Human resource outsourcing supervision method and system based on edge computing

Abstract

This invention discloses a human resource outsourcing supervision method and system based on edge computing, specifically relating to the fields of edge computing and human resource supervision technology. It addresses the problem that existing edge nodes rely on unreliable scheduling baseline data provided by outsourcing service providers, leading to unreliable supervisory conclusions. The method involves acquiring scheduling baseline data through edge nodes deployed at outsourcing work sites, collecting the actual spatiotemporal trajectories and behavioral event sequences of outsourced personnel, calculating spatial matching features, and generating logical consistency indicators using relative entropy. The system then assesses the credibility of the scheduling baseline data and, based on this assessment, adopts or corrects the data before performing compliance checks, enabling the edge end to autonomously verify the credibility of the baseline data.

Description

Technical Field

[0001] This invention relates to the fields of edge computing and human resource supervision technology, and more specifically, to a human resource outsourcing supervision method and system based on edge computing. Background Technology

[0002] In the field of human resource outsourcing supervision, outsourcing companies typically deploy edge computing nodes at the service provider's work sites. These nodes collect and process data such as attendance records, behavior monitoring, and task status at the edge, enabling real-time supervision of dispersed outsourced personnel. When performing compliance checks, edge nodes rely on dynamic baseline information provided by the outsourcing service provider's local management system, such as work schedules, job assignments, and task requirements. This baseline information serves as a reference for judging whether personnel behavior complies with regulations. In existing solutions, edge nodes by default receive the aforementioned baseline data from the service provider's management system, generate supervisory conclusions based on this data, and then upload the results to the outsourcing company's central platform. Because the edge computing architecture migrates data processing to the network edge closer to the data source, outsourcing companies can achieve parallel supervision of outsourcing activities across multiple locations while reducing the load on the central end.

[0003] The existing solutions mentioned above rely on benchmark information provided by outsourcing service providers as a premise for regulatory judgment. However, they do not take into account that service providers may provide delayed or inaccurate benchmark data for their own benefit. This makes compliance testing at the edge nodes lose a reliable reference basis, resulting in regulatory conclusions based on unreliable benchmarks that lack reliability. Outsourcing companies find it difficult to ensure from a technical perspective that the regulatory basis obtained at the edge truly reflects the actual work arrangements and requirements, thereby weakening the credibility of the overall regulatory system. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a human resource outsourcing supervision method and system based on edge computing to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A human resources outsourcing supervision method based on edge computing includes the following steps: S1: Edge nodes deployed at outsourced work sites acquire scheduling baseline data provided by the outsourcing service provider's management system. The scheduling baseline data includes the types of tasks that outsourced personnel should perform. S2: Edge nodes collect the actual spatiotemporal trajectories and behavioral event sequences of outsourced personnel through on-site sensing terminals; S3: Determine the preset spatial constraint range according to the type of task to be performed, calculate the proportion of the actual spatiotemporal trajectory falling into the preset spatial constraint range and the dispersion of the stationary point distribution, and generate spatial matching features; S4: Align the sequence of behavioral events according to the time windows divided by the preset task rhythm constraints corresponding to the type of task to be executed, count the frequency of events in each time window to obtain the observation frequency distribution, and calculate the relative entropy between the observation frequency distribution and the baseline event frequency distribution corresponding to the type of task to be executed as a logical consistency index. S5: Evaluate the credibility of scheduling baseline data based on spatial matching characteristics and logical consistency indicators. Adopt scheduling baseline data based on credibility or modify scheduling baseline data using actual spatiotemporal trajectories and behavioral event sequences before performing compliance checks.

[0006] Furthermore, edge nodes deployed at outsourced work sites acquire scheduling baseline data provided by the outsourcing service provider's management system. The scheduling baseline data includes the types of tasks that outsourced personnel should perform. This includes: the edge node sending a scheduling baseline data request to the outsourcing service provider's management system, the scheduling baseline data request carrying a list of outsourced personnel identifications under the edge node's jurisdiction; receiving the scheduling plan text returned by the outsourcing service provider's management system, the scheduling plan text containing the correspondence between outsourced personnel identifications and task item names; matching the task item names with the task type definition library pre-stored locally on the edge node to determine the task type that should be performed for each outsourced personnel identification; and recording the correspondence between outsourced personnel identifications and task types into the scheduling baseline data.

[0007] Furthermore, the edge node collects the actual spatiotemporal trajectory and behavioral event sequence of outsourced personnel through the field sensing terminal. This includes: the edge node using the outsourced personnel's identity identifier in the scheduling baseline data as an index to send a collection instruction containing the outsourced personnel's identity identifier to the field sensing terminal; the field sensing terminal periodically scans the positioning tags carried by the outsourced personnel according to the collection instruction to obtain location points with collection times, sorts the location points according to the collection times to form the actual spatiotemporal trajectory; the field sensing terminal captures the operation behavior records generated by the outsourced personnel during task execution to obtain operation type information and operation occurrence time, and parses the interaction signals between the outsourced personnel and the task equipment to obtain the interaction device identifier and interaction occurrence time; and arranges the operation behavior records and interaction signals in chronological order of operation occurrence time and interaction occurrence time to form a behavioral event sequence.

[0008] Furthermore, the signal strength between the positioning tag carried by the outsourced personnel and the built-in Bluetooth beacon of the mission device is detected. When the received signal strength is higher than the preset near-field threshold, it is determined to be a close-range contact. The interaction device identifier of the mission device to which the Bluetooth beacon belongs and the duration of the signal strength being higher than the preset near-field threshold are recorded. The interaction device identifier, duration, and time of interaction are recorded in the interaction signal.

[0009] Furthermore, based on the type of task to be performed, a preset spatial constraint range is determined. The proportion of position points falling within the preset spatial constraint range and the dispersion of dwell point distribution in the actual spatiotemporal trajectory are calculated to generate spatial matching features. This includes: edge nodes using the type of task to be performed in the scheduling baseline data as the query keyword to search the preset spatial constraint library, obtaining the preset spatial constraint range pre-associated with the type of task to be performed, traversing the position points in the actual spatiotemporal trajectory, determining whether the coordinates of each position point fall within the preset spatial constraint range, accumulating the number of position points falling within the preset spatial constraint range and dividing by the total number of position points in the actual spatiotemporal trajectory to obtain the position point proportion, extracting all position points falling within the preset spatial constraint range to form a valid position point set, calculating the geometric center coordinates of the valid position point set, calculating the standard deviation of the distance from each position point in the valid position point set to the geometric center coordinates as the dwell point distribution dispersion, and combining the position point proportion and the dwell point distribution dispersion into spatial matching features.

[0010] Furthermore, the edge node obtains the acquisition time carried by each location point in the set of valid location points, sorts the set of valid location points according to the order of acquisition time, extracts the time interval between the acquisition times of adjacent location points after sorting, removes location points whose time interval exceeds the preset trajectory break tolerance, and uses the remaining location points after removal as the set of valid location points again.

[0011] Furthermore, the behavioral event sequence is aligned with time windows divided according to the preset task rhythm constraints corresponding to the task type to be executed. The frequency of events within each time window is statistically analyzed to obtain the observation frequency distribution. The relative entropy between the observation frequency distribution and the baseline event frequency distribution corresponding to the task type to be executed is calculated as a logical consistency index. This includes: edge nodes using the task type to be executed in the scheduling baseline data as the query keyword to retrieve the preset task rhythm library, obtain the preset task rhythm constraints and baseline event frequency distribution pre-associated with the task type to be executed, extract the time window sequence arranged chronologically from the preset task rhythm constraints, align the occurrence time of each event in the behavioral event sequence with the time window sequence, group events whose occurrence time falls into the same time window into the same window aligned event subset, statistically analyze the number of events in each window aligned event subset to form the observation frequency distribution, calculate the probability ratio of the number of events corresponding to each time window in the observation frequency distribution to the number of events corresponding to the same time window in the baseline event frequency distribution, take the logarithm of the probability ratio, multiply it by the probability ratio itself, and sum the results. The summation result is used as the logical consistency index.

[0012] Furthermore, the edge nodes extract the operation type information and occurrence time of each event in the action event sequence. When the occurrence time is within the boundary tolerance range of the two time windows, the event is assigned to the window-aligned event subset corresponding to the time window with the higher matching degree, based on the degree of matching between the operation type information and the task rhythm corresponding to the two time windows.

[0013] Furthermore, the credibility of the scheduling baseline data is evaluated based on spatial matching features and logical consistency indicators. The scheduling baseline data is adopted based on the credibility, or it is corrected using actual spatiotemporal trajectories and behavioral event sequences before compliance checks are performed. This includes: edge nodes comparing the proportion of location points in the spatial matching features with a preset proportion threshold, and comparing the dispersion of the dwell point distribution in the spatial matching features with a preset dispersion threshold. If the proportion of location points is higher than the preset proportion threshold and the dispersion of the dwell point distribution is lower than the preset dispersion threshold, then the logical consistency indicator is compared with a preset consistency threshold. If the logical consistency indicator is lower than the preset consistency threshold, the scheduling baseline data is determined to be credible and adopted for compliance checks; otherwise, the scheduling baseline data is determined to be unreliable, and the task type to be executed is deduced using the preset spatial constraint range with the highest density of location points in the actual spatiotemporal trajectory to correct the scheduling baseline data before compliance checks are performed.

[0014] On the other hand, the present invention provides a human resource outsourcing supervision system based on edge computing, comprising the following modules: The data acquisition module is used to acquire scheduling baseline data provided by the outsourcing service provider's management system from edge nodes deployed at the outsourcing work site. The scheduling baseline data includes the types of tasks that outsourced personnel should perform. The sequence acquisition module is used by edge nodes to collect the actual spatiotemporal trajectories and behavioral event sequences of outsourced personnel through on-site sensing terminals; The feature generation module is used to determine the preset spatial constraint range according to the type of task to be performed, calculate the proportion of the actual spatiotemporal trajectory falling into the preset spatial constraint range and the dispersion of the stationary point distribution, and generate spatial matching features. The indicator generation module is used to align the behavioral event sequence with the time windows divided according to the preset task rhythm constraints corresponding to the task type to be executed, count the event frequency in each time window to obtain the observation frequency distribution, and calculate the relative entropy between the observation frequency distribution and the baseline event frequency distribution corresponding to the task type to be executed as a logical consistency indicator. The compliance detection module is used to evaluate the credibility of scheduling baseline data based on spatial matching characteristics and logical consistency indicators. After adopting the scheduling baseline data based on the credibility, or after correcting the scheduling baseline data using actual spatiotemporal trajectories and behavioral event sequences, compliance detection is performed.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. The reliability assessment of scheduling baseline data is decomposed into cross-validation of two independent dimensions: spatial matching features and logical consistency indicators. The edge nodes do not rely on the scheduling baseline data provided by outsourcing service providers as the sole premise for supervision. Instead, they use the actual spatiotemporal trajectories and behavioral event sequences collected by on-site sensing terminals to perform reverse verification of the scheduling baseline data. Spatial matching features measure the degree of consistency between the actual activities of outsourced personnel and the spatial attributes of the claimed tasks from two perspectives: the proportion of location points falling within the preset spatial constraint range and the dispersion of the distribution of dwell points. The logical consistency indicator quantifies the deviation between the behavioral rhythm of outsourced personnel and the temporal attributes of the claimed tasks by the relative entropy of the observation frequency distribution and the baseline event frequency distribution. The verification of both dimensions is completed locally at the edge nodes without relying on central intervention, so that the reliability judgment of the supervision basis can be completed at the data source.

[0016] 2. Based on a comprehensive evaluation mechanism that combines spatial matching features and logical consistency indicators, edge nodes can autonomously determine the reliability of scheduling baseline data. When the data is deemed unreliable, the node uses the pre-defined spatial constraint range with the highest density of location points in the actual spatiotemporal trajectory to infer the type of task to be performed and correct the scheduling baseline data. The correction process directly utilizes the sensing data already collected at the edge, without waiting for the outsourcing service provider to re-provide the baseline information. This ensures that compliance testing is based on cross-validated baseline data, avoiding the distortion of regulatory conclusions caused by outsourcing service providers providing delayed or inaccurate scheduling baseline data. Edge nodes complete the entire process of baseline data verification, correction, and compliance testing locally in a closed loop. Outsourcing companies can obtain reliable regulatory evidence without uploading the original sensing data to the central terminal. Attached Figure Description

[0017] Figure 1 This is a flowchart of a human resource outsourcing supervision method based on edge computing according to the present invention; Figure 2 This is a schematic diagram of the structure of a human resource outsourcing supervision system based on edge computing according to the present invention. Detailed Implementation

[0018] 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: Figure 1 This invention presents a human resource outsourcing supervision method based on edge computing, which includes the following steps: S1: Edge nodes deployed at outsourced work sites acquire scheduling baseline data provided by the outsourcing service provider's management system. The scheduling baseline data includes the types of tasks that outsourced personnel should perform. S2: Edge nodes collect the actual spatiotemporal trajectories and behavioral event sequences of outsourced personnel through on-site sensing terminals; S3: Determine the preset spatial constraint range according to the type of task to be performed, calculate the proportion of the actual spatiotemporal trajectory falling into the preset spatial constraint range and the dispersion of the stationary point distribution, and generate spatial matching features; S4: Align the sequence of behavioral events according to the time windows divided by the preset task rhythm constraints corresponding to the type of task to be executed, count the frequency of events in each time window to obtain the observation frequency distribution, and calculate the relative entropy between the observation frequency distribution and the baseline event frequency distribution corresponding to the type of task to be executed as a logical consistency index. S5: Evaluate the credibility of scheduling baseline data based on spatial matching characteristics and logical consistency indicators. Adopt scheduling baseline data based on credibility or modify scheduling baseline data using actual spatiotemporal trajectories and behavioral event sequences before performing compliance checks.

[0020] In S1, edge nodes deployed at outsourced work sites are computing devices located at the geographical locations where outsourced personnel perform tasks. These computing devices have network communication and data storage capabilities, and establish communication connections with the outsourcing service provider's management system via wired or wireless networks. The outsourcing service provider's management system is a computer system used by the outsourcing service provider to manage the scheduling plans of outsourced personnel. The outsourcing service provider's management system stores the scheduling plan text for each outsourced personnel. The scheduling plan text is a structured data record, and it includes at least an outsourced personnel identification field and a task item name field.

[0021] Before sending scheduling baseline data requests to the outsourcing service provider's management system, the edge node first determines the list of outsourced personnel identifiers under its jurisdiction. This list is obtained by the edge node from a pre-issued monitoring scope configuration file from the client company's central platform. This configuration file contains a complete list of outsourced personnel identifiers assigned to the edge node for monitoring. Each outsourced personnel identifier in the edge node's list serves as a unique identifier for the outsourced personnel. These identifiers can be employee IDs, registered mobile phone numbers, or biometric registration numbers. The outsourced personnel identifiers are agreed upon and synchronized by both the client company and the outsourcing service provider when they sign the outsourcing service contract.

[0022] The edge node sends a scheduling baseline data request to the outsourcing service provider management system. This request is a network message conforming to a pre-agreed data exchange protocol between the outsourcing company and the outsourcing service provider. The request carries a list of outsourced personnel identifiers under the edge node's jurisdiction, and the order of the fields in this list matches the order in the monitoring scope configuration file. Upon receiving the scheduling baseline data request, the outsourcing service provider management system parses the list of outsourced personnel identifiers. Using each identifier as a query condition, it retrieves the scheduling records corresponding to each identifier from the outsourcing service provider's locally stored scheduling database. The system then assembles the identifier and task name fields from these records into a scheduling plan text according to the pre-agreed data exchange protocol format and returns the scheduling plan text to the edge node via the communication network.

[0023] The system receives the scheduling plan text returned by the outsourcing service provider's management system. This text contains the mapping between outsourced personnel identifiers and task names. Each line in the scheduling plan corresponds to one outsourced personnel, and each line must contain at least one outsourced personnel identifier and one task name, separated by a comma or tab. The task name is a description of the specific work task assigned to the outsourced personnel by the outsourcing service provider. The task name is filled in by the outsourcing service provider according to its internal task naming rules. For example, a task name could be "Picking in Warehouse A," "Assembling at Production Line Station 3," "Early Shift Cashier at the Store," or "Sorting and Scanning in the Logistics South District."

[0024] Edge nodes pre-store a task type definition library locally. This library is a data table downloaded from the outsourcing enterprise's central platform and stored on local non-volatile storage during the deployment initialization phase. The task type definition library contains multiple mapping records, each containing a task item name and a corresponding task type. The library is constructed by the outsourcing enterprise, based on the service terms in the outsourcing service contract with the outsourcing service provider, pre-compiling a list of various task item names used by the outsourcing service provider in their business scenarios. Task item names with similar or identical meanings are grouped into a unified task type, forming a mapping relationship from task item name to task type, which is then written into the task type definition library. For example, the task item name "Warehouse A Picking" used by outsourcing service provider A and the task item name "Warehouse A Sorting" used by outsourcing service provider B both mean warehouse sorting operations; both are mapped to the same task type, "Warehouse Sorting," in the task type definition library. Updates to the task type definition library are completed by the outsourcing enterprise's central platform sending incremental update data to the edge nodes when it detects that an outsourcing service provider has added or changed a task item name.

[0025] The task item name corresponding to each outsourced personnel's identity in the scheduling plan text is matched with the mapping record in the task type definition library pre-stored locally on the edge node. The matching method is to use the task item name as the query keyword to search in the task type definition library. If the task item name is exactly the same as the task item name of a mapping record in the task type definition library, the corresponding task type to be executed in this mapping record is taken as the matching result. If no record with an exactly matching task item name is found in the task type definition library, the edge node sends a task type parsing request to the outsourcing enterprise center platform with the task item name as a request parameter. The outsourcing enterprise center platform searches for the corresponding task type to be executed in the global task type definition library and returns it to the edge node. The edge node supplements the correspondence between this task item name and the task type to be executed in the pre-stored task type definition library locally.

[0026] After determining the corresponding task types for each outsourced worker's identity, the mapping between outsourced worker identity and task type is recorded in the scheduling baseline data. The scheduling baseline data is a data structure built by the edge node in its local memory. It is stored in key-value pairs, where the key is the outsourced worker's identity and the value is the task type. An example key-value pair in the scheduling baseline data is outsourced worker identity WB20240501 corresponding to the task type warehouse picking. After the scheduling baseline data is built, the edge node persists it to a local log file for subsequent steps and compliance checks.

[0027] In S2, edge nodes and field sensing terminals establish a communication connection via a wireless local area network. The field sensing terminals are devices deployed at outsourced work sites that possess positioning signal reception and operational behavior collection capabilities. These terminals can be intelligent gateway devices integrating an ultra-wideband positioning base station and a Bluetooth signal receiving module. The positioning tags carried by outsourced personnel are wearable positioning devices issued to them. These tags periodically emit ultra-wideband pulse signals or Bluetooth broadcast frames. The identifier of the positioning tag and the outsourced personnel's identity are bound and paired during edge node deployment initialization, and this binding and pairing relationship is stored in the local configuration table of the field sensing terminal.

[0028] The edge node extracts all outsourced personnel identification identifiers from the scheduling baseline data and fills each identifier into the identifier field of the acquisition command. The acquisition command is a control message format agreed upon between the edge node and the field sensing terminal, containing a command type identifier and a list of outsourced personnel identification identifiers. The edge node sends the acquisition command to the field sensing terminal via Wi-Fi. After parsing the acquisition command, the field sensing terminal extracts the list of outsourced personnel identification identifiers and uses each outsourced personnel identification identifier in the list as an index to query the local configuration table, obtain the identifier of the location tag bound to the outsourced personnel identification identifier, and add the location tag identifier to the scanning and listening queue.

[0029] The on-site sensing terminal periodically scans the location tags in the scanning and monitoring queue. The time interval for each scan is set according to the task rhythm characteristics of the outsourced work site. The method for setting the time interval is as follows: the outsourcing company divides the working time into several time periods based on the work duration and task switching frequency stipulated in the outsourcing service contract. Each time period corresponds to a scanning cycle. The value of the scanning cycle ensures that at least three location points can be acquired during the execution of a single task. For example, if the task switching frequency is once per hour and a single task takes approximately 50 minutes to complete, the scanning cycle is set to 15 minutes. During each scan, the on-site sensing terminal receives pulse signals or broadcast frames emitted by the location tags. Based on the arrival time difference of the pulse signals or the received signal strength of the broadcast frames, it calculates the current spatial coordinates of the location tags. The spatial coordinates, along with the system clock reading at the moment the scan is completed, are used as the acquisition time to form a location point with the acquisition time. The data structure of the location point with the acquisition time includes the outsourced personnel identification field, longitude field, latitude field, floor height field, and acquisition time field.

[0030] The field sensing terminal will continuously scan and acquire all location points with acquisition times, arranging them in ascending order of acquisition time to form an actual spatiotemporal trajectory. The actual spatiotemporal trajectory is a sequence of location points arranged in chronological order. Each location point in the actual spatiotemporal trajectory uniquely corresponds to an acquisition time and a set of spatial coordinates. The actual spatiotemporal trajectory of outsourced personnel is stored in the local cache of the field sensing terminal and periodically synchronized to the edge nodes via a wireless local area network.

[0031] The method for capturing operation behavior records generated during outsourced personnel's task execution is as follows: the on-site sensing terminal is connected to the task terminal at the outsourced work site via a wired interface or near-field communication interface. The task terminal is the device operated by the outsourced personnel during task execution, such as a barcode scanner, workstation touchscreen, electronic scale, or RFID reader / writer. When the outsourced personnel perform an operation on the task terminal, the task terminal generates an operation behavior record, which includes operation type information, operation time, and operator identification. The on-site sensing terminal obtains the operation behavior records from the task terminal through polling or interruption, parses the operator identification in the operation behavior records, and compares the operator identification with the outsourced personnel's identity identifier. If the operator identification matches the outsourced personnel's identity identifier, the operation type information and operation time in the operation behavior record are retained; if the operator identification does not match the outsourced personnel's identity identifier, the operation behavior record is discarded.

[0032] The method for parsing the interaction signals between outsourced personnel and task equipment is as follows: The task equipment has a built-in Bluetooth beacon that periodically broadcasts Bluetooth broadcast frames at a fixed transmission power. Each broadcast frame contains the task equipment's interaction device identifier, a unique hardware number, which is programmed and fixed by the task equipment manufacturer at the factory. The positioning tag carried by the outsourced personnel has a built-in Bluetooth receiver module. When the tag receives a Bluetooth broadcast frame from the task equipment's built-in Bluetooth beacon, it extracts the interaction device identifier and measures the received signal strength. The tag then transmits the interaction device identifier and the received signal strength back to the field sensing terminal via its own uplink communication channel. The moment of this transmission is the moment the interaction occurred.

[0033] The received signal strength is compared with a preset near-field threshold, which is set based on the normal operating distance between the outsourced personnel and the task equipment, and the Bluetooth beacon transmission power. The preset near-field threshold is set by having the outsourcing company's technicians measure the typical distance between the positioning tag and the built-in Bluetooth beacon of the task equipment when the outsourced personnel are standing and operating it during the deployment phase at the outsourced work site. The reference received signal strength is calculated as: RSSIref = Ptx - (20 × log10(d) + 20 × log10(f) + 32.44); where RSSIref... f represents the reference received signal strength in dBm, Ptx represents the Bluetooth beacon transmit power in dBm, d represents the typical distance between the positioning tag and the built-in Bluetooth beacon of the task equipment when the outsourced personnel are standing and operating the task equipment in meters, and f represents the Bluetooth beacon operating frequency in MHz. The RSSIref is floated downwards by, for example, 2dBm to 3dBm as a preset near-field threshold to ensure that the received signal strength of the outsourced personnel in their normal operating position is consistently higher than the preset near-field threshold, while the received signal strength is lower than the preset near-field threshold when the outsourced personnel are far from the task equipment. When the received signal strength is higher than the preset near-field threshold, the field sensing terminal determines that the outsourced personnel and the task equipment are in close contact, records the interaction device identifier of the task equipment to which the Bluetooth beacon belongs, and starts timing. Timing stops when the received signal strength falls below the preset near-field threshold, obtaining the duration of the signal strength being higher than the preset near-field threshold. The field sensing terminal combines the interaction device identifier, duration, and time of interaction into an interaction signal.

[0034] The operation type information and operation occurrence time from the operation behavior records obtained from the task terminal are combined with the interaction device identifier, duration, and interaction occurrence time from the interaction signals parsed from the positioning tag's feedback signals. These are arranged chronologically according to the operation occurrence time and the interaction occurrence time, forming a behavior event sequence. Each element in the behavior event sequence is an event record, and each event record is sorted by its occurrence time. The behavior event sequence includes events from both operation behavior records and interaction signals; these two types of events are mixed and sorted according to their respective occurrence times to form a complete behavior event sequence. The field sensing terminal uploads the actual spatiotemporal trajectory and behavior event sequence to the edge node via a wireless local area network. The edge node stores the actual spatiotemporal trajectory and behavior event sequence for use in subsequent steps.

[0035] In S3, edge nodes extract the task types corresponding to the outsourced personnel's identities from the scheduling baseline data, and then retrieve the preset spatial constraint library using the task type as the query keyword. The preset spatial constraint library is a spatial constraint mapping table downloaded from the outsourcing enterprise's central platform and stored on local non-volatile storage media by the edge nodes during the deployment initialization phase. The preset spatial constraint library contains multiple mapping records, each containing a task type and a preset spatial constraint range. The pre-defined spatial constraint range is constructed as follows: During the deployment phase at the outsourced work site, the outsourcing company's technical personnel conduct on-site surveying of the work area corresponding to each task type. Using surveying equipment, they collect the spatial coordinate sequences of the work area boundaries. The closed polygonal or cuboid space enclosed by these coordinate sequences is stored in the pre-defined spatial constraint library as the pre-defined spatial constraint range. For example, when the task type is warehouse sorting, the pre-defined spatial constraint range is the area enclosed by the latitude and longitude coordinates of the rectangular boundary vertices of the warehouse sorting work area floor. Similarly, when the task type is production line assembly, the pre-defined spatial constraint range is the cuboid space jointly defined by the boundary vertices of the rectangular workbench of the production line assembly station and the boundary vertices of the workstation floor area. The pre-defined spatial constraint library is updated by the outsourcing company's central platform when the layout at the outsourced work site changes, by sending updated spatial constraint mapping records to the edge nodes.

[0036] The system iterates through the location points in the actual spatiotemporal trajectory. Each location point contains longitude, latitude, and floor height fields. It then determines whether the coordinates of each location point fall within a preset spatial constraint range. The determination method is as follows: If the preset spatial constraint range is a two-dimensional polygonal region, a ray is drawn from the horizontal coordinates of the location point in any direction. The number of intersections between the ray and the boundary of the preset spatial constraint range is counted. If the number of intersections is odd, the location point is considered to fall within the preset spatial constraint range; if the number of intersections is even, the location point is considered not to fall within the preset spatial constraint range. If the preset spatial constraint range is a three-dimensional cuboid space, a height dimension comparison is added to the two-dimensional determination. When the floor height field value of the location point is between the lower and upper boundaries of the height of the preset spatial constraint range, the horizontal direction inclusion determination result is combined to jointly determine whether the location point falls within the preset spatial constraint range.

[0037] The cumulative number of location points falling within the preset spatial constraint range is divided by the total number of location points in the actual spatiotemporal trajectory to obtain the location point percentage. The location point percentage is calculated as: R = Nin / Ntotal; where R represents the location point percentage with a dimension of 1, Nin represents the cumulative number of location points falling within the preset spatial constraint range, and Ntotal represents the total number of location points in the actual spatiotemporal trajectory.

[0038] Edge nodes extract all location points falling within a preset spatial constraint range to form a valid location point set. The valid location point set is a subset of the actual spatiotemporal trajectory, and each location point in the valid location point set satisfies the criterion that its coordinates fall within the preset spatial constraint range. Edge nodes obtain the acquisition time carried by each location point in the valid location point set and sort the valid location point set in ascending order of acquisition time to obtain a sorted sequence of valid location points.

[0039] The edge node extracts the time interval between the acquisition times of two adjacent position points in the sorted valid position point sequence. The time interval is calculated as: Δti = t(i+1) - ti; where Δti represents the time interval between the i-th position point and the (i+1)-th position point after sorting, in seconds; t(i+1) represents the acquisition time of the (i+1)-th position point after sorting; and ti represents the acquisition time of the i-th position point after sorting. The edge node compares each time interval with a preset trajectory break tolerance. The preset trajectory break tolerance is set based on the periodic scan time interval and the maximum allowable departure time at the outsourced work site. The preset trajectory break tolerance is set by taking twice the periodic scan time interval as the baseline interval tolerance, and adding the baseline interval tolerance to the maximum allowable departure time. The maximum allowable departure time is determined by the outsourcing company according to the reasonable departure time agreed upon in the outsourcing service contract, for example, the maximum allowable departure time is 10 minutes. If the time interval exceeds the preset trajectory break tolerance, it indicates that there is a time break between two adjacent location points that exceeds the reasonable departure time, and the i-th location point after sorting is removed; if the time interval does not exceed the preset trajectory break tolerance, the i-th location point after sorting is retained. The edge nodes will then use all the retained location points after removal as the set of valid location points again.

[0040] The edge nodes calculate the geometric center coordinates of the set of valid location points. The geometric center coordinates are calculated as: xc = Σ(xj) / M, yc = Σ(yj) / M, zc = Σ(zj) / M; where xc, yc, and zc represent the longitude, latitude, and floor height of the geometric center coordinates, respectively, in degrees, degrees, and meters; xj, yj, and zj represent the longitude, latitude, and floor height of the j-th location point in the set of valid location points; and M represents the total number of location points in the set of valid location points. The edge nodes also calculate the distance from each location point in the set of valid location points to the geometric center coordinates. Finally, the edge nodes calculate the standard deviation of the distances from each location point to the geometric center coordinates as the dispersion of the dwell point distribution. The dispersion of the dwell point distribution is calculated as: σs = Σ(dj - dmean). 2 / M, then take the square root of σs to get σ; where σ represents the dispersion of the stationary point distribution in meters, dj represents the distance from the j-th location point to the geometric center coordinates in meters, and dmean represents the average distance of each location point to the geometric center coordinates. dmean is calculated as: dmean=Σ(dj) / M.

[0041] The edge node combines the proportion of location points with the dispersion of dwell point distribution into a spatial matching feature. The spatial matching feature is a feature group containing two elements. The first element of the spatial matching feature is the proportion of location points, and the second element is the dispersion of dwell point distribution. The spatial matching feature is stored in the local memory of the edge node for subsequent steps to call.

[0042] In S4, edge nodes extract the task types corresponding to the outsourced personnel's identities from the scheduling baseline data and retrieve the preset task rhythm library using the task type as the query keyword. The preset task rhythm library is a rhythm mapping table downloaded from the outsourcing company's central platform and stored on local non-volatile storage media during the deployment initialization phase. The library contains multiple mapping records, each containing a task type, a preset task rhythm constraint, and a baseline event frequency distribution. The preset task rhythm constraints and baseline event frequency distribution are constructed as follows: after the outsourcing service contract is signed, the outsourcing company and the outsourcing service provider jointly review the standard operating procedures for each task type. The required steps and time intervals between these steps are extracted from the standard operating procedures and become preset task rhythm constraints. These constraints contain a sequence of time windows arranged chronologically. Each time window corresponds to a stage in the standard operating procedure. The start and end times of each time window in the sequence are calculated with the start time of the standard operating procedure as zero. Adjacent times in the time window sequence... A transition gap is reserved between windows to accommodate normal fluctuations in the operational rhythm. The width of the transition gap is determined by the outsourcing company based on the statistical variance of the operation time for that process stage, for example, rounding up to the nearest minute to twice the statistical variance. Simultaneously, the outsourcing company collects behavioral event records from outsourcing service providers during historical compliant execution periods. For each type of task to be executed, the average number of events occurring within each time window in the historical behavioral event records is calculated. The average number of events occurring in each time window is divided by the sum of the average number of events occurring in all time windows to obtain the baseline event probability for each time window. The baseline event probabilities for all time windows constitute the baseline event frequency distribution. Updates to the preset task rhythm library are completed by the outsourcing company's central platform when standard operating procedures change or outsourcing service provider historical data is updated, by sending updated mapping records to the edge nodes.

[0043] After obtaining the preset task rhythm constraints and baseline event frequency distribution pre-associated with the task type to be executed, the time window sequence arranged chronologically is extracted from the preset task rhythm constraints. Each time window in the time window sequence has a start time and an end time, and the time window sequence covers the complete time span from task start to task end. The edge node traverses each event in the behavioral event sequence. Each event carries occurrence time and operation type information. The edge node compares the occurrence time of each event with the start and end times of each time window in the time window sequence to determine which time window the event falls into.

[0044] When matching the occurrence time with the time window, a boundary tolerance range is set. The boundary tolerance range is set based on the width of the transition gap between adjacent time windows in the time window sequence. The boundary tolerance range is set by taking half the width of the transition gap between adjacent time windows. For example, if the transition gap width is 2 minutes, the boundary tolerance range is set to 1 minute. When the occurrence time falls within the closed interval from the start to the end of a time window and is not within the boundary tolerance range, the event is assigned to the window alignment event subset corresponding to that time window. When the occurrence time falls within the boundary tolerance range of two time windows (i.e., the difference between the occurrence time and the adjacent boundary times of the two adjacent time windows is less than the boundary tolerance range), the edge node extracts the event's operation type information, which is either an operation code or an operation name. The edge node compares the operation type information with the allowed operation type sets associated with each of the two time windows. The allowed operation type sets are the sets of operation actions specified for the corresponding process stage of the time window in the preset task rhythm constraints. The allowed operation type sets are jointly determined by the contracting company and the outsourcing service provider when constructing the preset task rhythm constraints. The comparison method involves determining whether the operation type information appears in the set of allowed operation types associated with a certain time window. If it does, the event is considered a match for that time window; otherwise, it is considered a mismatch. Edge nodes compare the operation type information with the matching results of the two time windows and assign the event to the window-aligned event subset corresponding to the time window that is considered a match. If the operation type information matches or does not match both time windows, the event is assigned to the window-aligned event subset corresponding to the time window whose occurrence time is closer to the middle time of its time window. The middle time of the time window is the arithmetic mean of the start and end times of that time window.

[0045] After the edge nodes complete the alignment of all events in the behavior event sequence, they count the number of events in the aligned event subsets of each window to form an observation frequency distribution. The observation frequency distribution is a sequence of event counts that corresponds one-to-one with the time windows in the time window sequence. The k-th element in the observation frequency distribution represents the total number of events falling into the k-th time window.

[0046] Edge nodes calculate the relative entropy between the observation frequency distribution and the baseline event frequency distribution as a logical consistency index. Edge nodes directly use the probability of occurrence of the baseline event corresponding to each time window in the baseline event frequency distribution as the baseline proportion value, with the sum of the baseline proportion values ​​being 1. Simultaneously, edge nodes divide the number of events corresponding to each time window in the observation frequency distribution by the sum of the number of events in all time windows in the observation frequency distribution to obtain the observation proportion distribution. The observation proportion value corresponding to each time window in the observation proportion distribution has a dimension of 1, and the sum of the observation proportion values ​​is 1. Edge nodes calculate the relative entropy of the observation proportion distribution relative to the baseline proportion distribution. The relative entropy is calculated as: DKL = Σ[Pobs(k) × log(Pobs(k) / Pref(k))]; where DKL represents the logical consistency index, with units of nat, Pobs(k) represents the observation proportion value corresponding to the k-th time window in the observation proportion distribution, Pref(k) represents the baseline proportion value corresponding to the k-th time window in the baseline proportion distribution, log represents the natural logarithm operation, and Σ represents the summation over all time windows. When Pobs(k) is 0, let Pobs(k) × log(Pobs(k) / Pref(k)) equal 0. Relative entropy measures the degree of difference between the actual time distribution of outsourced personnel's behavioral events and the expected time distribution of standard operating procedures. The closer the logical consistency index is to 0, the more consistent the rhythm of outsourced personnel's behavioral events is with the preset task rhythm constraints. Relative entropy is used instead of Euclidean distance or correlation coefficient because it can more sensitively capture anomalies when low-probability events are observed in the probability distribution. In the scenario of human resource outsourcing supervision, outsourced personnel's actions outside the expected time window or omission of key process stages will cause a significant shift in the observed proportion distribution relative to the baseline proportion distribution. Relative entropy can effectively quantify such shifts.

[0047] The edge nodes store the calculated logical consistency metrics in their local memory for use in subsequent steps.

[0048] In S5, edge nodes read spatial matching features and logical consistency indicators from local memory. Spatial matching features include the proportion of location points and the dispersion of dwell point distribution. The edge node compares the proportion of location points in the spatial matching features with a preset proportion threshold. The preset proportion threshold is set according to the minimum on-duty time ratio requirement of outsourcing personnel in the work area by the outsourcing company. The preset proportion threshold is set by the outsourcing company determining the minimum on-duty ratio based on the ratio of the effective working time length stipulated in the outsourcing service contract to the total dispatch time length. The difference between the minimum on-duty ratio multiplied by 1 and the tolerance coefficient is used as the preset proportion threshold. The tolerance coefficient is determined by the outsourcing company based on the average ratio of the allowed reasonable absence time to the total dispatch time in the outsourcing service provider's historical regulatory compliance records. For example, when the minimum on-duty ratio is 80% and the tolerance coefficient is 0.1, the preset proportion threshold is calculated as 80% × 0.9 = 72%. Edge nodes compare the dispersion of dwell point distribution in spatial matching features with a preset dispersion threshold. The preset dispersion threshold is set based on the geometric dimensions of the preset spatial constraint range and the normal activity radius of outsourced personnel when performing tasks within the preset spatial constraint range. The preset dispersion threshold is set by taking the radius of the circumcircle or circumcircle of the preset spatial constraint range as the reference radius. The typical movement radius of outsourced personnel when performing the type of task to be performed is obtained by the contracting company through the observation of the maximum movement distance of outsourced personnel during normal operation during the deployment phase at the outsourced work site. The typical movement radius is divided by the reference radius to obtain the normalized dispersion upper limit. The normalized dispersion upper limit is multiplied by the reference radius to obtain the preset dispersion threshold. For example, if the reference radius is 15 meters and the typical movement radius of outsourced personnel during normal operation is about 5 meters, the normalized dispersion upper limit is about 0.33, and the preset dispersion threshold is set to 15 × 0.33 = 5 meters.

[0049] Edge nodes compare the proportion of location points with a preset proportion threshold and the dispersion of dwell point distribution with a preset dispersion threshold. If the proportion of location points is higher than the preset proportion threshold and the dispersion of dwell point distribution is lower than the preset dispersion threshold, it indicates that the spatial activity range of outsourced personnel matches the preset spatial constraints corresponding to the task type in both coverage ratio and concentration. Edge nodes then compare the logical consistency index with the preset consistency threshold. The preset consistency threshold is set by the outsourcing company based on the maximum allowable deviation between the outsourcing personnel's behavioral rhythm and the standard operating procedure. The preset consistency threshold is set by the outsourcing company collecting historical values ​​of the logical consistency index corresponding to each task type of the outsourcing service provider during the historical compliance supervision period, sorting the historical values ​​from smallest to largest, and taking the value corresponding to the 95th percentile after sorting as the preset consistency threshold. If the amount of historical data is insufficient to calculate the 95th percentile, an initial reference value, such as 0.5, is taken as the preset consistency threshold, and it is updated according to the percentile method after accumulating sufficient historical data.

[0050] If the logical consistency index is lower than the preset consistency threshold, it indicates that the difference between the outsourced personnel's behavioral event rhythm and the preset task rhythm constraints is within the allowable range, and the edge node determines that the scheduling baseline data is reliable. The edge node adopts the scheduling baseline data to perform compliance checks. The compliance check method is as follows: the edge node takes the task type to be performed in the scheduling baseline data as the correct task type that the outsourced personnel should perform, compares the actual spatiotemporal trajectory and behavioral event sequence of the outsourced personnel with the preset spatial constraint range and preset task rhythm constraints corresponding to the task type to be performed, and generates a compliance check result. The compliance check result includes the outsourced personnel's identity, the task type to be performed, the spatial compliance judgment result, and the behavioral rhythm compliance judgment result. The edge node uploads the compliance check result to the outsourcing company's central platform.

[0051] If any one of the following three conditions is met, the edge node determines that the scheduling baseline data is unreliable: the logical consistency index is not lower than the preset consistency threshold, the proportion of location points is not higher than the preset proportion threshold, or the dispersion of the dwell point distribution is not lower than the preset dispersion threshold. After determining that the scheduling baseline data is unreliable, the edge node does not directly adopt it. Instead, it uses the preset spatial constraint range with the highest density of location points in the actual spatiotemporal trajectory to infer the task type to be executed in order to correct the scheduling baseline data. The method of inferring the task type using the preset spatial constraint range with the highest density of location points in the actual spatiotemporal trajectory is as follows: the edge node extracts all preset spatial constraint ranges from the preset spatial constraint library, checks each location point in the actual spatiotemporal trajectory against each preset spatial constraint range, counts the number of location points that fall into each preset spatial constraint range, selects the preset spatial constraint range with the highest number of location points falling into it as the preset spatial constraint range with the highest density of location points, and uses the task type associated with the preset spatial constraint range with the highest density of location points in the preset spatial constraint library as the inferred task type. The edge node replaces the original task types in the scheduling baseline data with the task types derived from reverse calculation to obtain the corrected scheduling baseline data. The edge node then adopts the corrected scheduling baseline data to perform compliance checks, and the compliance check method is the same as the method used to perform compliance checks when adopting the scheduling baseline data.

[0052] Example 2: Figure 2 A schematic diagram of a human resource outsourcing supervision system based on edge computing is provided. This system includes the following modules: The data acquisition module is used to acquire scheduling baseline data provided by the outsourcing service provider's management system from edge nodes deployed at the outsourcing work site. The scheduling baseline data includes the types of tasks that outsourced personnel should perform. The sequence acquisition module is used by edge nodes to collect the actual spatiotemporal trajectories and behavioral event sequences of outsourced personnel through on-site sensing terminals; The feature generation module is used to determine the preset spatial constraint range according to the type of task to be performed, calculate the proportion of the actual spatiotemporal trajectory falling into the preset spatial constraint range and the dispersion of the stationary point distribution, and generate spatial matching features. The indicator generation module is used to align the behavioral event sequence with the time windows divided according to the preset task rhythm constraints corresponding to the task type to be executed, count the event frequency in each time window to obtain the observation frequency distribution, and calculate the relative entropy between the observation frequency distribution and the baseline event frequency distribution corresponding to the task type to be executed as a logical consistency indicator. The compliance detection module is used to evaluate the credibility of scheduling baseline data based on spatial matching characteristics and logical consistency indicators. After adopting the scheduling baseline data based on the credibility, or after correcting the scheduling baseline data using actual spatiotemporal trajectories and behavioral event sequences, compliance detection is performed.

[0053] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0054] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each application, but such implementation should not be considered beyond the scope of this application.

Claims

1. A human resource outsourcing supervision method based on edge computing, characterized in that, Includes the following steps: S1: Edge nodes deployed at outsourced work sites acquire scheduling baseline data provided by the outsourcing service provider's management system. The scheduling baseline data includes the types of tasks that outsourced personnel should perform. S2: Edge nodes collect the actual spatiotemporal trajectories and behavioral event sequences of outsourced personnel through on-site sensing terminals; S3: Determine the preset spatial constraint range according to the type of task to be performed, calculate the proportion of the actual spatiotemporal trajectory falling into the preset spatial constraint range and the dispersion of the stationary point distribution, and generate spatial matching features; S4: Align the sequence of behavioral events according to the time windows divided by the preset task rhythm constraints corresponding to the type of task to be executed, count the frequency of events in each time window to obtain the observation frequency distribution, and calculate the relative entropy between the observation frequency distribution and the baseline event frequency distribution corresponding to the type of task to be executed as a logical consistency index. S5: Evaluate the credibility of scheduling baseline data based on spatial matching characteristics and logical consistency indicators. Adopt scheduling baseline data based on credibility or modify scheduling baseline data using actual spatiotemporal trajectories and behavioral event sequences before performing compliance checks.

2. The human resource outsourcing supervision method based on edge computing according to claim 1, characterized in that, Edge nodes deployed at outsourced work sites acquire scheduling baseline data provided by the outsourcing service provider's management system. The scheduling baseline data includes the task types that outsourced personnel should perform. This includes: the edge node sending a scheduling baseline data request to the outsourcing service provider's management system, the scheduling baseline data request carrying a list of outsourced personnel identifications under the edge node's jurisdiction; receiving the scheduling plan text returned by the outsourcing service provider's management system, the scheduling plan text containing the correspondence between outsourced personnel identifications and task item names; matching the task item names with the task type definition library pre-stored locally on the edge node to determine the task type that should be performed for each outsourced personnel identification; and recording the correspondence between outsourced personnel identifications and task types into the scheduling baseline data.

3. The human resource outsourcing supervision method based on edge computing according to claim 1, characterized in that, Edge nodes collect the actual spatiotemporal trajectories and behavioral event sequences of outsourced personnel through on-site sensing terminals. This includes: edge nodes using the outsourced personnel's identity identifier in the scheduling baseline data as an index to send a collection command containing the outsourced personnel's identity identifier to the on-site sensing terminal; the on-site sensing terminal periodically scans the positioning tags carried by the outsourced personnel according to the collection command to obtain location points with collection times, sorts the location points according to the collection times to form actual spatiotemporal trajectories, captures the operation behavior records generated by the outsourced personnel during task execution to obtain operation type information and operation occurrence time, and parses the interaction signals between the outsourced personnel and task equipment to obtain interaction device identifiers and interaction occurrence times, and arranges the operation behavior records and interaction signals in chronological order of operation occurrence times and interaction occurrence times to form behavioral event sequences.

4. The human resource outsourcing supervision method based on edge computing according to claim 3, characterized in that, The signal strength between the positioning tag carried by the outsourced personnel and the built-in Bluetooth beacon of the mission device is detected. When the received signal strength is higher than the preset near-field threshold, it is determined to be a close contact. The interaction device identifier of the mission device to which the Bluetooth beacon belongs and the duration of the signal strength being higher than the preset near-field threshold are recorded. The interaction device identifier, duration, and time of interaction are recorded in the interaction signal.

5. The human resource outsourcing supervision method based on edge computing according to claim 1, characterized in that, Based on the type of task to be performed, a preset spatial constraint range is determined. The proportion of position points falling within the preset spatial constraint range and the dispersion of dwell point distribution in the actual spatiotemporal trajectory are calculated to generate spatial matching features. This includes: edge nodes using the type of task to be performed in the scheduling baseline data as the query keyword to search the preset spatial constraint library, obtaining the preset spatial constraint ranges pre-associated with the type of task to be performed, traversing the position points in the actual spatiotemporal trajectory, determining whether the coordinates of each position point fall within the preset spatial constraint range, accumulating the number of position points falling within the preset spatial constraint range and dividing by the total number of position points in the actual spatiotemporal trajectory to obtain the position point proportion, extracting all position points falling within the preset spatial constraint range to form a valid position point set, calculating the geometric center coordinates of the valid position point set, calculating the standard deviation of the distance from each position point in the valid position point set to the geometric center coordinates as the dwell point distribution dispersion, and combining the position point proportion and the dwell point distribution dispersion to form a spatial matching feature.

6. The human resource outsourcing supervision method based on edge computing according to claim 5, characterized in that, Edge nodes acquire the acquisition time carried by each location point in the set of valid location points, sort the set of valid location points according to the order of acquisition time, extract the time interval between the acquisition times of adjacent location points after sorting, remove location points whose time interval exceeds the preset trajectory break tolerance, and re-select the remaining location points as the set of valid location points.

7. The human resource outsourcing supervision method based on edge computing according to claim 1, characterized in that, The behavioral event sequence is aligned with time windows divided according to the preset task rhythm constraints corresponding to the task type to be executed. The frequency of events within each time window is counted to obtain the observation frequency distribution. The relative entropy between the observation frequency distribution and the baseline event frequency distribution corresponding to the task type to be executed is calculated as a logical consistency index. This includes: edge nodes using the task type to be executed in the scheduling baseline data as the query keyword to retrieve the preset task rhythm library, obtain the preset task rhythm constraints and baseline event frequency distribution pre-associated with the task type to be executed, extract the time window sequence arranged chronologically from the preset task rhythm constraints, align the occurrence time of each event in the behavioral event sequence with the time window sequence, group events whose occurrence time falls into the same time window into the same window aligned event subset, count the number of events in each window aligned event subset to form the observation frequency distribution, calculate the probability ratio of the number of events corresponding to each time window in the observation frequency distribution to the number of events corresponding to the same time window in the baseline event frequency distribution, take the logarithm of the probability ratio, multiply it by the probability ratio itself and sum it up, and use the summation result as the logical consistency index.

8. A human resource outsourcing supervision method based on edge computing according to claim 7, characterized in that, Edge nodes extract the operation type information and occurrence time of each event in the action event sequence. When the occurrence time is within the boundary tolerance range of two time windows, the event is assigned to the window-aligned event subset corresponding to the time window with the higher matching degree, based on the degree of matching between the operation type information and the task rhythm corresponding to the two time windows.

9. A human resource outsourcing supervision method based on edge computing according to claim 1, characterized in that, The reliability of scheduling baseline data is assessed based on spatial matching features and logical consistency indicators. Based on the reliability, the scheduling baseline data is adopted or corrected using actual spatiotemporal trajectories and behavioral event sequences before compliance checks are performed. This includes: edge nodes comparing the proportion of location points in spatial matching features with a preset proportion threshold, and comparing the dispersion of dwell point distribution in spatial matching features with a preset dispersion threshold. If the proportion of location points is higher than the preset proportion threshold and the dispersion of dwell point distribution is lower than the preset dispersion threshold, then the logical consistency indicator is compared with a preset consistency threshold. If the logical consistency indicator is lower than the preset consistency threshold, the scheduling baseline data is determined to be reliable and adopted for compliance checks. Otherwise, the scheduling baseline data is determined to be unreliable, and the task type to be executed is deduced from the preset spatial constraint range where location points are most densely concentrated in the actual spatiotemporal trajectory to correct the scheduling baseline data before compliance checks are performed.

10. A human resource outsourcing supervision system based on edge computing, used to implement the human resource outsourcing supervision method based on edge computing as described in any one of claims 1-9, characterized in that, Includes the following modules: The data acquisition module is used to acquire scheduling baseline data provided by the outsourcing service provider's management system from edge nodes deployed at the outsourcing work site. The scheduling baseline data includes the types of tasks that outsourced personnel should perform. The sequence acquisition module is used by edge nodes to collect the actual spatiotemporal trajectories and behavioral event sequences of outsourced personnel through on-site sensing terminals; The feature generation module is used to determine the preset spatial constraint range according to the type of task to be performed, calculate the proportion of the actual spatiotemporal trajectory falling into the preset spatial constraint range and the dispersion of the stationary point distribution, and generate spatial matching features. The indicator generation module is used to align the behavioral event sequence with the time windows divided according to the preset task rhythm constraints corresponding to the task type to be executed, count the event frequency in each time window to obtain the observation frequency distribution, and calculate the relative entropy between the observation frequency distribution and the baseline event frequency distribution corresponding to the task type to be executed as a logical consistency indicator. The compliance detection module is used to evaluate the credibility of scheduling baseline data based on spatial matching characteristics and logical consistency indicators. After adopting the scheduling baseline data based on the credibility, or after correcting the scheduling baseline data using actual spatiotemporal trajectories and behavioral event sequences, compliance detection is performed.

Images