Method and system for optimizing the scheduling of unmanned sweeper area coverage

Abstract

The application discloses an unmanned cleaning vehicle area coverage optimization scheduling method and system, and particularly relates to the field of environmental sanitation scheduling, and is used for solving the problems that the area to be reswept is difficult to be accurately identified, the path of the reswept task is difficult to be inserted along the way, and the service delay is difficult to be quantitatively evaluated in the scene where temporary obstacles frequently occur, and is achieved by the following steps: generating a basic label with cleaning marks and obstacle marks based on a road section, constructing a reswept area record and merging to form a reswept task path, introducing an along-the-way overlap distance ratio and a service delay offset time, calculating a task allocation priority coefficient through a pre-defined scheduling matching model, cooperatively allocating between the reswept task path and a regular cleaning task, and combining an actual cleaning trajectory to feed back and adjust the reswept area identification rule and the reswept task path generation strategy, so as to realize the closed-loop optimization scheduling of the area coverage of the unmanned cleaning vehicle.

Description

Technical Field

[0001] This invention relates to the field of sanitation dispatching, and more specifically, to a method and system for optimizing the dispatching of unmanned sweeping vehicles for regional coverage. Background Technology

[0002] In real-world environments such as urban roads and industrial park roads, unmanned sweeping vehicles typically operate in fleets for extended periods. The system divides the entire road area into several cleaning zones based on electronic maps and established rules, and issues driving routes and operational instructions through a dispatch platform. During the cleaning process, temporary parking, construction obstructions, mobile trash cans, and material piles frequently occur along the roadside. These temporary obstacles periodically block a small area of ​​the roadside, preventing the unmanned sweeping vehicle from approaching and cleaning that area during the current shift. Simultaneously, the dispatch system needs to continuously record which areas have been cleaned and which have not, so that new coverage tasks can be scheduled in subsequent shifts. This achieves comprehensive coverage and optimized scheduling of the entire work area, which is the typical application scenario addressed by the unmanned sweeping vehicle area coverage optimization scheduling method and system described in the title of this invention.

[0003] However, in this scenario, existing area coverage and scheduling methods can only roughly distinguish between areas that have been cleaned and those that have not yet been cleaned. It is difficult to separately mark and manage areas that are bypassed due to temporary obstacles. Areas blocked and not cleaned during the current shift are often mixed with actually cleaned areas in the system records. Even after the temporary obstacles are removed, the scheduling platform struggles to accurately identify these remaining areas, let alone generate targeted cleanup tasks and allocate them appropriately to suitable unmanned sweeping vehicles. Over time, these small, previously obscured areas along the roadside are easily treated as covered and repeatedly ignored at the data level, manifesting as long-term neglected cleaning dead spots in the actual environment. The technical problem this invention aims to solve is how to clearly distinguish these temporarily uncleaned areas from the truly cleaned areas during the area coverage optimization scheduling process of unmanned sweeping vehicles, and automatically identify and reintegrate them into scheduling and path planning in subsequent operations, ensuring that the entire cleaning area maintains true and complete coverage even in dynamic environments.

[0004] To address the aforementioned problems, a technical solution is provided. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an unmanned sweeper area coverage optimization scheduling method and system. This method generates basic labels with cleaning and obstacle markings based on road segments, constructs records of areas to be re-sweeped, and merges them to form re-sweeping task paths. It introduces the ratio of overlapping distances along the route and service delay offset time, calculates task allocation priority coefficients using a predefined scheduling matching model, and performs collaborative allocation between re-sweeping task paths and regular cleaning tasks. Furthermore, it incorporates feedback adjustments based on actual cleaning trajectories, re-sweeping area identification rules, and re-sweeping task path generation strategies. This achieves closed-loop optimized scheduling of unmanned sweeper area coverage, thereby solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] The method for optimizing the scheduling of unmanned cleaning vehicles to cover designated areas includes the following steps:

[0008] S1 receives the current location cleaning status and temporary obstacle status reported by the unmanned sweeper, and generates basic labels with cleaning marks and obstacle marks for the road segment;

[0009] S2: Select small segments that have not been cleaned and have temporary obstacles from the basic tags, generate a record of areas to be re-cleaned and store it in the set of tasks to be re-cleaned;

[0010] S3, when the small segment of obstacle markers corresponding to the area to be rescanned changes from existing to disappearing, the adjacent areas to be rescanned are taken out from the set of tasks to be rescanned and merged into the path of the task to be rescanned.

[0011] S4. When allocating the task path to be re-sweeped, calculate the ratio of the overlapping distance along the route and the service delay offset time based on the combination of the unmanned sweeping vehicle and the task path. Input the two into the predefined scheduling matching model to obtain the task allocation priority coefficient. Based on the task allocation priority coefficient, determine the target task of each vehicle in the task path to be re-sweeped and the regular sweeping task.

[0012] S5 updates the basic labels and the record of areas to be re-sweeped based on the actual sweeping trajectory reported by the unmanned sweeper after the operation is completed. It also adjusts the identification rules for areas to be re-sweeped and the path generation strategy for tasks to be re-sweeped based on the execution results of the task allocation priority coefficient.

[0013] In a preferred embodiment, step S1 includes the following:

[0014] The work area is divided into a set of road segments. Each road segment is assigned a road segment number and its length is recorded. The system receives reports from unmanned sweepers regarding their current location, cleaning status, and temporary obstacle status. The reported records are mapped to the corresponding road segments through position and direction matching. The cumulative cleaning length and temporary obstacle occupancy length are accumulated at the road segment level. The cleaning coverage ratio is obtained based on the ratio of the cumulative cleaning length to the road segment length, and the temporary obstacle occupancy ratio is obtained based on the ratio of the temporary obstacle occupancy length to the road segment length. Based on historical work records and cleaning targets, preset coverage thresholds and preset obstacle thresholds are determined. A cleaning mark is generated by comparing the cleaning coverage ratio with the preset coverage threshold, and an obstacle mark is generated by comparing the temporary obstacle occupancy ratio with the preset obstacle threshold. The road segment number, cleaning mark, and obstacle mark are combined into a basic tag unit and written into a basic tag set.

[0015] In a preferred embodiment, step S2 includes the following:

[0016] The system reads the cleaning and obstacle markers of each road segment from the basic tag set, converts the cleaning markers into numerical cleaning markers, and converts the obstacle markers into numerical obstacle markers. It generates a rescan identifier based on the combination of numerical cleaning and obstacle markers. When the rescan identifier is one, it generates a rescan area record unit containing the road segment number, cleaning marker, obstacle marker, and rescan identification time. The system writes or updates the rescan task set using the road segment number as an index, so that each road segment in the rescan task set corresponds to only one rescan area record unit.

[0017] In a preferred embodiment, step S3 includes the following:

[0018] After the basic labels are updated, the latest cleaning mark and the latest obstacle mark are read from the basic label set according to the road segment number. They are compared with the historical obstacle marks stored in the task set according to the road segment number. Road segments marked as having temporary obstacles in the historical obstacle mark, having no temporary obstacles in the latest obstacle mark, and having not been cleaned in the latest cleaning mark are identified as road segments to be cleared and re-sweeped. Based on the geometric information of the centerline of the road segment and the topological connection relationship, the adjacency relationship between the centerline endpoints of the road segments to be cleared and re-sweeped is determined, and the driving direction is continuous. Road segments to be cleared and re-sweeped with adjacency relationship are grouped into the same area group to be cleared and re-sweeped.

[0019] In a preferred embodiment, step S3 further includes the following:

[0020] Generate a rescan task path covering all road segments within each obstacle clearance and rescan area group according to the continuous order of road segments in the road topology. Record the rescan task path in the rescan task set. In the rescan task set, write path association markers to the rescan area records belonging to each obstacle clearance and rescan area group. Establish a one-to-one correspondence between the rescan task set and the rescan task path set through the road segment number.

[0021] In a preferred embodiment, step S4 includes the following:

[0022] The scheduling platform obtains candidate task paths from the set of paths for tasks to be re-sweeped and the set of paths for regular cleaning tasks. It obtains the current position and remaining planned path of each unmanned cleaning vehicle from the set of unmanned cleaning vehicles. For each combination of unmanned cleaning vehicle and each candidate task path, the platform calculates the ratio of the total length of the road segments shared by the remaining planned path and the candidate task path to the total length of the remaining planned path at the road segment level as the along-road overlap distance ratio. The platform also selects the maximum difference of not less than zero from the difference between the actual service interval from the last cleaning to the current scheduling time and the corresponding target service interval for each road segment within the candidate task path, as the service delay offset time.

[0023] In a preferred embodiment, step S4 further includes the following:

[0024] The ratio of overlapping distances along the route and the service delay offset time are input into a predefined scheduling and matching model to obtain the task allocation priority coefficient. The candidate task paths corresponding to each unmanned sweeper are sorted according to the task allocation priority coefficient. The candidate task path with the highest task allocation priority coefficient and which is not occupied is selected as the target task path for each unmanned sweeper. The task allocation is completed between the task path to be re-sweeped and the regular sweeping task path.

[0025] In a preferred embodiment, step S5 includes the following:

[0026] After the operation is completed, the scheduling platform updates the cleaning and obstacle markers in the basic tag set according to the actual cleaning trajectory reported by the unmanned sweeper and maps the road segments. Based on this, it checks the records of areas to be re-sweeped in the set of tasks to be re-sweeped one by one. Records marked as cleaned are marked as completed and removed from the set of tasks to be re-sweeped. Records marked as uncleaned and with obstacles marked as having no temporary obstacles are retained and their re-sweeping identification time is updated. Records marked as uncleaned and with obstacles marked as having temporary obstacles are retained. On this basis, for each task path, the path fulfillment completeness ratio is obtained by the ratio of the sum of the planned lengths of the road segments in the path to the sum of the actual cleaning lengths in this round of operation. The service delay correction ratio is obtained by the ratio of the difference between the service delay offset time before scheduling and the service delay offset time after the operation to the service delay offset time before scheduling.

[0027] In a preferred embodiment, step S5 further includes the following:

[0028] A global path fulfillment integrity index is formed by summarizing the planned path length and actual cleaning length of each task path across the entire set of executed task paths. At the same time, a global service delay correction index is formed by summarizing the original total and corrected total service delay offset time of each task path. The driving coefficients for rule adjustment are constructed using the global path fulfillment integrity index and the global service delay correction index. Based on the distribution of the driving coefficients, path fulfillment integrity ratio and service delay correction ratio on different task paths, the rules for identifying areas to be re-scanned and the strategy for generating task paths to be re-scanned are iteratively adjusted.

[0029] The unmanned sweeper area coverage optimization and scheduling system includes:

[0030] Status labeling module: Receives the current location cleaning status and temporary obstacle status reported by the unmanned sweeper, and generates basic labels with cleaning and obstacle marks for road segments;

[0031] Rescan recognition module: Selects small segments that have not been cleaned and have temporary obstacles from the basic tags, generates a record of areas to be rescanned and stores it in the set of tasks to be rescanned;

[0032] Path generation module: When the detection of a small segment of obstacle marker corresponding to the area to be rescanned changes from existing to disappearing, the module takes the adjacent areas to be rescanned from the set of tasks to be rescanned and merges them into the path of the task to be rescanned.

[0033] Priority scheduling module: When allocating the path of the task to be re-sweeped, calculate the ratio of the overlapping distance along the route and the service delay offset time based on the combination of the unmanned sweeper and the task path. Input the two into the predefined scheduling matching model to obtain the task allocation priority coefficient. Based on the task allocation priority coefficient, determine the target task of each vehicle in the task path to be re-sweeped and the regular sweeping task.

[0034] Feedback optimization module: After the operation is completed, the basic labels and the record of areas to be cleaned again are updated according to the actual cleaning trajectory reported by the unmanned sweeper. The recognition rules for areas to be cleaned again and the path generation strategy for tasks to be cleaned again are adjusted according to the execution results of the task allocation priority coefficient.

[0035] The technical effects and advantages of the unmanned sweeper area coverage optimization scheduling method and system of the present invention are as follows:

[0036] This invention achieves a closed-loop processing of unmanned sweeping vehicles for the cleaning status and temporary obstacle changes of road segments in the same scheduling process by continuously linking basic tags, records of areas to be re-sweeped, task paths to be re-sweeped, the ratio of overlapping distances along the route, service delay offset time, and task allocation priority coefficients within the same scheduling process. This process involves discovery, temporary storage, merging, allocation and execution, and result writing back. Areas to be re-sweeped are no longer fixed by a one-time static judgment, but automatically evolve into continuous task paths to be re-sweeped after the obstacles disappear. These paths are then uniformly scheduled within the same priority system as regular cleaning tasks, reducing redundant detours and ineffective empty runs, and also reducing reliance on temporary path adjustments based on human experience.

[0037] Meanwhile, by retrospectively analyzing the actual cleaning trajectory after the operation, two types of quantitative feedback are constructed: path fulfillment integrity and service delay correction. The execution effect of the task allocation priority coefficient is mapped in real time to the adjustment of the rules for identifying areas to be re-sweeped and the path generation strategy for tasks to be re-sweeped. This allows the scheduling process to gradually converge as the distribution of road network obstacles and the pressure of cleaning change, thereby improving vehicle mileage utilization and the timeliness of re-sweeping response while ensuring the balance of road coverage. Attached Figure Description

[0038] Figure 1 This is a flowchart illustrating the area coverage optimization scheduling method for unmanned sweeping vehicles according to the present invention.

[0039] Figure 2 This is a schematic diagram of the structure of the unmanned sweeper area coverage optimization scheduling system of the present invention. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] Example 1: Figure 1 The present invention provides a method for optimizing the scheduling of unmanned sweeping vehicles for area coverage, including:

[0042] S1 receives the current location cleaning status and temporary obstacle status reported by the unmanned sweeper, and generates basic labels with cleaning and obstacle markings for the road segment.

[0043] S2: Select small segments that have not been cleaned and have temporary obstacles from the basic tags, generate a record of areas to be re-cleaned, and store it in the set of tasks to be re-cleaned.

[0044] S3, when the small segment of obstacle markers corresponding to the area to be rescanned changes from existing to disappearing, the adjacent areas to be rescanned are taken out from the set of tasks to be rescanned and merged into the path of the task to be rescanned.

[0045] S4. When allocating the task path to be re-sweeped, calculate the ratio of the overlapping distance along the route and the service delay offset time based on the combination of the unmanned sweeper and the task path. Input the two into the predefined scheduling matching model to obtain the task allocation priority coefficient. Based on the task allocation priority coefficient, determine the target task of each vehicle in the task path to be re-sweeped and the regular sweeping task.

[0046] S5 updates the basic labels and the record of areas to be re-sweeped based on the actual sweeping trajectory reported by the unmanned sweeper after the operation is completed. It also adjusts the identification rules for areas to be re-sweeped and the path generation strategy for tasks to be re-sweeped based on the execution results of the task allocation priority coefficient.

[0047] In scenarios where unmanned sweepers undertake long-term cleaning operations on urban and park roads, subsequent steps require identifying uncleaned areas and areas with temporary obstacles at the road segment level. The same set of road segment numbers is repeatedly referenced across multiple work cycles to generate records of areas requiring re-sweeping and re-sweeping task paths. Without first establishing a stable state description at the road segment level, any subsequent identification and scheduling of areas requiring re-sweeping will be difficult to base on a clear and consistent data foundation. Therefore, in step S1, the continuously reported current location, cleaning status, and temporary obstacle status of the unmanned sweeper are gradually transformed into a basic tag set with cleaning and obstacle markers, providing a unified data entry point for step S2 and subsequent steps.

[0048] Sub-step S1-1: Establishment of road segment set and calibration of road segment length.

[0049] Based on the existing road data within the work area, each road is divided into several road segments according to geometric continuity and management needs, forming a set of road segments. A unique road segment number is assigned to each road segment in the set, and the centerline geometry and length of the road segment are recorded, with the road segment length using a unit consistent with the mileage traveled by the unmanned sweeper. This ensures that subsequent processing always revolves around the same set of road segment numbers for state updates and tag generation, and any subsequent judgment regarding a particular segment can be traced back to the unique road segment number.

[0050] Sub-step S1-2 establishes the correspondence between the reported records and road segments.

[0051] The dispatch platform continuously receives reports from unmanned sweepers during operation. Each report contains at least three elements: current location, sweeping status, and temporary obstacle status. For each report, firstly, based on the geometric distance between the reported location and the centerline of each road segment, the closest candidate road segments are selected. Then, based on the relative relationship between the reported driving direction of the unmanned sweeper and the centerline direction of the candidate road segments, a road segment with the same direction and reasonable location is selected from the candidate road segments, establishing a one-to-one correspondence between the report and this road segment. This ensures that each report is assigned to a specific road segment, providing a source for updating the status of that road segment and laying the foundation for subsequent statistics on sweeping and obstacle status at the road segment level.

[0052] Sub-step S1-3 updates the cumulative cleaning length and the length occupied by temporary obstacles.

[0053] For each road segment in the set of road segments, establish two quantities: cumulative cleaning length and temporary obstacle occupancy length, both initially set to zero. For two consecutive reports from an unmanned sweeper on the same road segment, project the first and second reported positions along the centerline of the road segment to obtain the projected positions of the two points on the road segment. Then, calculate the arc length between these two points along the centerline of the road segment based on the geometric parameters of the road segment centerline. The unit of the arc length is consistent with the length of the road segment. When both reported cleaning statuses indicate that cleaning is in progress, this arc length is regarded as the increment of the cleaning length completed on the road segment during this period, and added to the original cumulative cleaning length of the road segment, so that the cumulative cleaning length gradually reflects the total driving length of the vehicle accumulated cleaning on the road segment.

[0054] For updating the length of temporary obstacle occupancy, when a reported temporary obstacle status indicates the existence of a temporary obstacle, the starting and ending points of the obstacle's impact are determined on the centerline of the road segment based on the obstacle range fed back by the unmanned sweeper's sensors. Then, the difference in arc length between these two points on the road segment is used to calculate the road length affected by the obstacle. This road length is then added to the corresponding road segment's temporary obstacle occupancy length, thus gradually accumulating the total length of the road segment obstructed by temporary obstacles within the current work cycle. Through this accumulation method, each road segment has a cumulative sweeping length and temporary obstacle occupancy length corresponding to the actual operation process at the end of a work cycle.

[0055] Sub-step S1-4: Calculation of sweeping coverage ratio and temporary obstacle occupancy ratio.

[0056] Given the aforementioned road segment length, cumulative cleaning length, and temporary obstacle occupancy length, a cleaning coverage ratio and a temporary obstacle occupancy ratio are calculated for each road segment. The cleaning coverage ratio represents the proportion of the road segment's length that has been cleaned to its total length. Specifically, it is calculated by ratioing the cumulative cleaning length to the road segment length, resulting in a dimensionless value between zero and one, representing the degree of cleaning coverage of the road segment. The temporary obstacle occupancy ratio represents the proportion of the road segment's length obstructed by temporary obstacles to its total length. This is also calculated by ratioing the temporary obstacle occupancy length to the road segment length, resulting in another dimensionless value, representing the degree to which the road segment is affected by temporary obstacles during the current work cycle. By transforming the cumulative results with length dimensions into proportional values ​​with a unified meaning, the cleaning and obstacle situations of road segments of different lengths and shapes can be compared on the same scale, creating conditions for using a unified threshold for marking and judgment.

[0057] Sub-step S1-5: The process of determining the preset coverage threshold and the preset obstacle threshold.

[0058] To ensure that the generation of cleaning and obstacle markers reflects both historical operational performance and road cleaning requirements, after multiple completed work cycles, a sequence of cleaning coverage ratios and temporary obstacle occupancy ratios for all road segments is collected at the end of each work cycle. For the cleaning coverage ratio, a target coverage level is set to reflect the proportion of road segments where cleaning requirements are expected to be met at the end of the work cycle. By selecting different candidate coverage levels from the cleaning coverage ratio sequence, the ratio between the number of road segments with a cleaning coverage ratio not lower than the candidate coverage level and the total number of road segments is calculated. This ratio is compared with the target coverage level, and the candidate coverage levels are gradually adjusted. The level that makes the statistical ratio closest to the target coverage level is selected as the preset coverage threshold.

[0059] For the temporary obstacle occupancy ratio, the focus is on examining the persistence of temporary obstacle occupancy for each road segment within a single work cycle. The differences between road segments that are consistently occupied for extended periods and those that are only occasionally occupied for short periods are statistically analyzed. Then, the obstacle judgment level that best distinguishes these two types of road segments is identified within the temporary obstacle occupancy ratio sequence, and this judgment level is used as the preset obstacle threshold. Through this series of analyses, the preset coverage threshold and preset obstacle threshold do not depend on fixed values ​​but are determined jointly by historical records and target requirements. This allows for long-term, consistent judgment of road segment status across different work cycles.

[0060] Sub-steps S1-6: Generation of basic tags and storage of the basic tag set.

[0061] Given that the sweeping coverage ratio and temporary obstacle occupancy ratio for each road segment have been calculated, and the preset coverage threshold and preset obstacle threshold have been determined, a basic label is generated for each road segment. For a given road segment, if its sweeping coverage ratio is not lower than the preset coverage threshold, the road segment is marked as swept; if the sweeping coverage ratio is lower than the preset coverage threshold, the road segment is marked as not swept. Similarly, if its temporary obstacle occupancy ratio is not lower than the preset obstacle threshold, the road segment is marked as having a temporary obstacle; if the temporary obstacle occupancy ratio is lower than the preset obstacle threshold, the road segment is marked as having no temporary obstacle. The road segment number, sweeping label, and obstacle label are combined to form a basic label unit, and all basic label units together constitute a basic label set.

[0062] The basic tag set serves as the input for step S2. In step S2, it is used to filter out road segments marked as uncleaned and obstacles marked as having temporary obstacles. These road segment numbers and their marking information are continuously referenced in subsequent steps to construct the upper-level data structure of the area to be re-scanned and the task path to be re-scanned.

[0063] Through the sub-step processing of step S1, the discrete reporting records of the unmanned sweeper during operation are mapped and accumulated along the centerline of road segments, so that each road segment has a cumulative sweeping length and a temporary obstacle occupancy length. This is further converted into sweeping coverage ratio and temporary obstacle occupancy ratio. Combined with preset coverage thresholds and preset obstacle thresholds determined based on historical operation data and road cleaning targets, basic labels with sweeping and obstacle markers are generated for each road segment, and these are summarized into a basic label set. In this way, subsequent steps can directly use the basic label set as a starting point to identify unkempt segments with temporary obstacles at the road segment granularity, thereby generating records of areas to be re-sweeped and task paths to be re-sweeped. This ensures that the entire invention has a clear and stable upstream data source in the area coverage optimization scheduling process.

[0064] Step S1 has generated a set of basic labels containing cleaning marks and obstacle marks for each road segment at the road segment granularity. In order to merge adjacent areas to form a task path to be re-sweeped after the obstacle disappears, it is necessary to first select the road segments that have not yet been swept and are currently occupied by temporary obstacles from these basic labels in step S2, and register and retain these road segments as areas to be re-sweeped separately at the data level, so as to facilitate the continuous use of the same road segment number for scheduling and path planning in subsequent operation cycles.

[0065] Sub-step S2-1: Reading the basic tag set and extracting the status of road segments.

[0066] The basic label units generated in step S1 are read sequentially according to the road segment numbers. Each basic label unit corresponds to a fixed road segment and contains at least three fields: road segment number, cleaning mark, and obstacle mark. For a basic label unit with a specific road segment number, the cleaning mark is read as the cleaning status of that road segment, and the obstacle mark is read as the temporary obstacle status of that road segment. By traversing all basic label units, a status list is formed that corresponds one-to-one between road segments and their cleaning and temporary obstacle statuses, providing a clear data source for subsequent rescan determination at the road segment level.

[0067] Sub-step S2-2: Numerical representation of cleaning state and obstacle state.

[0068] The cleaning and obstacle markers obtained in sub-step S2-1 are numerically processed to facilitate subsequent combination and judgment using a unified logic. For any road segment, if the cleaning marker indicates that it is not cleaned, the numerical cleaning marker for that road segment is set to one; if the cleaning marker indicates that it is cleaned, the numerical cleaning marker for that road segment is set to zero. Similarly, for any road segment, if the obstacle marker indicates that a temporary obstacle exists, the numerical obstacle marker for that road segment is set to one; if the obstacle marker indicates that there is no temporary obstacle, the numerical obstacle marker for that road segment is set to zero. Through this numerical processing method, the two types of states, originally represented by text, are uniformly converted into numerical identifiers with only zero and one values, so that the conditions of not being cleaned and the existence of a temporary obstacle can be directly reflected in the subsequent textual description through numerical combinations.

[0069] Sub-step S2-3: Generation of markers to be rescanned and screening of road segments.

[0070] Based on the numerical sweep markers and numerical obstacle markers in sub-step S2-2, a rescan marker is generated for each road segment. The rescan marker represents whether the road segment needs to be registered as a rescan area. For any road segment, when both the numerical sweep marker and the numerical obstacle marker are one, the rescan marker is set to one, indicating that the road segment is not swept and has temporary obstacles. In other combinations, the rescan marker is set to zero, indicating that the road segment will not be registered as a rescan area in this step. In this way, a set of rescan marker values ​​is formed for all road segments, clearly distinguishing between road segments that need to be registered as rescan areas and those that do not within the same data structure.

[0071] Sub-step S2-4: Generation and repeat check of records for the area to be rescanned.

[0072] For each road segment marked as "to be re-scanned," a record unit for the area to be re-scanned is generated and matched with the existing set of tasks to be re-scanned. For each road segment marked as "to be re-scanned," its road segment number, cleaning mark, and obstacle mark are first read. Then, combined with the current identification time, a re-scanning identification time is generated. These data are combined to form a record unit for the area to be re-scanned. The re-scanning identification time is used in subsequent steps to determine when this record enters the re-scanning state within the current work cycle. After generating a record unit for the area to be rescanned, a query is performed in the set of tasks to be rescanned using the road segment number as an index. If no record with the same road segment number exists in the set, the new record unit is directly written to the set, indicating that this road segment is entering the rescanning state for the first time. If a record with the same road segment number and in the rescanning state already exists in the set, no new record is created; instead, the cleaning marker, obstacle marker, and rescanning identification time in the record are updated to reflect the current state, ensuring that the record content in the set matches the latest state in the basic tag set. Through this checking and updating process, each road segment corresponds to at most one record unit for the area to be rescanned in the set of tasks to be rescanned.

[0073] Sub-step S2-5 Update and prepare for output of the set of tasks to be rescanned.

[0074] After completing the creation or update operations for all record units of the areas to be rescanned in sub-steps S2-4, the updated set of tasks to be rescanned is persistently saved as the output of step S2. The updated set of tasks to be rescanned only includes road segments marked as uncleaned and marked as having temporary obstacles under the current basic tag set. Each record carries the road segment number, the latest cleaning mark, the latest obstacle mark, the rescanning identifier, and the rescanning identification time. The set of tasks to be rescanned structurally forms a one-to-one correspondence with the set of road segments, providing a clear entry point for step S3 to directly find the corresponding record of the area to be rescanned by the road segment number and perform the merging of adjacent road segments when the obstacle status changes.

[0075] Through the sub-steps of step S2, based on the basic label set already formed in step S1, the cleaning and obstacle markers of each road segment are further numerically represented. The numerical cleaning and obstacle markers are combined to generate a re-scanning identifier, thereby filtering out road segments with cleaning markers indicating no cleaning and obstacle markers indicating temporary obstacles at the road segment level. A re-scanning area record unit is generated for these road segments, containing the road segment number, cleaning marker, obstacle marker, and re-scanning identification time. The repetition check logic ensures that the road segment number in the re-scanning task set is unique, providing clear and complete upstream data for the obstacle disappearance monitoring and re-scanning task path merging based on the re-scanning task set in the subsequent step S3.

[0076] Step S2 has already filtered out road segments marked as uncleaned and those marked as having temporary obstacles from the basic label set, and written them into the re-sweeping task set using the road segment number as an index, so that these temporarily obscured uncleaned locations are in a re-sweeping state at the data level. However, in subsequent operations, illegally parked vehicles, temporary barriers, and turnover materials on the roadside will be moved or removed, and the obstacle status will change from having temporary obstacles to having no temporary obstacles. If we still remain at the level of isolated road segment records, we cannot directly issue a structurally clear re-sweeping path to the unmanned sweeper. Therefore, in step S3, it is necessary to identify the road segments that have been cleared of obstacles and merge them into continuous re-sweeping task paths according to the road spatial topology, providing path-level input for subsequent task allocation.

[0077] Sub-step S3-1: Obstacle removal and rescanning of candidate road segments for identification.

[0078] After a new set of basic tags is refreshed, status information is read from both the basic tag set and the rescanning task set according to road segment number. For each road segment number, the latest cleaning mark and the latest obstacle mark are read from the basic tag set and recorded as the latest cleaning mark and the latest obstacle mark. The historical cleaning mark and the historical obstacle mark are read from the rescanning task set and recorded as the historical cleaning mark and the historical obstacle mark. The historical cleaning mark and the historical obstacle mark are the statuses written when the rescanning area record is generated in step S2, reflecting the situation when the road segment enters the rescanning task set. In this way, each road segment registered in the rescanning task set has two sets of information: the obstacle situation when entering the rescanning state and the current latest obstacle situation, providing a basis for determining whether the obstacle has been cleared.

[0079] Sub-step S3-2: Determine the road segment to be re-scanned after obstacle removal.

[0080] Based on the status read in sub-step S3-1, each road segment recorded in the set of tasks awaiting rescanning is assessed for obstacle removal. For a given road segment, if a historical obstacle marker indicates the presence of a temporary obstacle, the latest obstacle marker indicates the absence of a temporary obstacle, and the latest cleaning marker indicates it is still not cleaned, then the road segment is identified as a road segment awaiting rescanning. A historical obstacle marker indicating the presence of a temporary obstacle means the road segment was previously registered as a rescanning area due to an obstacle; a latest obstacle marker indicating the absence of a temporary obstacle means the obstacle has been removed; and a latest cleaning marker indicating it is not cleaned means that rescanning has not yet been performed after the obstacle was removed. Only when all three conditions are met simultaneously is the road segment added to the set of road segments awaiting rescanning for obstacle removal; other combinations are not processed for rescanning. This forms a group of road segments that are both registered in the set of tasks awaiting rescanning and are ready for rescanning at any time.

[0081] Sub-step S3-3 Obstacle removal and rescanning of spatial adjacency relationships in small road segments.

[0082] The spatial adjacency relationships within the set of road segments awaiting rescanning are identified. Using the centerline geometry and road topology established in step S1, each road segment is considered a directional centerline segment. For any two road segments awaiting rescanning, if the endpoint of the centerline of one road segment coincides with the starting point of the centerline of the other road segment in position or within the allowable accuracy range, and the travel directions of the two road segments are continuous at the connection point without reversal, then these two road segments are considered spatially adjacent. Within the set of road segments awaiting rescanning, the above determination is repeated for each pair of road segments. By recording which road segments are adjacent to each other, an adjacency relationship network covering only the road segments awaiting rescanning is constructed, laying the structural foundation for subsequent division of continuous regions.

[0083] Sub-step S3-4: Construction of the area group to be rescanned after obstacle removal.

[0084] Based on the adjacency network obtained in sub-step S3-3, the road segments awaiting rescanning are divided into several regions awaiting rescanning. The division process begins with a road segment that is not yet grouped but is already in the set of road segments awaiting rescanning. This segment is used as the starting road segment of a region group. Then, adjacent road segments awaiting rescanning are searched in the adjacency network and added to the current region group. This process is repeated for newly added road segments until all road segments connected by adjacency relationships are grouped into the same region group. For road segments awaiting rescanning that have not yet been grouped into any region group, new regions awaiting rescanning are constructed in the same manner. Through this process, the set of road segments awaiting rescanning is divided into several internally connected subsets. Each subset corresponds to a spatially contiguous region group, preventing the generation of subsequent paths from crossing unconnected regions.

[0085] Sub-step S3-5: Generation of the path to be rescanned.

[0086] For each group of areas awaiting rescanning, a rescanning task path covering all road segments within the group is generated. For a given group of areas awaiting rescanning, first, a road segment is selected as the starting point of the path. Prioritize road segments that are adjacent to only one other road segment awaiting rescanning in the adjacency network, allowing the path to extend from the edge of the area. After recording the starting road segment number in the path sequence, search the adjacency network for road segments adjacent to the current road segment that are not yet written into the path sequence. If multiple candidate road segments exist, select one according to a pre-defined road level, road direction, or number order as the next road segment, append it to the end of the path sequence, and update the current road segment to this new road segment. Repeat the above search and append operations until all road segment numbers within the entire group of areas awaiting rescanning have been written into the path sequence. The sequence of road segments in the path sequence corresponds to the actual driving sequence of the unmanned sweeper, which is used to form a continuous path for the task to be swept again. In a single execution, the entire area to be cleared and swept again can be covered at once.

[0087] For obstacle removal and rescanning areas with branching structures, the path generation process arranges the access order of multiple road segments according to the road topology. For example, it first proceeds along the main road, then enters the branch roads sequentially at appropriate locations, returns to the main road, and continues along the main road until all branches have been visited once. This ensures that the path remains as coherent as possible while covering all obstacle removal and rescanning road segments. After processing, a rescanning task path record is generated for each path, recording the identifier of the obstacle removal and rescanning area group, the ordered sequence of road segment numbers in the path, and the path generation time. These records are then written into the rescanning task path set.

[0088] The appropriate location refers to the point where the path has reached the intersection of the current road segment and the target branch road, and the unmanned sweeper can directly enter the target branch road in the direction of traffic without turning around at the intersection.

[0089] Sub-step S3-6: Update the record of the area to be rescanned and the path of the task to be rescanned.

[0090] The set of tasks to be re-scanned is updated to establish a stable association between the records of areas to be re-scanned and their corresponding paths. For each group of areas to be re-scanned, the path record corresponding to that group is read. The path number and the sequence of road segment numbers within the path are obtained from the path record. Then, the corresponding area record unit in the set of tasks to be re-scanned is searched by road segment number. The path number is written into these record units as a path association marker, indicating that the area record to be re-scanned has been merged into a certain path. Through this update operation, the records in the set of tasks to be re-scanned are no longer isolated road segments, but clearly indicate which path each road segment belongs to. This makes it convenient for step S4 to directly select a path from the set of paths to be re-scanned during task allocation, and to locate the specific road segment record through the path association marker when necessary.

[0091] Through the sub-steps described in step S3, based on the set of tasks to be re-scanned generated in step S2, and combining the latest cleaning markers and obstacle markers from the basic tag set, road segments that have been cleared but not yet cleared are selected for re-scanning. Spatial adjacency relationships are identified within this set of road segments, dividing them into several connected groups of areas to be re-scanned. A re-scanning task path covering all road segments is then generated for each group and written into the re-scanning task path set. Simultaneously, path association markers are written into the re-scanning task set to establish a consistent mapping between the re-scanning area records and the re-scanning task paths. In this way, in urban roads and park roads where temporary obstacles constantly appear and disappear, obscured, uncleaned areas can be promptly transformed into well-structured re-scanning task paths after the obstacles are cleared, providing a complete and coherent path-level foundation for task allocation in the subsequent step S4.

[0092] Step S3, after the obstacle disappears, merges adjacent areas to be re-sweeped into continuous re-sweeping task paths based on the road topology and writes them into the re-sweeping task path set. At this point, on the one hand, there are several re-sweeping task paths that need to be swept as quickly as possible without disrupting the overall movement of the convoy; on the other hand, there are also regular cleaning task paths that need to be continuously maintained. If a more detailed task selection mechanism is not introduced at this stage, and instead, tasks are simply executed according to the nearest distance or a fixed priority order, it will be difficult to simultaneously address both the need to insert re-sweeping tasks along the route and to prioritize handling road sections with severe service delays. Therefore, in step S4, it is necessary to construct two quantities based on the current remaining planned paths of the unmanned sweeping vehicles: the ratio of overlapping distances along the route and the service delay offset time. Then, a task allocation priority coefficient is formed through a scheduling matching model, thereby selecting a suitable target task path for each unmanned sweeping vehicle between the re-sweeping task paths and the regular cleaning tasks.

[0093] Sub-step S4-1: Summary of candidate task paths and unmanned sweeper status.

[0094] The scheduling platform internally organizes the basic data required for the current round of scheduling. First, it reads all re-scanning task path records from the re-scanning task path set. Each re-scanning task path record contains a path number and a sequence of road segment numbers arranged in the driving order. Then, it reads all regular cleaning task path records from the regular cleaning task path set. The record structure is similar to the re-scanning task paths, also existing in the form of path numbers and road segment number sequences. Next, it obtains the current position of each unmanned cleaning vehicle from the unmanned cleaning vehicle set and the current remaining planned path determined by the previous scheduling result. The current remaining planned path is also represented by a sequence of road segment numbers, allowing subsequent calculations to be performed on a unified data structure. Based on the above information, a task combination is constructed for each unmanned cleaning vehicle and each candidate task path, used as the calculation object for the along-route overlap distance ratio and service delay offset time.

[0095] Sub-step S4-2: Calculation process of the ratio of overlapping distances along the route.

[0096] At any given moment, an unmanned cleaning vehicle typically already has a remaining planned path. If a candidate task path shares a large number of identical road segments with this remaining planned path on the road network, then when executing this task path, most of the cleaning vehicle's travel distance is already within the existing plan. Only supplementary sweeping actions or slight adjustments are needed at certain locations, resulting in low insertion costs and good route alignment. Conversely, if the candidate task path and the remaining planned path have almost no overlapping road segments, the cleaning vehicle needs to deviate significantly from its existing travel direction, traveling an extra distance to complete the task, resulting in poor route alignment. The purpose of designing the route overlap distance ratio is to express this degree of route alignment using a ratio between zero and one.

[0097] In the specific calculation process, for a given unmanned sweeper, the set of road segment numbers contained in its current remaining planned path is read, and the total travel length of the remaining planned path is calculated based on the road segment length information. For a given candidate task path, the set of road segment numbers in the path record is read, and the geometric representation of the entire task path in the road network can also be constructed using the road segment length information. Then, at the road segment number level, the intersection set of the two is calculated, which is the set of road segments that simultaneously belong to both the remaining planned path and the candidate task path. For each road segment in the intersection set, the corresponding road segment length is extracted and added together to obtain the total overlap length of the two paths at the road segment level; simultaneously, the lengths of all road segments in the remaining planned path are accumulated to obtain the total length of the remaining planned path. The along-road overlap distance ratio is defined as the ratio between the total overlap length and the total length of the remaining planned path, forming a dimensionless value. This calculation is performed for each combination of unmanned sweeper and each candidate task path, obtaining the corresponding along-road overlap distance ratio under the premise that the total length of the remaining planned path is not zero. The closer the value is to one, the more roads the candidate task path shares with the current remaining planned path, and the more suitable it is to be inserted along the same route; the closer the value is to zero, the less common roads the candidate task path shares with the current remaining planned path, and the poorer the routeability.

[0098] Sub-step S4-3: Calculation of service delay offset time

[0099] For various road segments within urban and industrial parks, there are typically preset target service intervals, describing the timeframe within which at least one cleaning is expected to be completed. If any road segment within a task path exceeds the target service interval significantly since its last cleaning, it indicates a significant service lag and should be prioritized in task allocation. Conversely, if the most recent cleaning of any road segment within the path is within the target service interval, the service remains within the target range, and there's no need to artificially increase its priority in this scheduling round. The service delay offset is the time measure used to assess the degree of service lag within a candidate task path.

[0100] In the specific calculation process, for a given candidate task path, all road segment numbers are obtained from the path record. For each road segment number, the time of the most recent cleaning completion is read from the job record, and the target service interval duration corresponding to that road segment is read from the road service policy; the current scheduling time is used as a unified reference time. For each road segment, the actual interval duration from the most recent cleaning to the current scheduling time is calculated, and then compared with the target service interval duration to obtain a time difference. A positive difference indicates that the actual service interval has exceeded the target service interval, resulting in service delay; a difference of zero or negative indicates that it is still within the target service interval and does not constitute service lag. When constructing the service delay offset time, for road segments with a difference of zero or negative, a zero value is used in subsequent statistics to focus on the parts where service delay has already occurred. Subsequently, among all road segments involved in the entire candidate task path, the maximum value of the above differences is selected as the service delay offset time, using a time-based numerical value to summarize the service delay degree of the most severe road segment within the task path. The larger the service delay offset time, the more likely there are road segments with longer service lag times within the path, and the more priority the task path needs to be processed in this round of scheduling.

[0101] Sub-step S4-4: Scheduling matching model and task allocation priority coefficient calculation process.

[0102] A scheduling matching model is introduced to comprehensively assess the relationship between the ratio of overlapping distances along the same route and the service delay offset time. Relying solely on the ratio of overlapping distances along the same route only reflects the degree of route proximity, and relying solely on the service delay offset time only reflects the degree of service lag. The role of the scheduling matching model is to combine the two to form a priority coefficient for task allocation used for sorting and selection.

[0103] For each combination of an unmanned cleaning vehicle and a candidate task path, the along-path overlap ratio has been calculated in sub-step S4-2, and the service delay offset time corresponding to the task path has been calculated in sub-step S4-3. The scheduling matching model uses these two quantities as input and outputs a task allocation priority coefficient according to pre-defined rules or pre-trained functional relationships. The scheduling matching model is designed such that when the along-path overlap ratio increases, the task allocation priority coefficient increases accordingly, assuming the service delay offset time remains constant, reflecting that better along-path performance is more desirable; conversely, when the service delay offset time increases, the task allocation priority coefficient also increases, assuming the service delay is more severe, requiring priority processing. For task paths with excellent along-path performance but small service delay offset times, the scheduling matching model can provide a medium-to-high level of task allocation priority coefficient, making them more suitable for along-path task insertion; for task paths with average along-path performance but large service delay offset times, the scheduling matching model can provide a similarly high task allocation priority coefficient, prompting the scheduling platform to arrange appropriate detours to eliminate significant service delays. By setting the scheduling matching model in a nonlinear or piecewise manner, the change in the task allocation priority coefficient can reflect the combined effect of the two types of factors, without simply degenerating into a mechanical addition of the two quantities.

[0104] In one embodiment, a description of the pre-trained scheduling matching model is provided:

[0105] The first phase, training data construction, involves extracting multiple scheduling cycles from long-term operation records. Each scheduling cycle is considered a time slice, containing a set of unmanned cleaning vehicles participating in the scheduling and a set of candidate task paths. These candidate task paths include paths for tasks requiring re-sweeping and regular cleaning tasks. For each unmanned cleaning vehicle and each candidate task path combination within each time slice, a set of along-the-way overlap distance ratios is obtained using the method disclosed in step S4. Similarly, a set of service delay offset times is obtained using the method disclosed in step S4. Simultaneously, the scheduling execution records are reviewed to determine whether the task path was actually assigned to the unmanned cleaning vehicle within the time slice and whether the operational indicators for detour distance and service delay were met after execution. If both conditions are met, the combination is assigned a supervision label of 1; otherwise, it is assigned a supervision label of 0. This process generates a large number of samples, each consisting of the along-the-way overlap distance ratio, the service delay offset time, and a supervision label.

[0106] The second stage is the formal design stage of the scheduling and matching model. The model uses the ratio of overlapping distances along the route and the service delay offset time as inputs. To ensure a smooth response to inputs with different numerical ranges, the ratio of overlapping distances along the route is first transformed using a monotonically increasing logarithmic transformation. This weakens changes close to zero and gradually flattens changes close to one, yielding the along-route feature value. Then, the service delay offset time is transformed using a monotonically increasing logarithmic transformation to prevent excessive stretching of short and extremely long delays within the model, yielding the service feature value. Subsequently, an intermediate score is constructed, combining the along-route and service feature values ​​through a multiplicative structure. Specifically, the along-route feature value directly participates in the scoring, while the service feature value is amplified by a linear amplification factor. The two are multiplied to obtain a comprehensive score, which increases the comprehensive score when the route is good and further amplifies it when the service delay is severe. Finally, a monotonically increasing compression mapping function transforms the comprehensive score to between zero and one, serving as a task allocation priority coefficient. Higher comprehensive scores result in a task allocation priority coefficient closer to one, while lower comprehensive scores result in a task allocation priority coefficient closer to zero.

[0107] The third stage, the training objective and loss function design stage, aims to ensure that the output task allocation priority coefficient of the scheduling matching model aligns with historically well-performing combinations. A logarithmic loss is constructed between the supervision label and the task allocation priority coefficient for each training sample. If the supervision label for a combination is one, the logarithm of the task allocation priority coefficient is penalized in the loss function to encourage it to move closer to one. If the supervision label is zero, the logarithm of one minus the logarithm of the task allocation priority coefficient is penalized in the loss function to encourage it to move closer to zero. The loss terms for all vehicle-task path combinations across all time slices are summed and averaged to form an overall loss index. The training objective is to minimize this overall loss across all training samples. In this stage, the scheduling matching model tends to assign high task allocation priority coefficients to historically frequently used and effective combinations, and low task allocation priority coefficients to historically ignored or poorly performed combinations.

[0108] The fourth stage, parameter update and optimization, involves iteratively updating the parameters in the scheduling matching model, including the path transformation coefficient, service transformation coefficient, service impact adjustment coefficient, and compression mapping steepness. In each iteration, a subset of time slices is selected from the training sample set to form the training set for that round. Following the format given in the first two stages, the task allocation priority coefficient and loss value corresponding to this subset of samples are calculated. Then, based on the loss function, the direction of change for each model parameter is determined, and the parameters are fine-tuned along the direction that reduces the loss. The parameter update step size is controlled by the learning rate. During training, a subset of time slices is simultaneously retained as a validation set. Validation set samples do not participate in parameter updates; they are only used to monitor the trend of loss changes with the number of iterations. Training stops when the training set loss continues to decrease while the validation set loss no longer significantly decreases after multiple iterations, and the parameters obtained at this point are considered the final parameters.

[0109] The fifth stage is the hyperparameter setting and tuning stage. Hyperparameters include the learning rate, the number of time slices used in each iteration, and the upper limit of the number of training epochs. The learning rate controls the step size of parameter updates, the number of time slices in each iteration affects the stability of a single loss estimate, and the upper limit of the number of training epochs determines the duration of the entire optimization process. The learning rate is determined through preliminary experiments, testing a set of candidate values ​​to select one that allows for a rapid decrease in loss in the early stages without significant oscillations in the later stages. The number of time slices used in each iteration is selected based on the scale of historical data and computational resources, ensuring that the loss estimate in each epoch is representative without excessively consuming computational resources. The upper limit of the number of training epochs is determined by observing the loss curves of the training and validation sets. When the validation set loss enters a stable range and no longer improves, the number of training epochs is no longer increased. Regarding the initial values ​​of the parameters, the initial values ​​of the path transformation coefficient and the service transformation coefficient are selected to ensure that the two inputs fall within a medium range after logarithmic transformation. The initial value of the service influence adjustment coefficient is selected to be at a medium-to-small level so that pathality plays a major role in the early stages of training. The initial value of the compression mapping steepness parameter is selected to ensure that the task allocation priority coefficient is evenly distributed between zero and one, and is then automatically adjusted through training.

[0110] Phase 6 Online Operation: After offline training is completed and the scheduling matching model parameters are fixed, step S4 directly calls this predefined model during online scheduling. For each unmanned sweeper and each candidate task path combination, firstly, the along-route overlap distance ratio is calculated according to the logic disclosed in step S4, and then the corresponding service delay offset time is calculated according to the logic disclosed in step S4. These two quantities are then input into the along-route transformation function and the service transformation function respectively to obtain along-route feature values ​​and service feature values, further forming a comprehensive score. Finally, the task allocation priority coefficient is obtained through compression mapping. In each round of scheduling, for all candidate task paths of the same unmanned sweeper, the target task path is selected according to the task allocation priority coefficient from high to low. Already assigned task paths are eliminated through coordination among multiple vehicles, thereby achieving unified scheduling of re-sweeping task paths and regular sweeping task paths, transforming the empirical rules obtained from offline training into a quantitative basis for online task allocation.

[0111] Sub-step S4-5 is the task path allocation process based on the task allocation priority coefficient.

[0112] The existing task allocation priority coefficients are used to match unmanned cleaning vehicles with task paths. For each unmanned cleaning vehicle, the corresponding task allocation priority coefficients are collected from all candidate task paths that form combinations with it, forming a candidate task list. This candidate task list is sorted from high to low according to the task allocation priority coefficient. Assuming that no other vehicle has occupied a candidate task path, a target task path is selected for the unmanned cleaning vehicle starting from the highest-ranked task path. The corresponding path number is recorded as a new task path for this vehicle in this round of operation, and the task path is marked as assigned in the candidate task set. For other unmanned cleaning vehicles that have not yet obtained a target task path, the search continues in their respective candidate task lists, searching for unassigned task paths from high to low according to the task allocation priority coefficient, and so on, until no suitable task paths remain or all vehicles have been assigned.

[0113] After the task path is assigned, the target task path number is recorded for each unmanned sweeping vehicle, and the corresponding overlap distance ratio, service delay offset time and task assignment priority coefficient are written into the scheduling result record. This record will be used in step S5 to adjust the re-sweeping area identification rules and the re-sweeping task path generation strategy based on the actual execution effect.

[0114] Through the sub-steps of step S4, based on the set of paths for tasks to be re-sweeped and the set of paths for regular cleaning tasks, the ratio of overlapping distances along the route is introduced to describe the degree of overlap between candidate task paths and the current remaining planned paths. Service delay offset time is introduced to describe the service delay of the most lagging road segment within the task path. Then, the scheduling matching model is used to combine the two into a task allocation priority coefficient, and based on this, the target task path is selected from the candidate task paths for each unmanned cleaning vehicle. This ensures that the unmanned cleaning vehicle fleet can insert tasks to be re-sweeped along the route as much as possible during execution, while also giving priority to road segments with severe service delays. This provides a solid data foundation and a clear task execution trajectory for the effect evaluation and rule iteration in step S5.

[0115] Steps S1 to S4 have completed the entire process from collecting road segment status data, identifying areas to be re-sweeped, generating re-sweeping task paths, to allocating tasks based on the ratio of overlapping distances along the road and service delay offset time, forming a round of scheduling and issuing target task paths and the execution schedule of the unmanned sweeper. However, if the actual sweeping trajectory is not used to backtrack and evaluate the aforementioned data structure after the operation, the sweeping and obstacle markers in the basic labels cannot continuously match the real road conditions, and the quality of task allocation priority coefficients cannot be fed back into the re-sweeping area identification rules and the re-sweeping task path generation strategy, making it difficult for the overall scheduling logic to form a closed loop. Step S5 is precisely after the operation is completed, using the actual sweeping trajectory reported by the unmanned sweeper as the main line, on the one hand, updating the basic label set and the re-sweeping task set, and on the other hand, constructing two parameters around each task path: the path fulfillment integrity ratio and the service delay correction ratio. Then, at the path set level, global indicators and rule adjustment driving coefficients are extracted, so that the subsequent re-sweeping area identification rules and re-sweeping task path generation strategies can be gradually adjusted according to the running results.

[0116] Sub-step S5-1: Update the basic tag set.

[0117] After all tasks in the current work cycle are completed, the actual cleaning trajectory corresponding to this round of operation is obtained from each unmanned sweeper. The trajectory is represented by the position information arranged in chronological order. Based on the geometric information of the centerline of the road segment established in step S1, the actual cleaning trajectory is projected segment by segment onto the road segment set, and each road segment is assigned a number. For any road segment number, the actual cleaning length completed by the unmanned sweeper on that road segment during this round of operation is counted. This length does not exceed the length of the road segment itself. Combining the cumulative cleaning length in historical operations and the newly added cleaning length in this round, the total cleaning coverage ratio of the road segment at the current moment is obtained. This ratio is then compared with the pre-set coverage judgment conditions. When the coverage ratio meets the cleaning judgment conditions, the cleaning mark of the road segment in the basic label set is updated to "cleaned"; otherwise, it remains "uncleaned".

[0118] Regarding obstacle status, the unmanned sweeper continuously reports temporary obstacle detection results during operation, including the time and spatial location of the obstacle's appearance. For each road segment number, it is checked whether temporary obstacles still exist during the current round of operation. If no obstacle detection information matching the location of that road segment appears before the end of the operation, the temporary obstacle is considered to have been cleared, and the obstacle marker for the road segment in the basic tag set is updated to "no temporary obstacle." If obstacle detection information exists throughout the operation, the marker indicating the presence of temporary obstacles remains. Through this process, the sweeping markers and obstacle markers in the basic tag set are realigned with the actual execution results at the end of each round of operation, providing an accurate basis for updating the subsequent set of tasks to be re-sweeped and constructing evaluation parameters.

[0119] Sub-step S5-2: Verification and status adjustment of the set of tasks to be rescanned.

[0120] Based on the updated basic tag set, each record of the area to be rescanned in the rescanning task set is checked. The rescanning task set is indexed by road segment numbers, and each record of the area to be rescanned contains at least the road segment number, cleaning mark snapshot, obstacle mark snapshot, and the time of rescanning identification. For any record of the area to be rescanned, the road segment number is first read from the record, and then the latest cleaning mark and obstacle mark for that road segment are queried from the basic tag set.

[0121] When the cleaning marker in the basic label is "cleaned," it means that the road segment has been cleaned in this round of operation and is no longer an object to be re-cleaned in the future. At this time, the status of the corresponding area to be re-cleaned is adjusted to "completed," and it will not be included in the selectable set in the subsequent re-cleaning identification process. When the cleaning marker in the basic label is still "not cleaned," but the obstacle marker has been updated to "no temporary obstacle," it means that the road segment has not been cleaned in the current operation cycle, but there is no longer any temporary obstacle obstruction. This type of road segment should still be treated as a potential area to be re-cleaned in the next operation cycle and participate in the path merging process. Therefore, its record is retained in the re-cleaning task set, and the re-cleaning identification time field is updated to the end time of the current operation cycle, recording the latest identification time. When the cleaning marker in the basic label is still "not cleaned" and the obstacle marker is "temporary obstacle exists," it means that the temporary obstacle still exists. The corresponding area to be re-cleaned record continues to maintain its original status so that it can participate in the merging of the re-cleaning task path when the obstacle is removed in the future. This round of verification ensures that the content of the task set to be rescanned remains consistent with the basic label set at the end of the job, preventing any lag in the calculation of subsequent evaluation parameters.

[0122] Sub-step S5-3: Calculation of the path fulfillment integrity ratio.

[0123] After the basic tag set and the set of tasks to be re-sweeped are synchronized, the actual performance of each task path needs to be evaluated from the perspective of execution completeness. The task paths come from both the set of tasks to be re-sweeped and the set of regular sweeping tasks. Each path, when assigned, has a clearly defined sequence of road segments and corresponding lengths constituting the planned coverage area. If we only know whether a path has been assigned, without considering what percentage of the path is actually executed, we cannot determine whether the priority coefficient of the previous round of task allocation was reasonable in terms of route length and vehicle capacity matching. The path completion ratio describes the proportional relationship between the planned total length of roads to be swept in the path and the total length of roads actually swept in this round of operation; it is a key parameter for measuring execution completeness.

[0124] In practice, each task path is numbered, and all road segment numbers within that path are retrieved from the corresponding path record. Based on the length information in the road segment set, the lengths of all road segments within the path are summed to obtain the planned path length. This length reflects the total expected cleaning mileage when the path is assigned. After the operation is completed, the actual cleaning behavior of the unmanned sweeper on these road segments during this round of operation is statistically analyzed. For each road segment, the actual cleaning length covered in this round of operation is calculated. If the same segment is traversed repeatedly during actual execution, only the portion not exceeding the length of the road segment itself is counted to prevent double counting due to back-and-forth travel. After summing the actual cleaning lengths of all road segments within the path, the total actual execution length of the path is obtained.

[0125] The path completion ratio ranges from zero to one, and is calculated by dividing the actual total length of the path by the planned length. A ratio close to one indicates that almost all road segments covered by the planned path have been cleaned, demonstrating a high degree of completion. Conversely, a ratio significantly lower than one indicates that a considerable portion of the planned path was not covered in this round of cleaning, resulting in insufficient completion. This parameter helps identify long or complex paths that cannot be completed in a single operation during subsequent rule adjustments, providing a basis for optimizing the path generation strategy for re-scanning tasks.

[0126] Sub-step S5-4: Calculation of the service delay correction ratio.

[0127] Besides determining whether the path was executed completely, it's also necessary to evaluate the degree to which the task path corrected service latency in this round of operation from the perspective of service timeliness. Urban roads and park roads typically have target service intervals set for different road levels, requiring at least one cleaning to be completed within a certain time frame. If a path has accumulated a large service latency offset before scheduling, and the path is executed after scheduling with the most recent cleaning time of the relevant road segment updated, it indicates that the task allocation priority coefficient has played a role in the time dimension. If a path's service latency offset is significantly large before scheduling, but the service latency does not decrease significantly after scheduling, or even increases further, it indicates that the current rules for identifying areas to be re-scanned or the path generation strategy are insufficient in terms of service timeliness. The service latency correction ratio is used to measure the relative change in service latency offset before and after scheduling.

[0128] In step S4, the service delay offset time has been calculated for each task path during scheduling. The calculation involves finding the difference between the actual interval between the current scheduling time and the completion of the most recent cleaning and the corresponding target service interval for each road segment within the path. The largest non-negative value among these differences is selected as the service delay offset time for the entire path. In step S5, at the end of the job, the service delay offset time for each path at the end of the job is calculated again using the same method. At this point, due to the occurrence of this round of cleaning, the most recent cleaning time for some road segments within the path has been updated, and the new service delay offset time is generally no higher than the value at the scheduling time.

[0129] The service delay correction ratio can be understood as the proportion of the original service delay reduced by this round of operation relative to the original service delay. For paths where the service delay offset time before scheduling is greater than zero and the service delay offset time after the operation is no higher than the original value, the service delay correction ratio is obtained by subtracting the service delay offset time at the end of the operation from the original service delay offset time and then dividing by the original service delay offset time. The value is between zero and one; a value close to one indicates that the service delay is almost completely eliminated, while a value close to zero indicates that the delay reduction is limited. For paths where the service delay offset time before scheduling is not greater than zero, or paths where the service delay offset time after scheduling is higher than the original value, the service delay correction ratio is directly recorded as zero, indicating that the delay correction effect of this round of operation on that path is not significant. Through the service delay correction ratio, we can extract the path features that effectively alleviate long-term service lag problems in this round of operation from the time dimension, providing direction for subsequent adjustments to the rules for identifying areas to be rescanned and the path generation strategy.

[0130] Sub-step S5-5: Construction and application of global indicators and rule adjustment driving coefficients.

[0131] The path fulfillment integrity ratio and the service delay correction ratio respectively characterize the degree of path execution integrity and the degree of service delay improvement. However, rule adjustments often need to consider both the overall trend and local performance. Therefore, in addition to analyzing a single path, it is necessary to construct global indicators at the level of the entire path set and form rules to adjust the driving coefficients based on the global indicators.

[0132] Regarding global execution completeness, within the current job cycle, a set of all actually executed task paths is taken. For each path, its planned length and actual execution length are calculated and summed within the set to obtain the total planned length and total actual execution length of all paths. The ratio between the total actual execution length and the total planned length represents the overall path fulfillment status; this ratio is the global path fulfillment completeness index. A value close to one indicates that the vast majority of paths in this job cycle have been executed relatively completely in terms of length; a value that is too low indicates that many paths have not been fully covered in a single job.

[0133] Regarding global service timeliness, the service latency offset of all paths within the same set is statistically analyzed. For paths with positive service latency offsets before scheduling, the sum of their original service latency offsets is recorded as the total global original service latency. Simultaneously, the sum of the reductions in service latency offsets for these paths from the scheduling time to the end of the job is calculated as the total global service latency correction. The overall service latency improvement is expressed as the ratio between the total service latency correction and the total original service latency; this ratio is the global service latency correction index. A value close to one indicates that this round of operation has significantly alleviated the overall service lag problem, while a value close to zero indicates that the overall service latency has not been effectively alleviated.

[0134] The rule adjustment driving coefficient is constructed based on the global path fulfillment integrity index and the global service latency correction index, and the product of the two is used as the comprehensive quantity. When the product is close to one, it indicates that the current operation has achieved a good level in both path execution integrity and service latency improvement, and the existing rules for identifying areas to be rescanned are well coordinated with the path generation strategy and task allocation logic for rescanning tasks, requiring only minor adjustments locally. When the product remains low for several consecutive rounds, it indicates that at least one of the overall execution integrity and service latency improvement has consistently failed to meet expectations. In this case, the distribution of the path fulfillment integrity ratio and service latency correction ratio across various paths should be considered to identify areas with concentrated problems, and the rules in previous steps should be adjusted accordingly. For example, when insufficiently fulfilled paths are concentrated in areas with excessively long road segments or overly complex branch structures, the upper limit of path length and the order of path branch access can be adjusted in the path generation strategy for rescanning tasks. In areas where service latency correction is consistently low, the sensitivity to long-unswept road segments can be increased in the rules for identifying areas to be rescanned, making subsequent rescanning task paths more likely to cover these segments.

[0135] When the path fulfillment integrity ratio is consistently low and occurs frequently in areas where the planned path length exceeds the preset path length limit or the number of branch nodes exceeds the preset branch number limit, the path generation strategy for the task to be rescanned should be adjusted as follows. First, the planned path length, path completion ratio, and number of branch nodes within each task path are statistically analyzed over multiple work cycles. When the planned path length in the same type of road area exceeds the preset path length limit for multiple consecutive rounds and the corresponding path completion ratio remains below the preset completion threshold, the entire road segment is no longer combined into a single path according to the original complete topology during subsequent path generation. Instead, the original long path is split into several sub-paths with lengths not exceeding the preset path length limit, using branch intersection nodes as dividing points. For road structures containing multiple branch nodes, a main path is generated from continuous road segments in the trunk direction, and a separate branch path is generated from continuous road segments in each branch direction. Through this combination of splitting long paths and separating branch paths, the planned length of each task path in a single work cycle is controlled within the range that the vehicle's actual work capacity can complete, thereby improving the path completion ratio in subsequent operations in this type of area.

[0136] When the service latency correction ratio is consistently low and concentrated in areas where the service latency offset time is consistently positive, the rules for identifying areas to be rescanned should be adjusted as follows. First, the service delay offset time change trajectory of each road segment is statistically analyzed over multiple work cycles. Road segments whose service delay offset time consistently exceeds a preset service delay threshold are identified and marked as high-latency-sensitive road segments. In the next round of step S2, when a road segment in the basic label set is marked as uncleaned and has already been marked as a high-latency-sensitive road segment, even if the obstacle marker does not indicate the presence of a temporary obstacle, this road segment is recorded as a re-scanning area and written into the re-scanning task set. In steps S3 and S4, when generating re-scanning task paths and calculating task allocation priority coefficients, for task paths containing high-latency-sensitive road segments, under the premise of the same service delay offset time, paths containing high-latency-sensitive road segments are prioritized for allocation in this round. This ensures that long-uncleaned road segments are included more frequently in re-scanning task paths in subsequent work cycles, thereby improving the performance of the service delay correction ratio in these areas.

[0137] Through the above processing in step S5, after the operation is completed, the cleaning marks and obstacle marks in the basic label set are first updated using the actual cleaning trajectory, and the records of areas to be re-sweeped in the set of tasks to be re-sweeped are adjusted simultaneously. Then, at the task path level, the path fulfillment integrity ratio and the service delay correction ratio are constructed respectively. The former describes the proportional relationship between the planned path length and the actual execution length, and the latter describes the degree of improvement of the service delay offset time between the two moments before and after scheduling. Then, the global path fulfillment integrity index and the global service delay correction index are extracted from all execution path sets. The combination of the two is used to obtain the rule adjustment driving coefficient, which guides the identification rules of areas to be re-sweeped and the path generation strategy of tasks to be re-sweeped to be iteratively adjusted in the direction of more complete execution and more timely service. This enables the unmanned sweeper area coverage optimization scheduling method to form a closed-loop optimization mechanism in continuous operation. The task allocation priority coefficient used in the previous steps is also continuously verified and corrected through this feedback process.

[0138] Example 2: Figure 2 The present invention provides an unmanned sweeper area coverage optimization scheduling system, comprising:

[0139] Status labeling module: Receives the current location cleaning status and temporary obstacle status reported by the unmanned sweeper, and generates basic labels with cleaning and obstacle marks for road segments;

[0140] Rescan recognition module: Selects small segments that have not been cleaned and have temporary obstacles from the basic tags, generates a record of areas to be rescanned and stores it in the set of tasks to be rescanned;

[0141] Path generation module: When the detection of a small segment of obstacle marker corresponding to the area to be rescanned changes from existing to disappearing, the module takes the adjacent areas to be rescanned from the set of tasks to be rescanned and merges them into the path of the task to be rescanned.

[0142] Priority scheduling module: When allocating the path of the task to be re-sweeped, calculate the ratio of the overlapping distance along the route and the service delay offset time based on the combination of the unmanned sweeper and the task path. Input the two into the predefined scheduling matching model to obtain the task allocation priority coefficient. Based on the task allocation priority coefficient, determine the target task of each vehicle in the task path to be re-sweeped and the regular sweeping task.

[0143] Feedback optimization module: After the operation is completed, the basic labels and the record of areas to be cleaned again are updated according to the actual cleaning trajectory reported by the unmanned sweeper. The recognition rules for areas to be cleaned again and the path generation strategy for tasks to be cleaned again are adjusted according to the execution results of the task allocation priority coefficient.

[0144] Specifically, the above description is only a preferred embodiment of this application and is not intended to limit this application.

[0145] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0146] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for optimizing the scheduling of unmanned sweeping vehicles to cover specific areas, characterized in that: Including the following steps: S1 receives the current location cleaning status and temporary obstacle status reported by the unmanned sweeper, and generates basic labels with cleaning marks and obstacle marks for the road segment; S2: Select small segments that have not been cleaned and have temporary obstacles from the basic tags, generate a record of areas to be re-cleaned and store it in the set of tasks to be re-cleaned; S3, when the small segment of obstacle markers corresponding to the area to be rescanned changes from existing to disappearing, the adjacent areas to be rescanned are taken out from the set of tasks to be rescanned and merged into the path of the task to be rescanned. S4. When allocating the task path to be re-sweeped, calculate the ratio of the overlapping distance along the route and the service delay offset time based on the combination of the unmanned sweeping vehicle and the task path. Input the two into the predefined scheduling matching model to obtain the task allocation priority coefficient. Based on the task allocation priority coefficient, determine the target task of each vehicle in the task path to be re-sweeped and the regular sweeping task. The scheduling platform obtains candidate task paths from the set of task paths to be re-sweeped and the set of regular cleaning task paths. It obtains the current position and remaining planned path of each unmanned cleaning vehicle from the set of unmanned cleaning vehicles. For each combination of unmanned cleaning vehicle and each candidate task path, the platform calculates the ratio of the total length of the road segments shared by the remaining planned path and the candidate task path to the total length of the remaining planned path at the road segment level as the along-road overlap distance ratio. The platform also selects the maximum difference of not less than zero from the difference between the actual service interval from the last cleaning to the current scheduling time and the corresponding target service interval of each road segment in the candidate task path as the service delay offset time. S5 updates the basic labels and the record of areas to be re-sweeped based on the actual sweeping trajectory reported by the unmanned sweeper after the operation is completed. It also adjusts the identification rules for areas to be re-sweeped and the path generation strategy for tasks to be re-sweeped based on the execution results of the task allocation priority coefficient.

2. The method for optimizing the scheduling of unmanned sweeping vehicles for area coverage according to claim 1, characterized in that, Step S1 includes the following: The work area is divided into a set of road segments. Each road segment is assigned a road segment number and its length is recorded. The system receives reports from unmanned sweepers regarding their current location, cleaning status, and temporary obstacle status. The reported records are mapped to the corresponding road segments through position and direction matching. The cumulative cleaning length and temporary obstacle occupancy length are accumulated at the road segment level. The cleaning coverage ratio is obtained based on the ratio of the cumulative cleaning length to the road segment length, and the temporary obstacle occupancy ratio is obtained based on the ratio of the temporary obstacle occupancy length to the road segment length. Based on historical work records and cleaning targets, preset coverage thresholds and preset obstacle thresholds are determined. A cleaning mark is generated by comparing the cleaning coverage ratio with the preset coverage threshold, and an obstacle mark is generated by comparing the temporary obstacle occupancy ratio with the preset obstacle threshold. The road segment number, cleaning mark, and obstacle mark are combined into a basic tag unit and written into a basic tag set.

3. The method for optimizing the scheduling of unmanned sweeping vehicles for area coverage according to claim 2, characterized in that, Step S2 includes the following: The system reads the cleaning and obstacle markers of each road segment from the basic tag set, converts the cleaning markers into numerical cleaning markers, and converts the obstacle markers into numerical obstacle markers. It generates a rescan identifier based on the combination of numerical cleaning and obstacle markers. When the rescan identifier is one, it generates a rescan area record unit containing the road segment number, cleaning marker, obstacle marker, and rescan identification time. The system writes or updates the rescan task set using the road segment number as an index, so that each road segment in the rescan task set corresponds to only one rescan area record unit.

4. The method for optimizing the scheduling of unmanned sweeping vehicle area coverage according to claim 3, characterized in that, Step S3 includes the following: After the basic labels are updated, the latest cleaning mark and the latest obstacle mark are read from the basic label set according to the road segment number. They are compared with the historical obstacle marks stored in the task set according to the road segment number. Road segments marked as having temporary obstacles in the historical obstacle mark, having no temporary obstacles in the latest obstacle mark, and having not been cleaned in the latest cleaning mark are identified as road segments to be cleared and re-sweeped. Based on the geometric information of the centerline of the road segment and the topological connection relationship, the adjacency relationship between the centerline endpoints of the road segments to be cleared and re-sweeped is determined, and the driving direction is continuous. Road segments to be cleared and re-sweeped with adjacency relationship are grouped into the same area group to be cleared and re-sweeped.

5. The method for optimizing the scheduling of unmanned sweeping vehicles for area coverage according to claim 4, characterized in that, Step S3 also Includes the following: Generate a rescan task path covering all road segments within each obstacle clearance and rescan area group according to the continuous order of road segments in the road topology. Record the rescan task path in the rescan task set. In the rescan task set, write path association markers to the rescan area records belonging to each obstacle clearance and rescan area group. Establish a one-to-one correspondence between the rescan task set and the rescan task path set through the road segment number.

6. The method for optimizing the scheduling of unmanned sweeping vehicle area coverage according to claim 5, characterized in that, Step S4 also includes the following: The ratio of overlapping distances along the route and the service delay offset time are input into a predefined scheduling and matching model to obtain the task allocation priority coefficient. The candidate task paths corresponding to each unmanned sweeper are sorted according to the task allocation priority coefficient. The candidate task path with the highest task allocation priority coefficient and which is not occupied is selected as the target task path for each unmanned sweeper. The task allocation is completed between the task path to be re-sweeped and the regular sweeping task path.

7. The method for optimizing the scheduling of unmanned sweeping vehicles for area coverage according to claim 6, characterized in that, Step S5 includes the following: After the operation is completed, the scheduling platform updates the cleaning and obstacle markers in the basic tag set according to the actual cleaning trajectory reported by the unmanned sweeper and maps the road segments. Based on this, it checks the records of areas to be re-sweeped in the set of tasks to be re-sweeped one by one. Records marked as cleaned are marked as completed and removed from the set of tasks to be re-sweeped. Records marked as uncleaned and with obstacles marked as having no temporary obstacles are retained and their re-sweeping identification time is updated. Records marked as uncleaned and with obstacles marked as having temporary obstacles are retained. On this basis, for each task path, the path fulfillment completeness ratio is obtained by the ratio of the sum of the planned lengths of the road segments in the path to the sum of the actual cleaning lengths in this round of operation. The service delay correction ratio is obtained by the ratio of the difference between the service delay offset time before scheduling and the service delay offset time after the operation to the service delay offset time before scheduling.

8. The method for optimizing the scheduling of unmanned sweeping vehicles for area coverage according to claim 7, characterized in that, Step S5 also includes the following: A global path fulfillment integrity index is formed by summarizing the planned path length and actual cleaning length of each task path across the entire set of executed task paths. At the same time, a global service delay correction index is formed by summarizing the original total and corrected total service delay offset time of each task path. The driving coefficients for rule adjustment are constructed using the global path fulfillment integrity index and the global service delay correction index. Based on the distribution of the driving coefficients, path fulfillment integrity ratio and service delay correction ratio on different task paths, the rules for identifying areas to be re-scanned and the strategy for generating task paths to be re-scanned are iteratively adjusted.

9. An unmanned sweeper area coverage optimization scheduling system, used to implement the unmanned sweeper area coverage optimization scheduling method according to any one of claims 1-8, characterized in that, include: Status labeling module: Receives the current location cleaning status and temporary obstacle status reported by the unmanned sweeper, and generates basic labels with cleaning and obstacle marks for road segments; Rescan recognition module: Selects small segments that have not been cleaned and have temporary obstacles from the basic tags, generates a record of areas to be rescanned and stores it in the set of tasks to be rescanned; Path generation module: When the detection of a small segment of obstacle marker corresponding to the area to be rescanned changes from existing to disappearing, the module takes the adjacent areas to be rescanned from the set of tasks to be rescanned and merges them into the path of the task to be rescanned. Priority scheduling module: When allocating the path of the task to be re-sweeped, calculate the ratio of the overlapping distance along the route and the service delay offset time based on the combination of the unmanned sweeper and the task path. Input the two into the predefined scheduling matching model to obtain the task allocation priority coefficient. Based on the task allocation priority coefficient, determine the target task of each vehicle in the task path to be re-sweeped and the regular sweeping task. Feedback optimization module: After the operation is completed, the basic labels and the record of areas to be cleaned again are updated according to the actual cleaning trajectory reported by the unmanned sweeper. The recognition rules for areas to be cleaned again and the path generation strategy for tasks to be cleaned again are adjusted according to the execution results of the task allocation priority coefficient.

Images