is controlled to perform supplementary cleaning operations. This solves the problem that existing unmanned sweeping vehicles are unable to identify local garbage residue in a timely manner after completing a single cleaning and are prone to missing cleaning, thereby improving the targeting of unmanned sweeping vehicle supplementary cleaning operations and the overall cleaning effect. Attached Figure Description

[0008] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a schematic diagram of the steps of an unmanned sweeper cleaning method based on target distribution situation analysis according to the present invention; Figure 2 This is a schematic diagram of the telescopic arm of a front-mounted disc brush according to one embodiment; Figure 3 This is a schematic diagram of a front-mounted disk brush structure according to one embodiment; Figure 4 This is a schematic diagram of an embodiment of an unmanned sweeping vehicle; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0009] The technical method of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0010] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0011] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0012] To achieve the above objectives, please refer to Figures 1 to 4 This invention provides a method for unmanned sweeping vehicle cleaning based on target distribution pattern analysis, the method comprising the following steps: Step S1: Establish the cleaning operation area corresponding to the unmanned sweeper, and divide the cleaning operation area into multiple cleaning units; identify the road boundaries of the cleaning units based on the road point cloud data collected by LiDAR; In one embodiment, after the unmanned sweeping vehicle enters a preset work area, it establishes a corresponding cleaning work area based on integrated navigation and positioning information, and divides the cleaning work area into multiple cleaning units according to road length, road width, and road structure. The unmanned sweeping vehicle uses an onboard LiDAR to scan the current cleaning unit, obtain road point cloud data, and then filters the road surface point cloud from the road point cloud data. The road surface point cloud is segmented according to the road extension direction to form multiple road point cloud segments; the edge positions in each road point cloud segment are identified, the road boundary points on both sides of the road are extracted, and then connected according to the spatial continuity relationship between the road boundary points to form the road boundary corresponding to the cleaning unit. The road boundary may include the edge of the curb, the junction of the road and the green belt, the junction of the road and the building, or other edge positions that restrict the passage of the unmanned sweeping vehicle.

[0013] In another embodiment, assuming the unmanned sweeper's cleaning area is 800m long and 4m wide on average, this area is divided into 8 cleaning units, each 100m long. The vehicle-mounted LiDAR collects road point cloud data at a frequency of 10Hz, with each cleaning unit obtaining approximately 150,000 road point clouds. Taking the third cleaning unit as an example, the road point cloud is divided into 10 segments along the road's extension direction, each segment approximately 10m long. After filtering by ground points, approximately 80 road boundary points are identified on both sides of the road. Adjacent road boundary points are connected according to a spatial distance of less than 0.3m, forming a road boundary area approximately 100m long and 4.2m wide on average.

[0014] Step S2: Identify garbage targets along the road boundary and determine the garbage cleaning method for the garbage targets; identify obstacles at the road boundary and set the cleaning path; In one embodiment, an unmanned sweeping vehicle is controlled to move at a low speed along the road boundary, and continuously acquires images of garbage near the road boundary using a high-definition camera. The identified road boundary is mapped onto the garbage images, using the road boundary as the reference position for image detection, and multiple image detection zones are divided laterally along the road. The road background in the image detection zones is removed, and the remaining areas are used as candidate garbage areas. By analyzing the target texture, target size, and surface morphology of the candidate garbage areas, garbage targets are identified and classified into lightweight garbage, gravel garbage, silt-attached garbage, or dry dust garbage.

[0015] For different types of waste, the corresponding cleaning mechanism and operating mode are activated. For lightweight waste, the extension and rotation position of the front disc brush are adjusted, and it is used in conjunction with the central suction nozzle for sweeping. For gravel waste, the front disc brush is used to push the gravel towards the central roller brush area, and then the central roller brush turns it over and transports it towards the central suction nozzle. For muddy or sandy waste, the road surface is moistened by spraying water, and then the central roller brush is used for sweeping. For dry, dusty waste, the spray direction of the front nozzle is adjusted according to the location of the waste and the dust diffusion, forming a directional water spray to suppress dust.

[0016] Environmental perception devices such as LiDAR and high-definition cameras are used to identify obstacles near road boundaries and analyze the distance between the obstacles and the road boundaries. For areas where the unmanned sweeper cannot pass directly, a detour trajectory is set based on the edge-sweeping trajectory corresponding to the road boundary, and the detour trajectory is connected with the edge-sweeping trajectories before and after the obstacles to form the sweeping path of the unmanned sweeper.

[0017] Step S3: Control the unmanned sweeper to sweep the garbage targets along the sweeping path using a garbage sweeping method, and detect the distribution pattern of the targets; calculate the garbage target residue based on the target distribution pattern; In one embodiment, the unmanned sweeper moves along the cleaning path and controls the front disc brush, central roller brush, central suction nozzle, or front nozzle to perform operations based on the type of garbage corresponding to the current location. After completing the cleaning of one cleaning unit, the cleaned road area is detected again using a high-definition camera and LiDAR to identify any remaining garbage targets and record their location, quantity, coverage area, and aggregation status.

[0018] By comparing the distribution of target waste before cleaning with the residual target waste after cleaning, the target distribution pattern of the current cleaning unit is formed. The quantity residual ratio between the initial quantity of target waste and the area residual ratio between the area covered by residual waste and the area covered by initial waste are calculated. For multiple residual target wastes, the spatial spacing and degree of aggregation between the residual wastes can also be analyzed. Finally, by combining the quantity residual ratio, area residual ratio, and residual aggregation, the waste target residue level is calculated, ensuring that the waste target residue level reflects the actual cleaning effect of the current cleaning unit.

[0019] In another embodiment, suppose a cleaning unit identifies 40 debris targets before cleaning, with a total coverage area of ​​12 m². After the first cleaning, 8 more residual debris targets are detected, with a total residual debris coverage area of ​​2.1 m².

[0020] Of the eight remaining garbage targets, the quantity residual rate was 8 / 40, or 20%; the area residual rate was 2.1 / 12, or 17.5%. The detection revealed that five of the eight remaining garbage targets were concentrated within a 12m radius of the road boundary, while the remaining three were scattered in the central area of ​​the road. The residual rate can be calculated by combining the quantity residual rate (0.4), the area residual rate (0.4), and the residual aggregation degree (0.2). Assuming the normalized residual aggregation degree is 0.75, the garbage target residual rate can be calculated as: 0.4 × 0.20 + 0.4 × 0.175 + 0.2 × 0.75 = 0.30.

[0021] Step S4: Set the cleaning priority according to the target amount of garbage residue, and control the unmanned sweeper to perform supplementary cleaning operations according to the cleaning priority, and automatically dump the garbage when the bin is full.

[0022] In one embodiment, the residual amount of waste corresponding to multiple cleaning units is aggregated, and the cleaning units are sorted from high to low according to the residual amount of waste. For cleaning units with a high residual amount of waste, they are set as priority cleaning areas; for cleaning units with a low residual amount of waste, they are processed after the high-priority areas have been cleaned.

[0023] After receiving a re-sweeping task, the unmanned sweeper replans its route from its current location to the priority sweeping area and calls upon the corresponding sweeping mechanism to perform another sweep based on the location of any remaining trash. During the re-sweeping process, the loading status of the trash can is continuously monitored. When the trash can reaches the preset loading level, the current sweeping operation is paused, and the unmanned sweeper is guided to the preset trash dumping location. The trash can is then unloaded using the vehicle's automatic dumping mechanism, and the remaining sweeping tasks are resumed after the trash has been dumped.

[0024] In another embodiment, assuming that after completing all the initial cleaning, a total of 5 cleaning units are obtained, with target residual amounts of 0.12, 0.30, 0.08, 0.46, and 0.21 respectively. After sorting according to the target residual amount of 0.46, the 4th cleaning unit has a residual amount of 0.46 and is set as a level 1 cleaning area; the 2nd cleaning unit has a residual amount of 0.30 and is set as a level 2 cleaning area; the 5th cleaning unit has a residual amount of 0.21 and is set as a level 3 cleaning area; the 1st and 3rd cleaning units have residual amounts of 0.12 and 0.08 respectively, and can be temporarily not cleaned or set as low-priority cleaning areas.

[0025] Assuming the unmanned sweeper's trash can has a capacity of 240L, after completing the second and fourth cleaning units, the trash volume in the trash can reaches 205L, approximately 85% of the total capacity. At this point, the second cleaning is paused, and the sweeper travels along a preset route to the trash dumping point, where it controls the automatic dumping mechanism to unload the trash. Assuming the remaining trash volume in the trash can after unloading is 5L, the sweeper returns to the fifth cleaning unit to continue the second cleaning operation.

[0026] Preferably, in step S1, the identification of the road boundary of the cleaning unit based on the road point cloud data collected by the lidar includes: Ground points are filtered based on road point cloud data collected by lidar to obtain road ground point cloud; the cleaning unit is then segmented into multiple road point cloud segments based on the road ground point cloud. In one embodiment, after the unmanned sweeper enters the cleaning unit, it activates its onboard LiDAR to continuously scan the road area, collecting road point cloud data including the road surface, curbs, green belts, obstacles, and surrounding facilities. The collected road point cloud data is then filtered according to its height information and local planar features, retaining points located within the road surface area and exhibiting continuous planar distribution characteristics to obtain the road surface point cloud. Then, according to the unmanned sweeper's travel direction, the road surface point cloud is segmented, dividing the continuous road area into multiple adjacent road point cloud segments, so that each road point cloud segment corresponds to a local area on the road.

[0027] In another embodiment, assuming the unmanned sweeping vehicle uses a 32-line LiDAR, scanning a 120m long and 4m wide sweeping unit at a sampling frequency of 10Hz, approximately 1.8 million road point cloud data points are obtained after scanning the sweeping unit. Preliminary screening is performed based on the height information of each point, selecting points whose height difference relative to the road reference plane is within ±0.08m as candidate ground points. Further screening is performed based on local planar continuity, retaining approximately 1.26 million road ground point clouds. The road point cloud is then divided into 12 segments every 10m along the road's extension direction. Taking one road point cloud segment as an example, this segment corresponds to a road area 10m long and approximately 4m wide, containing approximately 105,000 road ground point clouds.

[0028] Road boundary points are identified by segmenting the road point cloud; the boundary points are connected according to their spatial continuity to form the road boundary of the cleaning unit.

[0029] In one embodiment, for each road point cloud segment, the road surface point cloud is scanned laterally along a direction perpendicular to the road's extension direction to identify locations where the road surface point cloud changes from a continuous distribution to a sparse distribution. Road boundary points on both sides of the road are extracted by combining the height changes and spatial intervals between adjacent points. The road boundary points within each road point cloud segment are sorted according to their spatial location, and road boundary points in adjacent road point cloud segments that are close in location and have the same extension direction are associated. The associated road boundary points are sequentially connected along the road's extension direction to form a continuous boundary line on one side of the road; the same method is used to form a continuous boundary line on the other side of the road, and these two continuous boundary lines are used as the road boundaries of the cleaning unit. For areas with localized green belts, curbs, or building edges, the boundary connection direction can be corrected based on the height changes and continuous extension state of the road boundary points to ensure that the formed road boundaries correspond to the actual cleanable area.

[0030] In another embodiment, it is assumed that in the aforementioned 12 road point cloud segments, the boundary points on the left and right sides of the road are identified for each segment. Taking the first road point cloud segment as an example, 92 road boundary points are identified on the left side of the road, and 88 road boundary points are identified on the right side of the road, with an average interval of approximately 0.11m between adjacent road boundary points. Connecting adjacent road boundary points with a spatial interval of less than 0.35m and an angle of less than 12° in the road's extension direction forms the local road boundary of the corresponding road point cloud segment.

[0031] After connecting the road boundary points of the 12 road point cloud segments, it is assumed that a total of 1080 continuous road boundary points are formed on the left side of the road and a total of 1036 continuous road boundary points are formed on the right side of the road. After boundary connection, the length of the continuous boundary on the left side of the road is approximately 119.6m, the length of the continuous boundary on the right side of the road is approximately 119.2m, and the average distance between the two road boundaries is 4.05m.

[0032] Preferably, in step S2, identifying garbage targets along the road boundary and determining the garbage cleaning method for the garbage targets includes: Collect garbage images along the road boundary and use the garbage images to identify garbage targets; classify garbage types according to garbage targets; determine the corresponding cleaning mechanism and operation mode for garbage types. For example, if the garbage type is lightweight garbage, a suction and sweeping mode using the front disc brush of an unmanned sweeper combined with the central suction nozzle is adopted. In one embodiment, as the unmanned sweeping vehicle travels along the road boundary, it continuously acquires images of trash in the road edge area using an onboard high-definition camera, and detects the trash in the images by combining the road boundary position. The road boundary is mapped onto the trash images, and used as a spatial reference for image detection, the road edge area is divided into multiple image detection regions. For each image detection region, background areas with continuous road texture features are removed, and the remaining areas are used as candidate trash regions. Then, target contours, surface textures, color distributions, and size information are extracted from the candidate trash regions, and the corresponding trash targets are identified based on these features.

[0033] After identifying the waste target, it is classified into different types based on its surface roughness, outline shape, size, and distribution. For example, waste with a relatively flat surface, light weight, and irregular flake or strip-shaped outline can be classified as lightweight waste; lightweight waste can include leaves, paper scraps, plastic flakes, and packaging fragments. For lightweight waste, the operating position of the front disc brush is adjusted according to its position relative to the road boundary, ensuring that the front disc brush's sweeping area covers the location of the lightweight waste, and the front disc brush pushes the lightweight waste from the road edge towards the center of the vehicle. The central suction nozzle is then activated, allowing the lightweight waste pushed by the front disc brush to enter the suction range of the central suction nozzle. A negative pressure airflow is created by the fan to suck up the lightweight waste, and the front disc brush and the central suction nozzle work together to form a suction and sweeping mode.

[0034] If the waste type is gravel, a vacuuming and sweeping mode using a front disc brush and a central roller brush in conjunction with a central suction nozzle is used; if the waste type is mud or sand, a wet sweeping mode using a central roller brush and water spray is used; if the waste type is dry dust, a water spraying and dust suppression mode using a front nozzle is used. Vacuuming and sweeping mode, wet sweeping mode, and water spraying and dust suppression mode are used as the waste cleaning methods.

[0035] In one embodiment, when the identified waste target has a high surface roughness, a granular or blocky outline, and its size is within a preset gravel size range, the corresponding waste target is classified as gravel waste. For gravel waste, the front disc brush is controlled to extend towards the road edge, and the dispersed gravel waste is pushed towards the corresponding area of ​​the central roller brush, forming a gravel accumulation area. The central roller brush is activated, and the gravel waste is turned over and collected through contact with the road surface, causing the gravel waste to be transported towards the central suction nozzle. The suction range of the central suction nozzle is adjusted according to the accumulation position of the gravel waste, and the corresponding fan is activated to suck up the gravel waste entering the suction range, so that the front disc brush, central roller brush, and central suction nozzle work together to form a suction and sweeping mode.

[0036] When the garbage is continuously attached to the road surface in a muddy or flaky form, and there is a clear adhesion between the target area and the road surface, it is classified as mud and sand-type attached garbage. For mud and sand-type attached garbage, the water spray device is controlled to spray water onto the attached area, wetting the attached mud and sand on the road surface and reducing its adhesion. Then, the central roller brush is activated, which rolls and scrubs the wetted mud and sand area, causing the attached mud and sand to detach from the road surface and concentrate in the area affected by the roller brush. As the vehicle moves along the road, the central roller brush continues to clean the scrubbing area, creating a wet sweeping mode by combining the water spray and the scrubbing action of the central roller brush.

[0037] When the garbage is in the form of fine particles, dry and dispersed, and easily generates dust under sweeping action, it is classified as dry dust garbage. For dry dust garbage, first identify the target area where the dry dust garbage is located, and adjust the spray direction of the front nozzles according to the vehicle's driving direction and the distribution of the garbage. After activating the front nozzles, the sprayed water mist covers the area where the dry dust garbage is located, pre-wetting the fine particles on the road surface and reducing the degree of particle stirring caused by vehicle movement or sweeping operations. The spray range of the front nozzles is continuously adjusted according to the driving status of the unmanned sweeper, so that the front nozzles form a water spraying and dust suppression mode corresponding to the road edge.

[0038] Most importantly, if the waste type is silt or sand-like attached waste, the wet sweeping mode using a centrally mounted roller brush combined with water spraying includes: If the type of waste is mud and sand, then a cleaning area is defined; control the front nozzle to spray water mist onto the cleaning area to moisten the mud and sand, thus creating a moist cleaning area. In one embodiment, when the identified waste type is silt-like attached waste, a corresponding attachment cleaning area is defined based on the distribution and outline of the silt-like attached waste on the road surface. An unmanned sweeper is controlled to travel to the vicinity of the attachment cleaning area, and the front nozzle is controlled to spray water mist towards the attachment cleaning area. Specifically, the water spray coverage range of the front nozzle is determined based on the distribution area of ​​the silt-like attached waste, and the water spray direction, spray angle, and water volume per unit time of the front nozzle are adjusted to ensure that the water mist evenly covers the area where the silt-like attached waste is located. During the water spraying process, the water mist gradually penetrates into the bonding area between the silt and the road surface to reduce the adhesion strength between the silt-like attached waste and the road surface, transforming the originally dry or semi-dry attached waste into a moist state that is easy to sweep, thereby forming a moist cleaning area.

[0039] The central roller brush is controlled to brush and sweep the mud and sand attached garbage in the wet cleaning area to form a mud and sand stripping state; continuous brushing is carried out in the mud and sand stripping state, and the water spraying operation of the front nozzle and the brushing operation of the central roller brush form a wet sweeping mode.

[0040] In one embodiment, after a wet cleaning area is formed, the central roller brush is controlled to move to the position corresponding to the wet cleaning area and is driven to rotate along the road surface, so that the brush bristles come into contact with the wetted mud and sand debris. Through the rotational friction and brushing pressure of the central roller brush, the mud and sand attached to the road surface are repeatedly disturbed, causing the mud and sand debris to gradually detach from the road surface. When the mud and sand in the wet cleaning area are detected to change from a continuously attached state to a partially loose, separated, or movable state, a mud and sand stripping state is confirmed. In the mud and sand stripping state, the central roller brush continues to perform brushing operations, and the front nozzle is controlled to spray water continuously or intermittently according to the wetness of the wet cleaning area, so that the mud and sand debris maintains a predetermined level of moisture during the brushing process. The water spraying operation of the front nozzle and the brushing operation of the central roller brush work together to form a wet sweeping mode for mud and sand attached garbage. The water spraying operation is used to continuously reduce the adhesion strength of mud and sand and suppress the dust generated during the brushing and sweeping process, while the brushing and sweeping operation is used to separate the wet mud and sand from the road surface.

[0041] Most importantly, if the waste type is gravel, the vacuuming mode using a front disc brush and a central roller brush in conjunction with a central suction nozzle includes: If the type of waste is gravel, control the front disc brush to move the gravel from the road boundary to the corresponding area of ​​the central roller brush to form a gravel accumulation area. In one embodiment, visual and distance sensors located at the front of the sweeping vehicle identify debris near the road boundary. When the identification results indicate that the particle size, shape, and accumulation state of the debris match the characteristics of gravel, the gravel distribution area on one side of the road boundary is determined. Based on the relative position between the gravel distribution area and the central roller brush, the rotation direction and swing angle of the front disc brush are adjusted, causing the outer edge of the front disc brush to move from the road boundary side towards the center of the vehicle, gradually conveying the dispersed gravel to the working area corresponding to the central roller brush. When the interval between multiple gravel distribution points is less than a preset aggregation distance, the gravel corresponding to adjacent gravel distribution points is concentrated to form a gravel aggregation area, and the length, width, and gravel distribution density of the gravel aggregation area are recorded.

[0042] The central roller brush is controlled to turn over the gravel and debris in the gravel accumulation area and transport the turned gravel and debris towards the central suction nozzle to form a gravel conveying area. The suction range of the central suction nozzle is determined according to the gravel conveying area, and the central suction nozzle is controlled to suck up the gravel and debris within the suction range to form a suction and sweeping mode.

[0043] In one embodiment, the central roller brush is located behind the front disc brush. By adjusting the rotation direction of the central roller brush, the part of the roller brush in contact with the ground forms a conveying trend from front to back and from the road boundary side towards the central suction nozzle. When the central roller brush enters the gravel accumulation area, it turns over the gravel waste with its bristles, causing the gravel accumulated on the ground to roll and shift, and move towards the central suction nozzle under the continuous rotation of the roller brush. The gravel conveying area is determined based on the distribution range of the turned gravel, and the width, length, and distance between the gravel conveying area and the central suction nozzle are obtained. When the gravel conveying area is within the effective suction range of the central suction nozzle, the central suction nozzle is controlled to start negative pressure suction, allowing the gravel waste to enter the collection channel through the suction nozzle. The front disc brush, central roller brush, and central suction nozzle work together in the sequence of conveying, turning, and suction to form a suction and sweeping mode for gravel waste.

[0044] Preferably, collecting garbage images along the road boundary and using the garbage images to identify garbage targets includes: The system controls the unmanned sweeping vehicle to move along the road boundary and continuously collects images of garbage using a high-definition camera; for the garbage images, the road boundary is mapped onto the garbage images to determine the road reference position; In one embodiment, the unmanned sweeping vehicle travels along pre-identified road boundaries, continuously capturing images of the road edge area using an onboard high-definition camera. The high-definition camera can be mounted at the front or front side of the vehicle, with its acquisition direction facing the cleaning area corresponding to the front disc brush, ensuring the acquisition range covers the road boundary and the area to be cleaned extending inwards from the road boundary. For the continuously acquired garbage images, based on the road boundary spatial coordinates identified by the LiDAR and the installation position relationship between the high-definition camera and the LiDAR, the road boundary is projected onto the corresponding garbage image.

[0045] Specifically, multiple road boundary points on the road edge can be transformed into the coordinate system of a high-definition camera and mapped to the pixel coordinates of the garbage image through camera projection. The mapped road boundary pixels are then connected according to their spatial positions to form the road boundary line in the image, which serves as the road reference position. This road reference position represents the actual location of the road edge in the garbage image, allowing the garbage identification area to extend inwards along the road boundary, thereby reducing the impact of green areas, building areas, or other non-sweeped areas on garbage target identification.

[0046] In another embodiment, assuming the unmanned sweeper moves along the road boundary at a speed of 3 km / h, a high-definition camera continuously captures garbage images at 25 frames / s, with a camera resolution of 1920×1080 pixels. After 20 seconds of continuous driving, a total of 500 garbage images are captured. Assuming the actual length of the road boundary in the vehicle's direction of travel is 16.7m, 60 road boundary points are selected for spatial mapping, of which 58 road boundary points are successfully mapped to their corresponding garbage images.

[0047] Taking one frame of a garbage image as an example, the lateral pixel coordinates of the 58 road boundary pixels are mainly distributed within the range of 180 to 260 pixels, while the vertical pixel coordinates are distributed within the range of 420 to 1010 pixels. After connecting adjacent road boundary pixels according to their spatial order, a continuous road boundary line is formed, and this road boundary line is used as the road reference position for that frame of the garbage image. As the unmanned sweeper continues to move, the average lateral offset of the road reference position in 20 consecutive frames is 6.8 pixels, thus reflecting the continuous change in the road edge position as the vehicle travels along the road boundary.

[0048] Starting from the road reference position, the garbage image is divided into multiple image detection zones; road background features in the image detection zones are removed, and the rest are candidate garbage regions; the target texture and target size are determined based on the candidate garbage regions; and garbage targets are identified based on the target texture and target size.

[0049] In one embodiment, starting from a road reference position, the garbage image is divided into multiple adjacent image detection zones extending inward along a direction perpendicular to the road boundary. Each image detection zone extends along the road boundary and corresponds to a road area within a different distance range. By using the road reference position as the starting point for dividing the image detection zones, the image detection area can be adjusted synchronously as the position of the road boundary changes, ensuring that the image detection zones correspond to the road areas that the unmanned sweeper actually needs to clean.

[0050] For each image detection band, the color distribution, texture continuity, and local grayscale changes in the image are analyzed to identify image regions with road background characteristics. Road backgrounds typically exhibit large-area continuous distribution, relatively gentle color changes, and high texture repetition. Pixel regions matching these road background characteristics are removed from the corresponding image detection band. Locally discontinuous regions remaining after removing the road background are considered candidate garbage regions.

[0051] Then, target texture and target size are extracted for the candidate waste areas. Target texture can include local grayscale variations, edge density, texture direction, and color variation features; target size can be obtained from the pixel width, pixel length, and occupied area of ​​the candidate waste area, and converted into an actual size range by combining the imaging parameters of the high-definition camera. Based on the target texture and target size, the candidate waste areas are classified and identified. For example, candidate areas with sheet-like textures, irregular edges, and small sizes can be identified as lightweight waste; candidate areas with granular or blocky textures and relatively distinct local edges can be identified as gravel waste; candidate areas with continuous attached textures and a significant color difference from the road background can be identified as silt-like attached waste; and candidate areas with fine granular textures and a dispersed distribution can be identified as dry, dusty waste.

[0052] Preferably, waste types are classified according to waste targets, including: Calculate the surface roughness of the target based on the waste target; determine the target size level based on the waste target; when the surface roughness of the target is less than the first roughness threshold and the target size level is small, classify the waste target as lightweight waste; In one embodiment, assuming the unmanned sweeper continuously identifies 10 garbage targets along the road boundary, the calculated surface roughness of each garbage target is 0.18, 0.23, 0.31, 0.62, 0.74, 0.27, 0.35, 0.81, 0.29, and 0.46, respectively; a first roughness threshold of 0.35 is set, and the maximum length and maximum width of the small-sized target are 150mm and 100mm, respectively. Specifically, the dimensions of the first target are 80mm × 45mm, the second target is 120mm × 70mm, the sixth target is 95mm × 60mm, the seventh target is 135mm × 85mm, and the ninth target is 110mm × 55mm. Since the surface roughness of all these targets is less than 0.35 and their dimensions are all within the small-sized range, the first, second, sixth, seventh, and ninth garbage targets can be classified as lightweight garbage.

[0053] When the surface roughness of the target is greater than the second roughness threshold and the target size is medium, the waste target is classified as gravel waste; when the texture of the waste target is continuously muddy and the target size is medium, the waste target is classified as silt-attached waste. In one embodiment, assuming the second roughness threshold is set to 0.60, the target length range corresponding to the medium size is 150mm to 500mm. Five consecutive waste targets are detected: the first target has a surface roughness of 0.68 and a size of 260mm × 180mm; the second target has a surface roughness of 0.76 and a size of 320mm × 210mm; the third target has a surface roughness of 0.41 and a size of 280mm × 190mm; the fourth target has a surface roughness of 0.72 and a size of 540mm × 260mm; and the fifth target has a surface roughness of 0.33 and a size of 350mm × 220mm. Since the surface roughness of the first and second targets is greater than 0.60 and their sizes are within the medium size range, the first and second targets are classified as gravel waste. Assuming that the third and fifth targets both exhibit continuous mud-like textures, with the mud-like area of ​​the third target being 0.052 m² and the mud-like area of ​​the fifth target being 0.087 m², and both being within the medium size range, the third and fifth targets are classified as silt-like attached waste. The fourth target, due to its size exceeding the medium size range, is not classified according to the above conditions.

[0054] When the texture of the waste target is dry and granular and the target size is small, the waste target is classified as dry dust waste.

[0055] In one embodiment, assuming there are six unclassified waste targets within the detection area, the number of particles for each target is 42, 58, 15, 76, 31, and 63, respectively, with corresponding average particle spacing of 4.2 mm, 3.8 mm, 12.6 mm, 3.1 mm, 8.5 mm, and 4.0 mm. Targets 1, 2, 4, and 6 all exhibit a dispersed distribution of multiple independent fine particles, with target sizes of 90 mm × 65 mm, 120 mm × 80 mm, 105 mm × 75 mm, and 140 mm × 90 mm, respectively, all within a small size range. Therefore, targets 1, 2, 4, and 6 are classified as dry dust waste. Target 3, although small in size, has a low particle count and large particle spacing, failing to meet the dry granular distribution characteristics. Target 5 exhibits a locally continuous coverage texture and is therefore not classified as dry dust waste. Through this method, the unmanned sweeper can classify waste targets according to their surface roughness, size class, and texture distribution characteristics.

[0056] Preferably, if the type of waste is lightweight waste, the sweeping mode using the front disc brush of the unmanned sweeper combined with the central suction nozzle includes: If the type of waste is light waste, calculate the extension distance of the front disc brush based on the distance between the light waste and the road boundary; determine the working position of the front disc brush based on the extension distance of the front disc brush. In one embodiment, after the unmanned sweeper identifies lightweight waste along the road boundary, it uses the road boundary as a lateral position reference and calculates the lateral distance between the center of the lightweight waste and the road boundary. Based on the lateral distance and the initial retracted position of the front disc brush, it calculates the required lateral displacement of the front disc brush and converts this lateral displacement into the extension distance of the telescopic sweeping arm. Specifically, the front disc brush is connected to the vehicle body via a telescopic sweeping arm, which houses a telescopic drive mechanism. This telescopic drive mechanism can be an electric push rod, a hydraulic cylinder, or a linear drive motor. The controller sends a position adjustment command to the telescopic drive mechanism based on the calculated telescopic distance, driving the front disc brush to move from its initial position towards the road boundary. A position sensor mounted on the telescopic sweeping arm detects the actual extension position of the disc brush. When the actual extension distance corresponds to the target telescopic distance, the telescopic drive mechanism stops, ensuring that the sweeping range of the front disc brush covers the area containing the lightweight waste, and the current position is designated as the operating position of the front disc brush.

[0057] The rotation speed of the front disc brush is determined based on the distribution area of ​​the lightweight waste; the suction duration of the central suction nozzle is determined based on the quantity of lightweight waste; and the suction and sweeping mode is set based on the working position, rotation speed, and suction duration of the front disc brush.

[0058] In one embodiment, a regional statistical analysis is performed on lightweight waste near the road boundary to calculate the road surface area occupied by the lightweight waste, and the disk brush operation load is divided according to the distribution area of ​​the lightweight waste. When the distribution area of ​​lightweight waste is small, the front disk brush is kept at a low speed by the brush motor controller; as the distribution area of ​​lightweight waste increases, the output speed of the brush motor is increased, so that the disk brush forms a larger sweeping frequency per unit time. Specifically, the distribution area of ​​lightweight waste can be divided into multiple area intervals, each corresponding to a different brush motor speed setting. The brush motor drives the front disk brush to rotate, so as to push the lightweight waste near the road boundary to the suction area corresponding to the central suction nozzle.

[0059] The system counts the amount of lightweight waste in the current cleaning area and adjusts the continuous operating time of the central suction nozzle based on the amount of waste. When the amount of lightweight waste is small, the suction system maintains suction for a shorter time after the disc brush completes one sweep; when the amount of lightweight waste increases, the continuous working time of the central suction nozzle is extended, allowing the lightweight waste conveyed by the front disc brush to be continuously sucked into the waste collection bin. The controller combines the working position of the front disc brush, the output speed of the brush motor, and the continuous suction time of the central suction nozzle into corresponding equipment operating parameters, and controls the front disc brush and the central suction nozzle to work together according to the equipment operating parameters to form a suction and sweeping mode suitable for lightweight waste.

[0060] Preferably, determining the extension distance of the front disc brush based on the distance between the lightweight waste and the road boundary includes: The lateral offset of the lightweight waste is calculated based on the distance between the lightweight waste and the road boundary; the lateral offset is compared with the preset initial operating position of the front disc brush to calculate the disc brush position compensation. In one embodiment, a lateral reference line is determined based on the position of the road boundary in the coordinate system of the unmanned sweeper, and the lateral coordinates of the center position of the lightweight waste relative to the lateral reference line are obtained. The lateral distance between the center position of the lightweight waste and the road boundary is taken as the lateral offset of the lightweight waste, denoted as L1. The lateral position of the front disc brush in the initial working state is preset, denoted as L0, and the position difference between the lateral offset and the initial working position of the front disc brush is calculated to obtain the disc brush position compensation amount ΔL, i.e., ΔL = L1 - L0. When the lightweight waste is located outside the initial working position of the front disc brush, the front disc brush is controlled to extend towards the road boundary according to the position difference; when the lightweight waste is located inside the initial working position of the front disc brush, the front disc brush is controlled to retract towards the center of the vehicle according to the position difference. Thus, the front disc brush can adjust its lateral position according to the actual distribution position of the lightweight waste relative to the road boundary.

[0061] Adjust the maximum telescopic arm length of the front disc brush according to the disc brush position compensation amount, use the maximum telescopic arm length to identify the disc brush sweeping range, stop the telescopic arm when the disc brush sweeping range covers the location of light waste, and record the telescopic distance of the front disc brush.

[0062] In one embodiment, the target lateral position of the front disc brush is determined based on the disc brush position compensation amount, and the telescopic sweeping arm is gradually extended or retracted according to the target lateral position. The front disc brush corresponds to different sweeping radii at different extension lengths, and the sweeping range is determined based on the rotation radius of the front disc brush, the center position of the disc brush, and the maximum telescopic sweeping arm length. When the sweeping range formed during the telescopic sweeping arm adjustment has not yet covered the location of the lightweight waste, the length of the telescopic sweeping arm continues to be adjusted; when the outer boundary of the sweeping range reaches or exceeds the location of the lightweight waste, the telescopic sweeping arm is stopped, and the current telescopic sweeping arm length is determined as the actual extension distance of the front disc brush. A certain coverage margin can be reserved outside the location of the lightweight waste, so that the sweeping range of the disc brush can not only reach the center of the waste but also cover the surrounding area, thus avoiding waste leakage due to vehicle deviation.

[0063] In another embodiment, assuming the initial telescopic sweeping arm length of the front disc brush is 0.45m, the maximum telescopic length is 0.90m, and the effective sweeping radius of the disc brush itself is 0.32m, the required telescopic sweeping arm lengths are calculated to be 0.70m, 0.57m, 0.48m, 0.45m, 0.52m, and 0.65m respectively after adjusting the telescopic arm length according to the corresponding disc brush position compensation for the six identified lightweight wastes. The corresponding lateral sweeping ranges of the disc brush can cover road areas from the road boundary of 0.38m to 1.02m, 0.25m to 0.89m, 0.16m to 0.80m, 0.13m to 0.77m, 0.20m to 0.84m, and 0.33m to 0.97m respectively. When the center of each lightweight waste is within the corresponding disc brush sweeping range, the telescopic arm is stopped, and the actual telescopic distance of the front disc brush is recorded. Therefore, the front-mounted disc brush can extend and retract at different distances for lightweight waste in different locations, so that the sweeping range of the disc brush corresponds to the distribution location of the lightweight waste.

[0064] Preferably, determining the rotation speed of the front disc brush based on the distribution area of ​​lightweight waste includes: The coverage area corresponding to the lightweight waste is determined based on the distribution area of ​​the lightweight waste; the single operation area of ​​the front disc brush is calculated based on the coverage area; the number of times the front disc brush is used for coverage is calculated based on the coverage area and the single operation area. In one embodiment, the edge contours of each lightweight waste target are determined based on the lightweight waste targets identified in the road image. Adjacent waste targets with a spacing less than a preset connection distance are connected to form a continuous distribution area corresponding to the lightweight waste. The coverage range of the lightweight waste is determined based on the outer boundary of the continuous distribution area, and the actual area corresponding to the coverage range is calculated, denoted as S1. The single-operation area of ​​the front disc brush is determined based on the effective sweeping width of the front disc brush and the forward distance of the unmanned sweeper during a single cleaning process, denoted as S2. S2 can be determined based on the product of the effective sweeping width and the single forward distance. The coverage range of the lightweight waste is compared with the single-operation area of ​​the front disc brush to determine the number of coverage times N required to complete the corresponding coverage range, i.e., N = S1 / S2. When the calculation result has a decimal, the number of coverage times is rounded up to ensure that the actual operating range of the front disc brush can cover the entire area where the lightweight waste is located. The number of repeated sweeping operations of the front disc brush in the corresponding road area is determined based on the number of coverage times.

[0065] The sweeping load of light waste is assessed based on the number of coverages; a drive adjustment command for the front disc brush is generated based on the sweeping load; the drive adjustment command is sent to the sweeping motor of the front disc brush; the output speed of the sweeping motor is adjusted by the motor controller of the sweeping motor; and the rotation speed of the front disc brush is determined by the output speed of the sweeping motor.

[0066] In one embodiment, it is assumed that work areas covered 1 to 2 times are defined as low cleaning load, work areas covered 3 to 4 times are defined as medium cleaning load, and work areas covered 5 times or more are defined as high cleaning load. Assuming the rated rotation speed of the front disc brush is 240 r / min, for low cleaning load areas, the motor controller adjusts the output speed of the brush motor to the corresponding front disc brush rotation speed of 150 r / min according to the drive adjustment command; for medium cleaning load areas covered 3 and 4 times, the front disc brush rotation speed is adjusted to 190 r / min and 220 r / min respectively; when a certain area of ​​lightweight waste requires 6 covers according to calculation, it is defined as high cleaning load, and the front disc brush rotation speed is adjusted to 240 r / min. Assuming the unmanned sweeper passes through four areas covered by lightweight waste in sequence, with corresponding coverage times of 2, 4, 3, and 6 times respectively, the output speed of the sweeping brush motor is adjusted sequentially according to the corresponding drive adjustment commands, so that the rotation speed of the front disc brush is 150 r / min, 220 r / min, 190 r / min, and 240 r / min respectively. Thus, the disc brush rotation speed is differentiated according to the distribution area of ​​lightweight waste and the cleaning needs in different road areas.

[0067] Preferably, if the waste type is dry waste with dust, the dust suppression mode using front-mounted nozzles includes: If the waste type is dry waste with dust, then determine the dust target area; determine the dust diffusion direction based on the dust target area; determine the water spray direction of the front nozzle based on the dust diffusion direction; In one embodiment, when the road debris identification result indicates that the debris type is dry dust, the dust target area is determined based on the location of the debris target in the road image and the surrounding particulate distribution range. Multiple consecutively acquired road images are compared to identify the changing direction of the particulate distribution boundary in the dust target area, and the main dust diffusion direction is determined in conjunction with the driving direction of the unmanned sweeper. The target water spray direction of the front nozzle is determined based on the dust diffusion direction, so that the water mist sprayed from the front nozzle is directed towards the dust diffusion area. When the dust mainly diffuses outwards from the road, the front nozzle is controlled to deflect outwards from the road; when the dust mainly diffuses along the vehicle's driving direction, the front nozzle is controlled to spray water in front of the vehicle. Thus, the water spray direction of the front nozzle is adjusted according to the actual dust diffusion state, so that the water mist can cover the dust target area and its main diffusion area.

[0068] The front nozzles are controlled to spray water directionally onto the dust target area according to the water spray direction, forming a dust suppression coverage area; the water spray state of the front nozzles is adjusted according to the dust suppression coverage area to form a water spraying dust suppression mode.

[0069] In one embodiment, assuming the effective spray distance of the front nozzle is 2.5m, the spray angle of a single nozzle is 40°, and the initial water flow rate is 6L / min, when the area of ​​the dust target area is 1.8m², the dust suppression coverage area formed according to the water spray direction is 1.4m², which does not completely cover the dust target area. In this case, the spray angle of the front nozzle is adjusted to 55°, expanding the dust suppression coverage area to 2.1m², and continuous water spraying is used for dust suppression. Assuming another dust target area is 0.6m², and the corresponding dust suppression coverage area is 0.9m², the water flow rate is adjusted to 4L / min, and an intermittent water spraying state of 3 seconds followed by 2 seconds of pause is adopted. Assuming three dust-generating target areas are detected consecutively, with areas of 0.5m², 1.3m², and 2.4m² respectively, corresponding dust suppression coverage areas of 0.8m², 1.6m², and 2.8m² will be formed. These areas will be sprayed with intermittent water, continuous water at a medium flow rate, and continuous water at an increased spray angle, respectively. Therefore, the spray pattern of the front nozzles can be dynamically adjusted according to the dust suppression coverage of different dust-generating target areas, forming a water spraying dust suppression mode adapted to different dust distribution ranges.

[0070] Preferably, step S2, which involves identifying obstacles at the road boundary and setting the cleaning path, includes: Collect road environment data along the road boundary, identify obstacles to traffic based on the road environment data; determine the location of obstacles based on the obstacles; and determine the area of ​​influence of obstacles centered on the location of the obstacles. In one embodiment, assuming the unmanned sweeping vehicle moves along a road boundary of 50m in length, three obstacles are continuously identified in the road environment data. The first obstacle is 0.35m from the road boundary, with dimensions of 0.60m long and 0.40m wide; the second obstacle is 0.80m from the road boundary, with dimensions of 1.20m long and 0.50m wide; and the third obstacle is 0.45m from the road boundary, with dimensions of 0.75m long and 0.60m wide. Setting a safe obstacle avoidance distance of 0.30m, the outer contours of each of the three obstacles are extended outward by 0.30m to obtain their respective obstacle influence areas. At this point, the lateral coverage of the first obstacle's influence area is approximately 0.70m, the second obstacle's influence area is approximately 1.10m, and the third obstacle's influence area is approximately 1.20m. The locations of multiple obstacles and their influence areas are then marked.

[0071] Determine the passable area based on the distance between the area affected by the obstacle and the road boundary; set the vehicle driving direction in the passable area and determine the detour trajectory along the vehicle driving direction; connect the detour trajectory with the edge cleaning trajectory corresponding to the road boundary to form a cleaning path.

[0072] In one embodiment, assuming the interval between the road boundary and the obstacle-affected area corresponding to the trash can is 0.75m, which is less than the minimum passage width of 1.40m corresponding to the 1.2m width of the unmanned sweeper and the safety margin, the unmanned sweeper cannot pass directly between the road boundary and the trash can. The 2.2m wide road area on the side away from the road boundary is defined as the passable area. Assuming the unmanned sweeper is currently traveling along the road boundary at a speed of 0.8m / s, it begins to deviate from the edge-sweeping trajectory 1.5m from the front of the trash can's affected area and moves laterally 0.9m towards the inside of the road to form an entry trajectory; it then bypasses the 1.2m long obstacle-affected area along its original direction of travel, forming a detour trajectory; after the vehicle passes the obstacle-affected area, it gradually moves 0.9m towards the road boundary within a 1.5m return distance, re-entering the edge-sweeping trajectory. Assuming the edge-sweeping trajectory extends 20m before and 25m before and after the obstacle, the 20m initial edge-sweeping trajectory, the approximately 4.2m detour trajectory, and the 25m subsequent edge-sweeping trajectory are connected to form a complete sweeping path. Using this method, when obstacles exist at the road boundary, the unmanned sweeper can automatically adjust its trajectory according to the obstacle's influence range, improving the path continuity, obstacle avoidance safety, and road boundary sweeping coverage.

[0073] It should be noted that you should refer to [link / reference]. Figure 2 The front of the vehicle is equipped with a millimeter-wave radar for all-weather distance measurement, speed measurement, and target tracking safety perception; the left and right sides are equipped with left and right side brushes respectively for cleaning road debris along the edges; the middle of the vehicle is equipped with a central roller brush to gather debris towards the suction inlet; and the right side is equipped with a telescopic arm mechanism to control the extension and retraction of the side brushes, thereby adapting to cleaning operations at different road edge positions.

[0074] It should be noted that you should refer to [link / reference]. Figure 3 The diagram shows the side brush drive motor, telescopic mechanism, passive anti-collision mechanism, side brush disc, side brush bristles, and water mist nozzles. The drive motor provides rotational power for the side brush, the telescopic mechanism adjusts the working range of the side brush, the passive anti-collision mechanism is used to reduce collision damage, the water mist nozzles are used to spray water to suppress dust during the sweeping process, and the side brush bristles push the garbage from the road edge toward the vehicle's sweeping area by rotating.

[0075] It should be noted that you should refer to [link / reference]. Figure 4 The vehicle is equipped with components such as a main LiDAR, a front-facing camera, a remote driving camera, a combined navigation device, a panoramic camera, a millimeter-wave radar, and a domain controller. The main LiDAR is used to perceive the road environment and obstacles around the vehicle, the combined navigation device is used to determine the vehicle's position and driving posture, multiple cameras are used to acquire image information of the road in front of and around the vehicle, and the domain controller processes various sensor data to support the unmanned sweeper to achieve environmental perception, positioning and navigation, path planning, and autonomous cleaning operations.

[0076] It's important to clarify the difference between millimeter-wave radar and lidar. LiDAR and millimeter-wave radar (mmWaveRadar) are two mainstream sensing sensors used in autonomous driving, robot navigation, and intelligent security. Their core differences lie in their working principles, detection performance, environmental adaptability, cost, and application scenarios. I. Working Principle and Signal Characteristics: LiDAR: Emits a laser beam (light wave, extremely short wavelength, typically 905nm or 1550nm), calculates target distance by measuring the laser's "time of flight (ToF)" or phase difference, and generates a high-density 3D point cloud through scanning (mechanical or solid-state). Millimeter-wave radar: Emits millimeter waves (electromagnetic waves, wavelength 1-10mm, common frequency bands are 24GHz, 77GHz, 79GHz), calculates distance by measuring the echo time difference of the electromagnetic wave, calculates target velocity through the Doppler effect, and calculates angle through an antenna array. II. Comparison of Detection Performance: Accuracy and Resolution: LiDAR boasts extremely high accuracy (centimeter / millimeter level), extremely high angular resolution (0.1° level), and high point cloud density, enabling precise reconstruction of the three-dimensional outline of objects. Millimeter-wave radar exhibits relatively moderate accuracy (decimeter / meter level), lower angular resolution (1°-5°), and sparse point clouds, making it difficult to distinguish object details (e.g., difficulty in differentiating pedestrians from stationary obstacles). Velocity Measurement: LiDAR velocity measurement requires multi-frame point cloud computing, resulting in relatively limited accuracy. Millimeter-wave radar offers extremely high velocity measurement accuracy (directly measuring radial velocity through Doppler frequency shift), demonstrating a significant advantage in tracking dynamic targets. Detection Range: The effective detection range of LiDAR is typically 150-250 meters (some long-range models can reach over 300 meters), with point cloud density decreasing at longer distances. Millimeter-wave radar performs well at medium to long distances (up to over 350 meters), exhibiting greater stability in long-range detection. III. Environmental Adaptability: LiDAR is significantly affected by adverse weather conditions. Rain, snow, fog, and dust can scatter or attenuate laser light, leading to shorter detection range and increased point cloud noise; strong light (such as backlight) may interfere with the receiver. However, its nighttime performance is excellent (actively emitting light, not dependent on ambient light). Millimeter-wave radar: operates in all weather conditions and has strong anti-interference capabilities. Millimeter waves have strong penetration capabilities through rain, fog, snow, and dust, are minimally affected by severe weather, and have stable nighttime performance; however, they reflect less light from non-metallic objects (such as plastic and fabric), making them prone to missed detections, and are susceptible to multipath interference in complex metallic environments (such as tunnels). IV. Cost and Computing Power Requirements LiDAR: relatively expensive (typically several hundred to several thousand US dollars per unit, with high-line-count models being even more expensive), and has a large point cloud data volume (MB level / frame), requiring high processor computing power. Millimeter-wave radar: relatively inexpensive (typically tens to hundreds of US dollars per unit), has a small point cloud data volume (KB level / frame), requires less processor computing power, and is easy to mass-produce. V. Typical Application Scenarios LiDAR: Suitable for scenarios with high requirements for high-precision 3D modeling and fine obstacle recognition, such as: environmental perception and obstacle avoidance in advanced autonomous driving (L3 and above), high-precision map surveying, robot navigation, and city-level 3D scanning.Millimeter-wave radar: Suitable for scenarios requiring all-weather detection, long-range speed measurement, and cost sensitivity, such as adaptive cruise control (ACC), automatic emergency braking (AEB), blind spot detection (BSD), intelligent traffic speed measurement, and perimeter security. In summary, the two are not "substitutes" but rather "complementary." In practical applications (such as advanced autonomous driving), a multi-sensor fusion solution of "LiDAR + millimeter-wave radar + camera" is typically used. LiDAR provides a high-precision 3D environment model, while millimeter-wave radar provides all-weather speed measurement and distance redundancy, achieving complementary advantages.

[0077] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A sweeping method for unmanned sweepers based on target distribution pattern analysis, characterized in that, Includes the following steps: Step S1: Establish the cleaning operation area corresponding to the unmanned sweeper, and divide the cleaning operation area into multiple cleaning units; identify the road boundaries of the cleaning units based on the road point cloud data collected by LiDAR; Step S2: Identify garbage targets along the road boundary and determine the garbage cleaning method for the garbage targets; identify obstacles at the road boundary and set the cleaning path; Step S3: Control the unmanned sweeper to sweep the garbage targets along the sweeping path using a garbage sweeping method, and detect the distribution pattern of the targets; calculate the garbage target residue based on the target distribution pattern; Step S4: Set the cleaning priority according to the target amount of garbage residue, and control the unmanned sweeper to perform supplementary cleaning operations according to the cleaning priority, and automatically dump the garbage when the bin is full.

2. The unmanned sweeping vehicle cleaning method based on target distribution situation analysis according to claim 1, characterized in that, In step S1, the road boundary of the cleaning unit is identified based on the road point cloud data collected by the lidar, including: Ground points are filtered based on road point cloud data collected by lidar to obtain road ground point cloud; the cleaning unit is then segmented into multiple road point cloud segments based on the road ground point cloud. Road boundary points are identified by segmenting the road point cloud; the boundary points are connected according to their spatial continuity to form the road boundary of the cleaning unit.

3. The unmanned sweeping vehicle cleaning method based on target distribution situation analysis according to claim 1, characterized in that, Step S2 involves identifying garbage targets along the road boundary and determining the garbage collection method based on these targets, including: Collect garbage images along the road boundary and use the garbage images to identify garbage targets; classify garbage types according to garbage targets; determine the corresponding cleaning mechanism and operation mode for garbage types. For example, if the garbage type is lightweight garbage, a suction and sweeping mode using the front disc brush of an unmanned sweeper combined with the central suction nozzle is adopted. If the waste type is gravel, a vacuuming and sweeping mode using a front disc brush and a central roller brush in conjunction with a central suction nozzle is used; if the waste type is mud or sand, a wet sweeping mode using a central roller brush and water spray is used; if the waste type is dry dust, a water spraying and dust suppression mode using a front nozzle is used. Vacuuming and sweeping mode, wet sweeping mode, and water spraying and dust suppression mode are used as the waste cleaning methods.

4. The unmanned sweeping vehicle cleaning method based on target distribution situation analysis according to claim 3, characterized in that, Collecting images of litter along road boundaries and using these images to identify litter targets includes: The system controls the unmanned sweeping vehicle to move along the road boundary and continuously collects images of garbage using a high-definition camera; for the garbage images, the road boundary is mapped onto the garbage images to determine the road reference position; Starting from the road reference position, the garbage image is divided into multiple image detection zones; road background features in the image detection zones are removed, and the rest are candidate garbage regions; the target texture and target size are determined based on the candidate garbage regions; and garbage targets are identified based on the target texture and target size.

5. The unmanned sweeping vehicle cleaning method based on target distribution situation analysis according to claim 3, characterized in that, Waste types are categorized according to their destination, including: Calculate the surface roughness of the target based on the waste target; determine the target size level based on the waste target; when the surface roughness of the target is less than the first roughness threshold and the target size level is small, classify the waste target as lightweight waste; When the surface roughness of the target is greater than the second roughness threshold and the target size is medium, the waste target is classified as gravel waste; when the texture of the waste target is continuously muddy and the target size is medium, the waste target is classified as silt-attached waste. When the texture of the waste target is dry and granular and the target size is small, the waste target is classified as dry dust waste.

6. The unmanned sweeping vehicle cleaning method based on target distribution situation analysis according to claim 3, characterized in that, If the waste type is lightweight waste, the sweeping and vacuuming modes using the front disc brush of the unmanned sweeper combined with the center suction nozzle include: If the type of waste is light waste, calculate the extension distance of the front disc brush based on the distance between the light waste and the road boundary; determine the working position of the front disc brush based on the extension distance of the front disc brush. The rotation speed of the front disc brush is determined based on the distribution area of ​​the lightweight waste; the suction duration of the central suction nozzle is determined based on the quantity of lightweight waste; and the suction and sweeping mode is set based on the working position, rotation speed, and suction duration of the front disc brush.

7. The unmanned sweeping vehicle cleaning method based on target distribution situation analysis according to claim 6, characterized in that, The extension distance of the front disc brush is determined based on the distance between lightweight waste and the road boundary, including: The lateral offset of the lightweight waste is calculated based on the distance between the lightweight waste and the road boundary; the lateral offset is compared with the preset initial operating position of the front disc brush to calculate the disc brush position compensation. Adjust the maximum telescopic arm length of the front disc brush according to the disc brush position compensation amount, use the maximum telescopic arm length to identify the disc brush sweeping range, stop the telescopic arm when the disc brush sweeping range covers the location of light waste, and record the telescopic distance of the front disc brush.

8. The unmanned sweeping vehicle cleaning method based on target distribution situation analysis according to claim 6, characterized in that, The rotation speed of the front disc brush is determined based on the distribution area of ​​the lightweight waste, including: The coverage area corresponding to the lightweight waste is determined based on the distribution area of ​​the lightweight waste; the single operation area of ​​the front disc brush is calculated based on the coverage area; the number of times the front disc brush is used for coverage is calculated based on the coverage area and the single operation area. The sweeping load of light waste is assessed based on the number of coverages; a drive adjustment command for the front disc brush is generated based on the sweeping load; the drive adjustment command is sent to the sweeping motor of the front disc brush; the output speed of the sweeping motor is adjusted by the motor controller of the sweeping motor; and the rotation speed of the front disc brush is determined by the output speed of the sweeping motor.

9. The unmanned sweeping vehicle cleaning method based on target distribution situation analysis according to claim 3, characterized in that, If the waste type is dry waste with dust, the water spraying mode using front nozzles for dust suppression includes: If the waste type is dry waste with dust, then determine the dust target area; determine the dust diffusion direction based on the dust target area; determine the water spray direction of the front nozzle based on the dust diffusion direction; The front nozzles are controlled to spray water directionally onto the dust target area according to the water spray direction, forming a dust suppression coverage area; the water spray state of the front nozzles is adjusted according to the dust suppression coverage area to form a water spraying dust suppression mode.

10. The unmanned sweeping vehicle cleaning method based on target distribution situation analysis according to claim 1, characterized in that, Step S2 involves identifying obstacles at the road boundary and setting the cleaning path, including: Collect road environment data along the road boundary, identify obstacles to traffic based on the road environment data; determine the location of obstacles based on the obstacles; and determine the area of ​​influence of obstacles centered on the location of the obstacles. Determine the passable area based on the distance between the area affected by the obstacle and the road boundary; set the vehicle driving direction in the passable area and determine the detour trajectory along the vehicle driving direction; connect the detour trajectory with the edge cleaning trajectory corresponding to the road boundary to form a cleaning path.

Images