A Dynamic Scheduling Method for Scenic Area Service Personnel Based on Real-Time Visitor Data

By using a distributed sensing node array and dynamic weight adjustment, combined with the difference in backlog index between adjacent grids, the problem of lagging pedestrian flow parameters and inaccurate resource allocation in the scheduling of scenic area service personnel was solved, thus achieving precise scheduling and efficient response of scenic area service personnel.

CN123117631APending Publication Date: 2026-11-17ZHEJIANG RIGUANGYU CULTURAL TOURISM DEVELOPMENT CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611274292.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-21
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

The existing methods for dispatching service personnel in scenic areas lack quantitative perception of regional population flow trends, making it impossible to reflect dynamic changes in a timely manner. This leads to distorted allocation of service personnel and an inability to effectively identify service pressure points in local grids, resulting in inaccurate resource allocation.

Method used

By acquiring real-time pedestrian flow data through a distributed array of sensing nodes, dynamically adjusting the calculation weights of pedestrian flow direction parameters, and combining the difference in backlog index between adjacent grids to generate service demand decision labels, precise personnel scheduling can be achieved.

Benefits of technology

It enables sensitive reflection of pedestrian flow and identification of local pressure in scenic areas, improves the accuracy and efficiency of service personnel dispatch, shortens waiting time, and enhances service quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123117631A_ABST
    Figure CN123117631A_ABST
Patent Text Reader

Abstract

This invention discloses a dynamic scheduling method for scenic area service personnel based on real-time data on visitor flow, belonging to the field of intelligent scenic area management technology. The method acquires real-time visitor flow and dwell time data through a distributed sensing node array, calculates visitor flow direction parameters for each grid area after dividing the scenic area into grid partitions, constructs a real-time scheduling backlog index table containing service request queue length and average waiting time using these parameters, generates service demand decision labels based on the difference in backlog indices between the target grid and adjacent grids, and finally generates a global scheduling instruction queue containing service personnel identification codes, target grid coordinates, and scheduling time windows based on the decision labels. This method can dynamically perceive local service pressure differences, generate targeted personnel scheduling instructions, and improve the service response efficiency of the scenic area.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of intelligent management technology for scenic areas, specifically a method for dynamic scheduling of scenic area service personnel based on real-time data on visitor flow. Background Technology

[0002] Large scenic areas face the problem of lagging staff scheduling during peak visitor periods. Existing staff scheduling methods mostly rely on the experience and judgment of managers or on setting fixed thresholds based on peak visitor flow at single points. They lack a quantitative understanding of the flow trends between different areas within the scenic area, making it difficult to pre-schedule staff before service demand is fully apparent.

[0003] Conventional technical solutions use real-time passenger flow and fixed service request data in each area for scheduling. The calculation of regional passenger flow parameters typically uses net flow or density change values ​​directly, with fixed historical weights for smoothing. This approach ignores the differentiated impact of passenger exchange frequency between adjacent areas on current regional passenger flow trends. When abnormal passenger flow occurs in some areas due to activities or congestion, the fixed-weight calculation method cannot reflect this dynamic transfer relationship in a timely manner, causing passenger flow parameter predictions to lag behind actual changes, resulting in distorted subsequent service personnel allocation.

[0004] Meanwhile, existing scheduling decisions typically trigger personnel reinforcement requests solely based on whether the number of service requests or waiting times within a single grid exceed preset values, without considering the relative service pressure between the target grid and its surrounding grids. In scenic areas, the severity of service backlog in a particular grid is often closely related to the capacity status of adjacent grids. Evaluating the backlog of a single grid in isolation can easily lead to an average distribution of scheduling resources across multiple pressure areas, making it difficult to quickly identify the local grids with the most significant service pressure relative to their surrounding areas, and hindering the concentration of service personnel in truly critical locations.

[0005] Accurately characterizing the dynamic impact of inter-regional personnel exchange on local flow trends and utilizing the differences in service pressure between adjacent regions to generate priority-discriminative scheduling decisions are problems that need to be solved to improve the precision of scenic area service personnel scheduling. Summary of the Invention

[0006] This paper proposes a method for dynamic scheduling of scenic area service personnel based on real-time data of people flow. The method dynamically adjusts the calculation weight of people flow direction parameters according to the frequency of people exchange between adjacent grids, and generates service demand decision labels based on the difference in backlog index between the target grid and adjacent grids, so as to realize accurate personnel scheduling by dynamically sensing the differences in regional service pressure.

[0007] To achieve the above objectives, this invention provides the following technical solution: This invention proposes a dynamic scheduling method for scenic area service personnel based on real-time data on pedestrian flow. It acquires real-time pedestrian flow and dwell time data at scenic area entrances, attraction nodes, and public service areas through a distributed sensing node array, enabling dynamic optimization of scenic area service resources. The method includes: utilizing a sensing network composed of near-field communication beacons and infrared pyroelectric sensors to continuously collect data on pedestrian entry and exit numbers and dwell time status, and dividing the scenic area into grid-like zones. For each grid, the number of people entering and leaving within multiple consecutive sampling periods is counted, the net inflow is calculated, and the pedestrian flow density increment is derived by combining the grid area. Then, the current pedestrian flow density increment is weighted and summed with historical data to generate a regional pedestrian flow direction parameter per unit time. The weights are dynamically adjusted according to the frequency of pedestrian exchange between the target grid and adjacent grids, allowing the parameter to more sensitively reflect the intensity of local pedestrian flow interaction.

[0008] Based on regional pedestrian flow parameters, a real-time dispatch backlog index table for public service areas in scenic areas is constructed. The request initiation time and service type code of service facilities within each grid are obtained, and requests are grouped by grid to generate a queue of pending requests. Preferably, when service request data contains an urgency code, requests are sorted from highest to lowest urgency, and then by request initiation time from earliest to latest for requests of the same urgency, forming a priority queue of pending requests. The current service request queue length for each grid is recorded, and the waiting time of the earliest initiated but not yet served request is used as the current average waiting time for that grid. The grid identifier, queue length, and average waiting time are associated and stored in the backlog index table to accurately depict the service pressure distribution in each area.

[0009] Subsequently, service demand decision labels are determined for each grid based on the real-time scheduling backlog index table. The backlog index of the target grid is calculated, which is the current service request queue length multiplied by the current average waiting time, and the average backlog index of all adjacent grids is calculated. The difference is obtained by subtracting the target grid's backlog index from the average backlog index of adjacent grids. When the difference is greater than a preset first threshold, a high-priority service demand decision label is set; when the difference is less than a preset second threshold, a low-priority label is set; and when the difference is between the two, a normal priority label is set. This process enables scheduling decisions to highlight areas of high local pressure while also considering global balance.

[0010] A global scheduling instruction queue is generated based on service demand decision tags. The system retrieves personnel information tables for on-duty service personnel, including their identification code, current location coordinates, current status code, and the type code of the service tool they are carrying. After filtering out low-priority grids, the grids are sorted from highest to lowest backlog index to form a list of grids to be scheduled. Service personnel currently in an idle state are selected as candidate schedulers. For the highest-priority target grid in the list, the idle service personnel closest to the grid and with the best match between their tool type and service request type are selected. The target grid coordinates are assigned, and a scheduling time window is generated. This time window is determined by calculating the theoretical arrival time from the service personnel's current location to the center of the target grid, taking the larger value between this and the target grid's current average waiting time as the lower limit of the scheduling time window, and then multiplying the lower limit by a preset time buffer coefficient to obtain the upper limit. After allocation, a ternary scheduling instruction containing the service personnel's identification code, target grid coordinates, and scheduling time window is stored at the end of the global scheduling instruction queue, and the service personnel is removed from the candidate list. This process is repeated until all grids to be scheduled are allocated or the candidate list is cleared.

[0011] As a technical solution of this invention, the matching degree of service tool types is considered when scheduling service personnel, prioritizing the scheduling of service personnel with the highest tool matching degree among multiple service personnel with similar distances, thereby improving the quality of service response. The entire method forms a closed loop from real-time data collection to personnel scheduling through grid-based pedestrian flow perception, dynamic weighted pedestrian flow direction calculation, priority queue management, and backlog index comparison scheduling decisions. It can effectively adapt to the non-uniform characteristics of the spatiotemporal distribution of pedestrian flow in scenic areas, shorten service waiting time, improve service personnel utilization, and alleviate local service pressure accumulation.

[0012] The technical effects and advantages provided by the present invention in the above technical solution are as follows: In calculating the regional pedestrian flow direction parameters per unit time, the incremental pedestrian flow density is weighted and summed with the historical incremental pedestrian flow density from the previous sampling period. The weight of this weighted sum is dynamically adjusted based on the frequency of pedestrian exchange between the target grid and its neighboring grids. Specifically, a sliding window is used to statistically analyze and normalize the exchange frequency between the target grid and each neighboring grid, which serves as the weight component related to the corresponding neighboring grid in the next period's weighted summation. This scheme ensures that, in the current sampling period, the more frequently neighboring grids exchange people with the target grid, the greater the weight of their historical flow data in the prediction of the next period. This allows the regional pedestrian flow direction parameters to adaptively reflect the differences in cross-regional tourist transfer patterns within the scenic area. When frequent pedestrian exchanges occur between certain neighboring grids due to path changes or events, this dynamic adjustment allows the target grid's pedestrian flow direction parameters to promptly follow changes in actual exchange intensity, avoiding the response lag caused by fixed weights and providing a more quantitative basis for subsequent service scheduling that closely reflects the actual flow trends.

[0013] When generating service demand decision labels for each grid, the service request queue length and average waiting time of the target grid and adjacent grids are extracted from the real-time scheduling backlog index table. A backlog index is constructed, and the average backlog index of adjacent grids is calculated. The grids are then classified into high-priority, normal-priority, and low-priority categories based on the difference between the target grid's backlog index and the average backlog index of adjacent grids. This approach does not use the absolute value of a single threshold as the scheduling trigger condition, but rather measures the degree to which the service pressure of the target grid exceeds that of surrounding grids. This can suppress high-priority alarms across the entire scenic area caused by a synchronous increase in overall visitor flow, concentrating service resources only on local grids where the backlog is significantly higher than the surrounding average level. Service personnel scheduling no longer relies on manual observation or uniform rules across the entire area; it can automatically identify pressure points that require priority response, achieving differentiated and precise deployment with limited service personnel. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0015] Figure 1 This is a flowchart of a method for dynamically scheduling scenic area service personnel based on real-time data on visitor flow; Figure 2 It is a flowchart of dynamic weight adjustment based on sliding window and normalized exchange frequency; Figure 3 This is a flowchart of the process for generating the global dispatch instruction queue for scenic area service personnel. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0017] See Figure 1 This invention provides a method for dynamic scheduling of scenic area service personnel based on real-time data of people flow, including steps one to five.

[0018] Step 1: Deploy a distributed sensing node array at the scenic area entrance, key attractions, and public service areas. This array includes near-field communication beacons and infrared pyroelectric sensors. Near-field communication beacons are used to sense mobile terminals or identification devices carried by tourists entering their effective communication range, recording events of tourists entering and leaving specific sensing ranges. Infrared pyroelectric sensors detect changes in human infrared radiation within the area in a non-contact manner, outputting electrical signals indicating the presence and movement of people. The sensing data from both sensors are aggregated to form real-time pedestrian flow data and pedestrian dwelling data. Real-time pedestrian flow data reflects the number of people passing through the coverage area of ​​each sensing node per unit time and their direction of movement, while pedestrian dwelling data reflects the duration and concentration of tourists within a certain area.

[0019] Step Two: After obtaining real-time pedestrian flow and occupancy data, the scenic area's geographic space is divided into grid-based zones. The scenic area map is evenly divided into several grids according to preset side lengths, and each grid is assigned a unique grid identifier. Based on the sensing data from the sensing nodes associated with each grid, the regional pedestrian flow direction parameters for each grid per unit time are calculated. The regional pedestrian flow direction parameters comprehensively reflect the changing trends of net inflow or outflow of people and pedestrian flow density within that grid, providing a basis for subsequent backlog index calculation and scheduling decisions.

[0020] Step 3: Construct a real-time dispatch backlog index table by combining the service requests received by each service facility within the scenic area's public service area. Service request data, including the request initiation time and service type code, is obtained from service terminals or service request systems within each grid. The service request data is grouped by grid identifier code, forming a queue of pending requests within each grid. The real-time dispatch backlog index table records the current service request queue length and average waiting time for each grid, stored using the grid identifier code as an index.

[0021] Step 4: Based on the real-time scheduling backlog index table, determine the service demand decision label for each grid. For the target grid, calculate its backlog index, and simultaneously obtain the backlog indices of adjacent grids, calculating the difference between the backlog indices. Based on the relationship between the difference and a preset threshold, determine the service demand decision label of the target grid as high priority, low priority, or normal priority. The service demand decision label is used to identify the urgency of service personnel scheduling for each grid.

[0022] Step 5: Generate a global dispatch instruction queue for scenic area service personnel based on service demand decision tags. First, obtain the personnel information table of on-duty service personnel. Then, filter the grids that need to be dispatched according to the service demand decision tags and sort them by backlog index. From the idle service personnel, sequentially match the personnel closest to the grid to be dispatched with the personnel with the best tool matching degree, generate dispatch instructions, and the instructions include the service personnel identification code, target grid coordinates, and dispatch time window. Store them in the global dispatch instruction queue until all grids to be dispatched are assigned or there are no dispatchable personnel.

[0023] Example 1: In practice, the process of calculating the regional pedestrian flow direction parameters for each grid within a unit of time in step two includes the following implementation methods.

[0024] The scenic area map is divided into multiple grids according to preset side lengths, with each grid associated with a unique grid identifier. The preset side lengths are set to fixed values ​​based on the total area of ​​the scenic area and spatial resolution requirements, for example, a side length of 10 meters. The grid identifiers use numerical or string encoding to uniquely identify each grid within the scenic area's geographic information system.

[0025] For the target grid, the number of people entering and exiting is statistically analyzed by the distributed sensing node array over multiple consecutive sampling periods. The sampling period is set to a fixed time interval, for example, 30 seconds. At the end of each sampling period, the near-field communication beacons and infrared pyroelectric sensors in the distributed sensing node array report sensing data. The sensing data includes the number of people entering and leaving the target grid, denoted as a first count value and a second count value, respectively. The first count value represents the total number of people entering the target grid from outside during the current sampling period, and the second count value represents the total number of people leaving the target grid from inside during the current sampling period. The first and second count values ​​are obtained by summing data collected by all sensing nodes deployed at the target grid boundary and within the grid.

[0026] The net inflow to the target grid is calculated based on the first and second count values. The net inflow is obtained by subtracting the second count value from the first count value. A positive net inflow value indicates a net inflow of personnel into the target grid, while a negative net inflow value indicates a net outflow of personnel from the target grid.

[0027] Dividing the net inflow by the area of ​​the target grid yields the increment of personnel flow density for that grid. The area of ​​the target grid is determined by the preset side length of the grid; when the grid is square, the area equals the square of the preset side length. The increment of personnel flow density reflects the intensity of change in net inflow or outflow of personnel per unit area.

[0028] The incremental population flow density is weighted and summed with the historical population flow density increment from the previous sampling period to obtain the population flow direction parameter per unit time area for the target grid. The weighted sum uses dynamically adjusted weighting coefficients, which are adjusted based on the frequency of population exchange between the target grid and adjacent grids. The formula for calculating the weighted sum is: in, Indicates the target mesh at the 1st Human flow direction parameters per unit time area in each sampling period; Indicates the target mesh at the 1st The increment of population flow density calculated from each sampling period; Indicates the target mesh at the 1st Human flow direction parameters per unit time area in each sampling period; This represents the dynamically adjusted weighting coefficient. The value range is a real number greater than 0 and less than 1, and the specific value is determined by the personnel exchange frequency between the target grid and adjacent grids. The weighting coefficient is dynamically adjusted. The method for obtaining the data is as follows: Identify the grids that share a boundary with the target grid, forming a set of adjacent grids; count the number of personnel exchanges between the target grid and each adjacent grid in the set of adjacent grids within a preset observation period; normalize the number of personnel exchanges corresponding to each adjacent grid to obtain a normalized exchange frequency; then use the average of the normalized exchange frequencies corresponding to all adjacent grids as a dynamically adjusted weighting coefficient. The value of is determined by the frequency of personnel exchange between the target grid and its adjacent grids. The average value is larger when the frequency of personnel exchange between the target grid and its adjacent grids is high. The value approaches 1, causing the current increase in population flow density. Pedestrian flow direction parameters within a unit time region The middle part has a relatively large weight; when the frequency of personnel exchange between the target grid and adjacent grids is low, the average value is small. The value approaches 0, making the historical unit time area population flow direction parameter The weight of the parameter for pedestrian flow direction per unit time area is increased. The changes are smoother. By dynamically adjusting the weighting coefficients as described above, the calculated pedestrian flow direction parameters per unit time area can adaptively reflect the dynamic interaction intensity between the target grid and the surrounding grids.

[0029] Example 2: In specific implementation, please refer to Figure 2 The specific implementation method for dynamically adjusting the weights of the weighted sum based on the frequency of personnel exchange between the target grid and adjacent grids is as follows.

[0030] A sliding window of length M is defined, containing the number of personnel exchanges between the target grid and each adjacent grid for M consecutive sampling periods. The number of sampling periods M is set according to the time-varying characteristics of personnel flow in the scenic area; M is set to 10, meaning the sliding window covers the most recent 10 sampling periods. When the sampling period is set to 30 seconds, the time span of the sliding window is 300 seconds. The sliding window moves forward one sampling period after each sampling period, discarding the data from the earliest sampling period and adding the data from the latest sampling period. The number of personnel exchanges refers to the total number of people entering an adjacent grid from the target grid and the total number of people entering the target grid from the same adjacent grid within a sampling period. Each event of personnel crossing the boundary between two grids is jointly detected and recorded by near-field communication beacons and infrared pyroelectric sensors in the distributed sensing node array on both sides of the boundary. The number of personnel exchanges between two grids is obtained by accumulating these records within the sampling period.

[0031] For M sampling periods within the sliding window, the number of personnel exchanges between the target grid and one adjacent grid is counted, yielding the total number of exchanges between the target grid and this adjacent grid within the sliding window. Dividing the total number of exchanges by M gives the exchange frequency between the target grid and this adjacent grid. The exchange frequency represents the average number of personnel exchanges between the target grid and this adjacent grid per unit sampling period. This process is repeated to calculate the exchange frequency between the target grid and each adjacent grid in the adjacent grid set. The adjacent grid set consists of all grids sharing a boundary with the target grid, and the total number of adjacent grids in the adjacent grid set is denoted as K.

[0032] The exchange frequency between the target grid and each of its neighboring grids is normalized. The normalization formula is: in, Represents the target mesh and the set of neighboring meshes. Normalized exchange frequency corresponding to each adjacent grid This represents the index of the adjacent grid within the adjacent grid set. The value of is an integer from 1 to K; Represents the target mesh and the set of neighboring meshes. The exchange frequency between adjacent grids; This represents the total number of adjacent grids in the adjacent grid set; Represents the target mesh and the set of neighboring meshes. The exchange frequency between adjacent grids For the ordinal variable in the summation operation, The value of is an integer from 1 to K; This represents the sum of exchange frequencies between the target mesh and all adjacent meshes in the set of adjacent meshes. After normalization, the normalized exchange frequencies corresponding to all adjacent meshes satisfy... .

[0033] The normalized exchange frequency corresponding to each adjacent grid is used as the weight when the target grid is weighted and summed with its corresponding adjacent grids in the next sampling period for the incremental increase in population flow density. The weighting coefficients are dynamically adjusted for each weighted summation when calculating the population flow direction parameters of the target grid within a unit time region. The weighting factor is determined by the normalized exchange frequencies of all adjacent grid cells. Specifically, it is calculated by averaging the normalized exchange frequencies of all adjacent grid cells and using this average as a dynamically adjusted weighting factor. The value of is determined by the average normalized exchange frequency of all adjacent grids. A higher average value indicates a higher level of personnel exchange activity between the target grid and its neighboring grids. The weighting coefficient is then dynamically adjusted. A larger value indicates that the increase in population flow density during the current sampling period accounts for a larger proportion of the population flow direction parameters per unit time area; when the average normalized exchange frequency corresponding to all adjacent grids is small, it indicates that the overall population exchange activity between the target grid and its adjacent grids is low, and the weighting coefficient is dynamically adjusted. Smaller values ​​enhance the smoothing effect of the historical population flow direction parameter within a given time period.

[0034] Example 3: In practice, the specific implementation process of constructing the real-time scheduling backlog index table for the public service area of ​​the scenic spot in step three is as follows.

[0035] The system acquires service request data for all service facilities within each grid of the scenic area's public service zone. Each service facility within the zone is equipped with a service request terminal, which receives service requests initiated by tourists. The service request data includes the request initiation time and service type code. The request initiation time is recorded by the service request terminal upon receiving the request, with a time accuracy of seconds. The service type code is a predefined numerical code, uniquely corresponding to a specific type of public service within the scenic area. For example, service type code 01 corresponds to cleaning services, service type code 02 to guided tours and information, and service type code 03 to emergency assistance.

[0036] Service request data is grouped according to grid identifier codes. A grid identifier code is a unique identifier assigned to each grid during grid division, represented by a fixed-length string or number. The grouping operation combines all service request data corresponding to the same grid identifier code into the same group, forming a service request data set for that grid. After grouping, each grid corresponds to a service request data set, and each record in the service request data set includes the request initiation time and service type code.

[0037] The service request data within each grid is sorted according to the request initiation time. The sorting rule is as follows: compare the initiation times of every two service request data items, and arrange them in ascending order of initiation time; that is, the service request data with the earliest initiation time is placed at the beginning, and the service request data with the latest initiation time is placed at the end. After sorting, the service request data set is transformed into an ordered queue of pending requests. Each element in the queue is a service request data item, with the first element of the queue being the service request with the earliest initiation time, and the last element being the service request with the latest initiation time.

[0038] Record the length of each pending request queue, which serves as the current service request queue length for the corresponding grid. The length of the pending request queue is the number of service request data elements contained in the queue. The current service request queue length is an integer greater than or equal to zero. When there is no service request data in the pending request queue, the current service request queue length is zero.

[0039] The waiting time for the earliest initiated but not yet served request in each grid is obtained and used as the current average waiting time for that grid. The earliest initiated but not yet served request is the service request corresponding to the first element of the pending request queue. The request initiation time of this service request is obtained, and the time difference between the processing time and the request initiation time is calculated; this time difference is the waiting time. The processing time is the time when the real-time scheduling backlog index table is retrieved. The unit of the current average waiting time is seconds, and the value of the current average waiting time is a real number greater than or equal to zero. When the pending request queue is empty, the current average waiting time is zero.

[0040] The grid identifier, current service request queue length, and current average wait time for each grid are associated and stored in the real-time scheduling backlog index table. The real-time scheduling backlog index table is a dynamically updated data table. Each row in the table corresponds to a grid, and each row contains three columns: the first column is the grid identifier, the second column is the current service request queue length, and the third column is the current average wait time. The real-time scheduling backlog index table is recalculated and updated in each build cycle. The duration of the build cycle is consistent with the duration of the sampling cycle; for example, the build cycle is set to 30 seconds.

[0041] Optionally, the service request data also includes an urgency code. The urgency code is a predefined numeric code used to identify the urgency level of the service request. The value of the urgency code ranges from 1 to 5, where 1 represents the highest urgency and 5 represents the lowest urgency. Before sorting the service request data by request initiation time, it is first sorted by urgency code. The sorting rule is as follows: Prioritize sorting by urgency code from lowest to highest; that is, service request data with urgency code 1 is placed before service request data with urgency code 2. If two or more service request data have the same urgency code, then within the service request data with the same urgency code, they are further sorted by request initiation time from earliest to latest. Through this sorting method, the service request data set is organized into a priority queue of pending requests. The first element of the priority queue is the service request with the highest urgency and earliest initiation time, and the last element is the service request with the lowest urgency and latest initiation time. The length of the priority-based pending request queue and the waiting time of the first element are recorded, serving as the current service request queue length and current average waiting time for the corresponding grid, respectively. The storage method is the same as when there is no urgency coding, where the grid identifier, current service request queue length, and current average waiting time are associated and stored in the real-time scheduling backlog index table.

[0042] Example 4: In practice, the specific implementation process for determining the service demand decision labels for each grid in step four is as follows.

[0043] From the real-time scheduling backlog index table constructed in step three, extract the current service request queue length and current average waiting time of the target mesh. The real-time scheduling backlog index table is queried using the mesh identifier as an index, searching for the row corresponding to the target mesh, and retrieving the values ​​of the current service request queue length and current average waiting time from the row. The current service request queue length of the target mesh is denoted as... The current average waiting time of the target grid is denoted as . . Integers greater than or equal to zero. For real numbers greater than or equal to zero, The unit is seconds.

[0044] Calculate the backpacking index of the target mesh. The formula for calculating the backpacking index is: in, Indicates the backlog index of the target grid; Indicates the current length of the service request queue for the target mesh; This represents the current average waiting time for the target grid. Backlog Index Its dimension is the product of time and quantity. The higher the value, the more severe the backlog of service requests within the target mesh. When The value is zero or When the value is zero, the backlog index The value is zero.

[0045] Calculate the average of the backlog indexes of all adjacent grids of the target grid to obtain the average backlog index of adjacent grids.

[0046] The specific method for calculating the average adjacent grid packing index is as follows: Based on the grid division results in step two, determine the grids that share a boundary with the target grid, and take all the grids that share the boundary as the set of adjacent grids of the target grid. The number of adjacent grids in the adjacent grid set is denoted as... , The value of is determined by the position of the target grid within the grid division. If the target grid is located inside the scenic area grid division region, then The value is 4; if the target grid is located on the boundary of the scenic area grid division area and is not at a corner, then The value is 3; if the target grid is located in the corner of the scenic area grid division area, then The value of is 2.

[0047] The backlog index is retrieved from the real-time scheduling backlog index table to find the current service request queue length and current average waiting time for each neighboring grid in the neighboring grid set. For each neighboring grid in the neighboring grid set, after obtaining its current service request queue length and current average waiting time, the backlog index of the neighboring grid is calculated. The calculation method for the backlog index of the neighboring grid is the same as that for the target grid, which is to multiply the queue length by the average waiting time. The backlog indices of all neighboring grids in the neighboring grid set are added together to obtain the total backlog index. The total backlog index is then divided by... The average of the packing exponents of adjacent grids is obtained, denoted as . The calculation of the mean of the adjacent grid packing index is expressed as follows: ,in Represents the first in the set of adjacent grids The backlog index of adjacent grids, This represents the index of the adjacent grid within the adjacent grid set. The value range is from 1 to Integers.

[0048] The difference in the backlog index is obtained by subtracting the mean backlog index of the target grid from the mean backlog index of its adjacent grids. This difference in backlog index is denoted as... The calculation formula is: . A positive number indicates that the density of the target grid is higher than the average density of its neighboring grids. A negative number indicates that the backlog of the target grid is lower than the average backlog of its neighboring grids.

[0049] Set a preset first threshold and a preset second threshold. The first threshold is denoted as... The second threshold is denoted as , and All are real numbers greater than zero, and The value is greater than The value of the first threshold. The setting is based on: statistical analysis of historical service data of the scenic area, and taking the 90th percentile of the difference in grid backlog index of all historical data as the first threshold. The value of the second threshold. The setting is based on the following: the 10th percentile of the difference in grid backlog indexes in historical data is taken as the second threshold. The value of the first threshold. With the second threshold It is automatically calculated and loaded based on historical data each time the system is initialized, and does not change with real-time data changes.

[0050] The difference in backlog index With the first threshold and the second threshold Compare them.

[0051] If the difference in backlog index Greater than the first threshold ,Right now If so, the service requirement decision label of the target grid will be set to high priority.

[0052] If the difference in backlog index Less than the second threshold ,Right now If so, the service requirement decision label of the target grid will be set to low priority.

[0053] If the difference in backlog index Greater than or equal to the second threshold And less than or equal to the first threshold ,Right now Then, the service requirement decision label of the target grid will be set to normal priority.

[0054] Service demand decision tags are stored in the scheduling decision data table, which is indexed by grid identifier codes. Each row records the grid identifier code and the corresponding service demand decision tag for a grid. The service demand decision tag takes one of three predefined values: "HIGH" for high priority, "LOW" for low priority, and "NORMAL" for normal priority.

[0055] Example 5: In specific implementation, please refer to Figure 3 The specific implementation process of generating the global scheduling instruction queue for scenic area service personnel in step five is as follows.

[0056] Retrieve the personnel information table for all currently on-duty service personnel in the scenic area. The personnel information table is stored in the scenic area's personnel management database. It includes each on-duty service personnel's identification code, current location coordinates, current status code, and the type of service tool they are carrying. The service personnel identification code is a string and is a unique code assigned to each service personnel. The current location coordinates include longitude and latitude coordinates, uploaded to the scenic area's personnel management database by the positioning device carried by the service personnel at a fixed reporting interval of 5 seconds. The current status code is an integer, with values ​​of 0, 1, and 2. 0 represents an idle state, indicating that the service personnel are not currently performing any service tasks and are available for dispatch; 1 represents a service state, indicating that the service personnel are performing assigned service tasks; and 2 represents a temporarily unavailable state, indicating that the service personnel are temporarily unable to be dispatched due to unforeseen circumstances. The service tool type code is a string that records the type of service tool currently carried by the service personnel. There is a predefined mapping relationship between the service tool type code and the service type code. For example, the service tool type code "TOOL01" corresponds to the cleaning service with service type code "01", and the service tool type code "TOOL02" corresponds to the tour guide and consultation service with service type code "02".

[0057] Based on the service demand decision labels of each grid determined in step four, grids with low-priority service demand decision labels are filtered out, resulting in a set of target grids to be scheduled. All grids in the scheduling decision data table are traversed, and the service demand decision label of each grid is read. Grids with a service demand decision label value of "LOW" are removed from the scheduling consideration, while grids with service demand decision labels values ​​of "HIGH" and "NORMAL" are retained. The retained grids constitute the set of target grids to be scheduled.

[0058] The grids in the target grid set to be scheduled are sorted from largest to smallest according to their corresponding backlog index, forming a sorted list of grids to be scheduled. The backlog index of each grid in the target grid set is retrieved from the real-time scheduling backlog index table. The backlog index is calculated by multiplying the current service request queue length by the current average waiting time. Grids are sorted from largest to smallest according to their backlog index values, with the grid with the largest backlog index at the top of the sorted list and the grid with the smallest backlog index at the bottom. If two grids have the same backlog index value, their current average waiting times are compared, and the grid with the larger current average waiting time is sorted first.

[0059] Select service personnel whose current status code is idle from the personnel information table as candidate dispatchers. Iterate through each record in the personnel information table, read the value of the current status code, filter out records with a current status code value of 0, and extract the service personnel identification code, current location coordinates, and service tool type code from the filtered records to form a set of candidate dispatchers.

[0060] For the first target grid in the grid sorting list, select idle service personnel matching the first target grid from the candidate dispatcher set. The selection method is as follows: calculate the Euclidean distance between the current position coordinates of each candidate dispatcher in the candidate dispatcher set and the center point coordinates of the first target grid. The formula for calculating the Euclidean distance is: in, Indicates candidate schedulers With the first target grid The Euclidean distance between them; Indicates candidate schedulers The longitude coordinates of the current location; Indicates candidate schedulers The latitude coordinates of the current location; Indicates the first target grid The longitude coordinates of the center point; Indicates the first target grid The coordinates of the center point are in latitude. The unit of distance is the same as the unit of coordinate values, using the metric system.

[0061] After calculating the Euclidean distances between all candidate schedulers and the first target grid, the candidate scheduler corresponding to the minimum Euclidean distance is identified. If the difference between the Euclidean distance and the minimum value for multiple candidate schedulers is within a preset distance threshold, the matching degree between the service tool type codes carried by the multiple candidate schedulers and the service type codes of the service request data in the pending request queue within the first target grid is further compared. The matching degree is calculated as follows: extract the service type code that appears most frequently in the pending request queue of the first target grid, and determine whether there is a predefined mapping relationship between the service tool type code carried by the candidate scheduler and the service type code that appears most frequently. If a mapping relationship exists, the matching degree is recorded as 1; otherwise, the matching degree is recorded as 0. From the multiple candidate schedulers whose Euclidean distance is within the distance threshold, the candidate scheduler with a matching degree of 1 is selected as the idle service personnel matched with the first target grid. The distance threshold is set at 20 meters. This value is based on the fact that the side length of a grid in the public service area of ​​the scenic spot is set at 10 meters. The distance threshold is the value of the side length of two grids, which can include service personnel within the adjacent grid range in the comparison range for nearby dispatch.

[0062] Assign the grid identifier code of the first target grid and a preset scheduling time window to the idle service personnel that match the first target grid, and generate a scheduling instruction with a length of ternary elements, which include the service personnel identifier code, the target grid coordinates and the scheduling time window.

[0063] The scheduling time window is determined as follows: Obtain the Euclidean distance between the current location coordinates of the idle service personnel matching the first target grid and the center point coordinates of the first target grid. The Euclidean distance is calculated using the same formula as described above. The calculated Euclidean distance is denoted as... Euclidean distance Divide the theoretical arrival time by the preset average movement speed of service personnel within the scenic area. The preset average movement speed is set at 1.2 meters per second, a value determined based on the average of on-site measurements of service personnel's walking speed within the scenic area. Obtain the current average waiting time within the first target grid. The lower limit of the scheduling time window is determined by the greater of the theoretical arrival time and the current average waiting time. The upper limit of the scheduling time window is obtained by multiplying the lower limit by a preset time buffer coefficient. The time buffer coefficient is set to 1.5, a value determined based on the experience required for redundancy in service personnel arrival time in scenic area scheduling management. A buffer of 1.5 ensures that service personnel arrive within the window period even if delays are caused by route congestion or temporary obstacles. The lower and upper limits of the scheduling time window are combined to form the scheduling time window for idle service personnel matching the first target grid. The scheduling time window is represented as a time interval, with the lower limit being the start time of the time interval and the upper limit being the end time of the time interval.

[0064] The generated scheduling instructions are stored at the end of the global scheduling instruction queue. The global scheduling instruction queue is a linear queue data structure, and newly generated scheduling instructions are inserted from the tail of the queue. After the storage operation is completed, the service personnel to be scheduled are removed from the candidate scheduler set, and the number of elements in the candidate scheduler set is reduced by one.

[0065] After generating the scheduling instruction for the first target grid, it is determined whether there are still target grids in the grid to be scheduled that have not been assigned a scheduling instruction and whether the candidate scheduling personnel set is empty. If there is a next target grid in the grid to be scheduled and the candidate scheduling personnel set is not empty, the above operations of selecting matching idle service personnel, calculating the scheduling time window, generating scheduling instructions, and removing scheduled personnel are repeated for the next target grid in the grid to be scheduled. This process continues until all target grids in the grid to be scheduled have been assigned scheduling instructions or the candidate scheduling personnel set becomes empty. When the candidate scheduling personnel set becomes empty and there are still target grids in the grid to be scheduled that have not been assigned a scheduling instruction, the scheduling requests of the target grids without assigned scheduling instructions are left for processing in the next scheduling cycle. The start and end times of the next scheduling cycle are determined by the boundaries of the sampling period.

[0066] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for dynamic scheduling of scenic area service personnel based on real-time data on visitor flow, characterized in that, Includes the following steps: Step 1: Real-time pedestrian flow data and personnel dwell data are acquired by deploying a distributed sensing node array at the scenic area entrance, attraction nodes, and public service areas. The distributed sensing node array includes near-field communication beacons and infrared pyroelectric sensors. Step 2: Based on real-time pedestrian flow data and personnel stay data, the scenic area is divided into grid zones, and the pedestrian flow direction parameters of each grid within a unit of time are calculated. Step 3: Based on the regional pedestrian flow direction parameters, construct a real-time scheduling backlog index table for the public service areas of the scenic area. The real-time scheduling backlog index table includes the current service request queue length and average waiting time for each grid. Step 4: Based on the real-time scheduling backlog index table, determine the service demand decision label for each grid. The service demand decision label is generated based on the difference between the backlog index of the target grid and the backlog index of the adjacent grid. Step 5: Based on the service demand decision tags, generate a global dispatch instruction queue for scenic area service personnel. The global dispatch instruction queue includes the service personnel identification code to be dispatched, the target grid coordinates, and the dispatch time window.

2. The method for dynamic scheduling of scenic area service personnel based on real-time data flow as described in claim 1, characterized in that, Step two involves calculating the regional pedestrian flow direction parameters for each grid unit of time, specifically as follows: The scenic area map is divided into multiple grids according to the preset side length, and each grid is associated with a unique grid identifier code; For the target grid, the number of people entering and exiting the grid is counted in the distributed sensing node array over multiple consecutive sampling periods. The number of people entering and exiting the grid includes a first count value for entering the grid and a second count value for leaving the grid. Based on the first and second count values, calculate the net inflow to the target grid; Divide the net inflow by the area of ​​the target grid to obtain the increase in the population flow density of the target grid; The incremental population flow density is weighted and summed with the historical population flow density increment of the previous sampling period to obtain the population flow direction parameter of the target grid per unit time area. The weight of the weighted summation is dynamically adjusted according to the frequency of population exchange between the target grid and adjacent grids.

3. The method for dynamic scheduling of scenic area service personnel based on real-time data flow as described in claim 2, characterized in that, The weights of the weighted summation are dynamically adjusted based on the frequency of personnel exchange between the target grid and its neighboring grids, specifically: Set a sliding window of length M, which contains the number of personnel exchanges between the target grid and each adjacent grid for M consecutive sampling periods; Calculate the exchange frequency between the target grid and each adjacent grid within a sliding window, where the exchange frequency is equal to the number of personnel exchanges divided by M; The exchange frequency between the target grid and each of its neighboring grids is normalized to obtain the normalized exchange frequency corresponding to each neighboring grid. The normalized exchange frequency is used as the weight when the target grid is weighted and summed with the corresponding adjacent grid in the next sampling period to calculate the incremental population flow density.

4. The method for dynamic scheduling of scenic area service personnel based on real-time data flow as described in claim 1, characterized in that, Step three involves constructing a real-time scheduling backlog index table for the scenic area's public service areas, specifically as follows: Obtain service request data for all service facilities within each grid in the public service area of ​​the scenic spot. The service request data includes the request initiation time and service type code. The service request data is grouped according to the grid identifier code, and the service request data in each grid is sorted according to the time of request initiation to form the pending request queue of that grid. Record the length of each pending request queue, which serves as the current service request queue length for that grid; Get the waiting time of the earliest initiated but not yet served request in each grid, and use it as the current average waiting time for that grid; The grid identifier code, current service request queue length, and current average waiting time of each grid are associated and stored in the real-time scheduling backlog index table.

5. The method for dynamic scheduling of scenic area service personnel based on real-time data flow as described in claim 1, characterized in that, Step four, which determines the service demand decision labels for each grid, specifically involves: From the real-time scheduling backlog index table constructed in step three, extract the current service request queue length and the current average waiting time of the target grid, and calculate the backlog index of the target grid, which is equal to the current service request queue length multiplied by the current average waiting time. Calculate the average of the backlog indexes of all neighboring grids of the target grid to obtain the average backlog index of neighboring grids; The difference in the backlog index is obtained by subtracting the backlog index of the target grid from the average backlog index of the adjacent grids. If the difference in the backlog index is greater than a preset first threshold, the service demand decision label of the target grid is set to high priority. If the difference in the backlog index is less than a preset second threshold, the service demand decision label of the target grid is set to low priority. If the difference in the backlog index is between the first threshold and the second threshold, the service demand decision label of the target grid is set to normal priority.

6. The method for dynamic scheduling of scenic area service personnel based on real-time data flow as described in claim 5, characterized in that, The calculation of the average of the packing exponents of all adjacent grids of the target grid is specifically as follows: Based on the mesh division results in step two, determine the meshes that share the boundary with the target mesh, and use them as the set of adjacent meshes of the target mesh; The current service request queue length and current average waiting time of each grid in the adjacent grid set are queried from the real-time scheduling backlog index table, and the backlog index of each adjacent grid is calculated. The sum of the backlog indices of all adjacent grids in the adjacent grid set is obtained; The average backlog index of adjacent grids is obtained by dividing the sum of the backlog indices by the number of adjacent grids in the adjacent grid set.

7. The method for dynamic scheduling of scenic area service personnel based on real-time data flow as described in claim 1, characterized in that, Step five involves generating a global dispatch instruction queue for scenic area service personnel, specifically as follows: Obtain a personnel information table of all currently on-duty service personnel in the scenic area. The personnel information table includes each service personnel's identification code, current location coordinates, current status code, and service tool type code. Based on the service demand decision labels of each grid determined in step four, grids with low priority service demand decision labels are filtered out to obtain the set of target grids to be scheduled. The grids in the set of target grids to be scheduled are sorted from largest to smallest according to their corresponding backlog index to form a sorted list of grids to be scheduled; Select service personnel whose current status code is idle from the personnel information table and use them as candidate dispatchers; For the first target grid in the grid sorting list to be scheduled, select the idle service personnel closest to the target grid from the candidate scheduling personnel, assign the grid identifier code of the target grid and the preset scheduling time window to the idle service personnel, and generate a scheduling instruction with a length of ternary, wherein the ternary includes the service personnel identifier code, the target grid coordinates and the scheduling time window; The generated scheduling instructions are stored at the end of the global scheduling instruction queue, and the scheduled service personnel are removed from the candidate scheduler list. Repeat the above steps until all target grids in the grid sorting list have been assigned scheduling instructions or the candidate scheduler list is empty.

8. The method for dynamic scheduling of scenic area service personnel based on real-time data of visitor flow according to claim 7, characterized in that, The process of assigning the target grid identifier code and a preset scheduling time window to the idle service personnel is as follows: Obtain the Euclidean distance between the current location coordinates of the idle service personnel and the center point coordinates of the target grid; Divide the Euclidean distance by the preset average moving speed of the service personnel in the scenic area to obtain the theoretical arrival time. Obtain the current average waiting time within the target grid, and use the larger value between the theoretical arrival time and the current average waiting time as the lower limit of the scheduling time window; Multiply the lower limit of the scheduling time window by a preset time buffer coefficient to obtain the upper limit of the scheduling time window; The lower and upper limits of the scheduling time window are combined to form the scheduling time window for the idle service personnel.

9. The method for dynamic scheduling of scenic area service personnel based on real-time data flow as described in claim 4, characterized in that, The service request data also includes an urgency level code. When sorting the service request data according to the time the request was initiated, it is first sorted from high to low according to the urgency level code. Under the same urgency level, it is then sorted from early to late according to the time the request was initiated, thus forming a priority queue of pending requests.

10. The method for dynamic scheduling of scenic area service personnel based on real-time data flow as described in claim 7, characterized in that, When selecting the idle service personnel closest to the target grid from the candidate dispatch personnel, the matching degree between the service tool type code carried by the service personnel and the service type code in the service request data within the target grid is also considered. If there are multiple service personnel that are close to each other, the service personnel with the highest tool matching degree are selected first.