Intelligent personnel scheduling management method and system based on internet of things

Abstract

The present application relates to the technical field of Internet of Things, in particular to a personnel intelligent scheduling management method based on Internet of Things, comprising: acquiring personnel position, physiology, behavior and other multi-source data, and fusing surrounding personnel and historical relationship to calculate individual risk score; generating aggregated risk topology map through preset kernel function; analyzing risk peak in the map, matching optimal performer through multi-objective optimization algorithm and generating scheduling instruction. The present application converts discrete individual risk into intuitive global risk view, and realizes automation and optimization of scheduling decision based on dynamic analysis of risk situation, thereby improving risk early warning accuracy and emergency response efficiency.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology, specifically to an IoT-based intelligent personnel scheduling and management method and system. Background Technology

[0002] The operation and management of specific locations is extremely complex, requiring stringent dynamic allocation of human resources and risk control. Introducing advanced data processing methods to improve management efficiency and decision-making quality is a core objective.

[0003] Existing personnel management systems mostly employ scheduling methods based on static rules or fixed cycles, resulting in rigid resource allocation and an inability to cope with sudden changes in manpower needs and complex task assignments. Decision-making processes rely on managers' personal experience, lacking data-driven risk assessment and dynamic optimization capabilities, leading to low operational efficiency and hidden management risks. Therefore, existing management methods urgently need to address how to integrate multi-dimensional operational data to achieve dynamic, intelligent, and risk-aware resource scheduling, moving beyond static scheduling.

[0004] To address this, a method and system for intelligent personnel scheduling and management based on the Internet of Things is proposed. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for intelligent personnel scheduling and management based on the Internet of Things. By constructing a multi-layered intelligent perception and decision-making framework that integrates individual physiological / behavioral, social context and spatial dimensions, it aims to solve the problems of delayed risk identification, fragmented situational awareness and low scheduling response efficiency in the management of specific locations.

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

[0007] A method for intelligent personnel scheduling and management based on the Internet of Things, comprising:

[0008] The system acquires multi-source data on individuals within the monitored area in real time. This multi-source data includes each individual's real-time location, physiological data, and behavioral data. Based on a context-related risk assessment model, it dynamically calculates the current individual risk score for each individual. The calculation of the individual risk score integrates multi-source data collected by the individual, real-time multi-source data from other individuals in the surrounding area, and a dynamically updated knowledge graph of interpersonal relationships. The context-related risk assessment model includes a data input and preprocessing unit, an individual state assessment unit, a social context analysis unit, and a risk fusion decision unit. The data input and preprocessing unit receives the multi-source data and performs cleaning, normalization, and feature extraction on it. A standardized input data stream is generated; the individual status assessment unit calculates the basic risk level of the target person in an isolated state based on the standardized input data stream of the target person; the social context analysis unit queries the nodes and relationships of the target person and surrounding people in the personnel relationship knowledge graph according to the real-time location of the target person, analyzes and quantifies the potential risk impact of the current social environment through graph computation and reasoning, and generates a context risk correction factor; the risk fusion decision unit fuses the basic risk level output by the individual status assessment unit and the context risk correction factor output by the social context analysis unit, calculates it through a weighted algorithm, and outputs the individual risk score of the target person;

[0009] The physical space of the monitored area is discretized into a two-dimensional and / or three-dimensional digital grid composed of multiple grid points, and the initial risk value of all grid points is set to zero. For each person in the monitored area, a risk distribution function is constructed using their current real-time location as the center and their individual risk score as the amplitude, applying a preset kernel function to characterize the range and intensity of the individual risk of that person in space. Each grid point in the digital grid is traversed, and the function value of the risk distribution function for all persons at that grid point is calculated and accumulated to obtain the cumulative aggregate risk value at that grid point. All grid points and their corresponding aggregate risk values ​​are rendered to generate an aggregate risk topology map, and different aggregate risk values ​​are visualized through different color gradients.

[0010] The aggregated risk topology map is analyzed in real time to identify risk peak areas with risk values ​​higher than a preset threshold. The risk cause type and dynamic evolution trend of the risk peak area are determined. Based on the risk peak area, risk level, risk type and dynamic evolution trend, a multi-objective optimization scheduling algorithm is used to match the optimal executor and generate a scheduling instruction containing a suggested response level.

[0011] Preferably, the step of generating the scheduling instruction includes: extracting features from the identified risk peak region to obtain its location, risk level, risk type, and dynamic evolution trend, as input parameters for scheduling decision; for each executor, comprehensively evaluating it based on the input parameters using the multi-objective optimization scheduling algorithm to calculate sub-indicators; performing weighted calculations on multiple sub-indicators for each executor to determine the optimal executor in the current situation; and generating a scheduling instruction for the optimal executor, the scheduling instruction including intervention location, risk type summary, and suggested response level.

[0012] Preferably, the sub-indicators include intervention time, skill matching, and task interruption cost. The intervention time is determined based on the executor's current real-time location and the location of the risk peak area, using a path planning algorithm to calculate the estimated travel time of the shortest reachable path. The skill matching is determined by comparing and quantifying the preset required skills corresponding to the risk type of the risk peak area with the skill items and proficiency levels in the executor's pre-existing skill profile. The task interruption cost is a quantified cost value comprehensively evaluated based on the preset priority of the task currently being performed by the executor and whether the task can be safely interrupted.

[0013] Preferably, the personnel relationship knowledge graph is generated through social network analysis methods, specifically including: based on the real-time location data, continuously recording close contact events of n personnel within a preset distance threshold within the monitoring area, and storing the contact duration and identity information; when a risk peak area appears on the aggregated risk topology map and / or an externally input security incident report is received, performing correlation analysis on the close contact events related to the incident before and after the incident time point; based on the results of the correlation analysis, quantitatively updating the relationship attributes between the personnel involved in the close contact events; and storing the updated relationship attributes in a relational database for use by the context-related risk assessment model when calculating individual risk scores.

[0014] An IoT-based intelligent personnel dispatch and management system includes:

[0015] Data acquisition module: acquires multi-source data of each person within the monitoring area in real time, including real-time location data, physiological data, and behavioral data of each person; Risk assessment module: dynamically calculates the current individual risk score for each person based on a context-related risk assessment model; the calculation of the individual risk score integrates the multi-source data collected by the person, real-time multi-source data of other people in the surrounding area, and a dynamically updated knowledge graph of personnel relationships; the context-related risk assessment model includes a data input and preprocessing unit, an individual state assessment unit, a social context analysis unit, and a risk fusion decision unit; the data input and preprocessing unit... The system is used to receive the multi-source data, clean, normalize, and extract features from the multi-source data to generate a standardized input data stream; the individual state assessment unit calculates the basic risk level of the target person in an isolated state based on the standardized input data stream of the target person; the social context analysis unit is used to query the nodes and relationships of the target person and surrounding people in the personnel relationship knowledge graph according to the real-time location of the target person, and analyze and quantify the potential risk impact of the current social environment through graph computation and reasoning to generate a context risk correction factor; the risk fusion decision unit is used to fuse the basic risk level output by the individual state assessment unit and the data. The social context analysis unit outputs the context risk correction factor, which is then calculated using a weighted algorithm to output the individual risk score of the target person. The topology generation module discretizes the physical space of the monitored area into a two-dimensional and / or three-dimensional digital grid composed of multiple grid points, and sets the initial risk value of all grid points to zero. For each person within the monitored area, using their current real-time location as the center and their individual risk score as the amplitude, a preset kernel function is applied to construct a risk distribution function characterizing the range and intensity of the individual risk's influence in space. The module iterates through each grid point in the digital grid, calculating and accumulating the risk scores of all individuals. The risk distribution function is defined at the grid point, yielding the cumulative aggregated risk value at that grid point. All grid points and their corresponding aggregated risk values ​​are rendered to generate an aggregated risk topology map, with different aggregated risk values ​​visualized using different color gradients. The scheduling decision module analyzes the aggregated risk topology map in real time, identifies risk peak areas where the risk value exceeds a preset threshold, determines the risk cause type and dynamic evolution trend of the risk peak areas, and, based on the risk peak areas, risk levels, risk types, and dynamic evolution trends, uses a multi-objective optimization scheduling algorithm to match the optimal executor and generate scheduling instructions containing suggested response levels.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0017] 1. This invention constructs a context-related risk assessment model, which not only analyzes individual states but also innovatively builds and utilizes a knowledge graph of interpersonal relationships to transform discrete, unstructured historical interaction data into a structured relationship network. This allows risk assessment to move beyond simple historical event queries, enabling deep insights into potential conflicts arising from complex interpersonal networks through graph reasoning and analysis. It achieves a shift from superficial monitoring to deep knowledge-driven risk prediction, significantly improving the reliability and foresight of early warnings. This method can identify hidden risks that are difficult to detect using traditional methods earlier and more accurately, significantly enhancing the reliability and foresight of early warnings.

[0018] 2. The aggregated risk topology map proposed in this invention integrates and elevates massive, discrete personnel risk data into a continuous and intuitive global risk distribution map. Managers no longer need to monitor individual alerts one by one; instead, they can observe the risk like weather forecasts. Figure 1 In this way, the distribution, form, and evolution of the risk "high-pressure zones" and "low-pressure zones" in the entire monitoring area can be clearly understood at a glance. This greatly simplifies the information processing burden and realizes a fundamental transformation from fragmented point-based monitoring to integrated, global situational awareness.

[0019] 3. This invention is based on a comprehensive assessment of the causes, levels, and dynamic evolution trends of risk peaks in the risk topology map. Employing a multi-objective optimization scheduling algorithm, it can automatically match the optimal intervention executor and provide suggested response levels based on multiple dimensions such as response time, skill matching degree, and task interruption cost. This fundamentally changes the traditional scheduling model that relies on human experience and suffers from delayed responses, ensuring that every intervention is a precise response with optimal resources and appropriate strategies, achieving a dual improvement in the efficiency of resource utilization and the effectiveness of handling emergencies. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a method for intelligent personnel scheduling and management based on the Internet of Things, as proposed in an embodiment of the present invention.

[0021] Figure 2 This is a flowchart of the context-related risk assessment model proposed in an embodiment of the present invention;

[0022] Figure 3 This is a system architecture diagram of an IoT-based intelligent personnel dispatch and management system proposed in an embodiment of the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Example 1

[0025] Please see Figures 1 to 2 This invention provides a method for intelligent personnel scheduling and management based on the Internet of Things, the technical solution of which is as follows:

[0026] A method for intelligent personnel scheduling and management based on the Internet of Things, such as Figure 1 As shown, it includes:

[0027] The system acquires multi-source data on individuals within the monitored area in real time. This multi-source data includes each individual's real-time location, physiological data, and behavioral data. Based on a context-related risk assessment model, it dynamically calculates the current individual risk score for each individual. The calculation of the individual risk score integrates multi-source data collected by the individual, real-time multi-source data from other individuals in the surrounding area, and a dynamically updated knowledge graph of interpersonal relationships. The context-related risk assessment model includes a data input and preprocessing unit, an individual state assessment unit, a social context analysis unit, and a risk fusion decision unit. The data input and preprocessing unit receives the multi-source data and performs cleaning, normalization, and feature extraction on it. A standardized input data stream is generated; the individual status assessment unit calculates the basic risk level of the target person in an isolated state based on the standardized input data stream of the target person; the social context analysis unit queries the nodes and relationships of the target person and surrounding people in the personnel relationship knowledge graph according to the real-time location of the target person, analyzes and quantifies the potential risk impact of the current social environment through graph computation and reasoning, and generates a context risk correction factor; the risk fusion decision unit fuses the basic risk level output by the individual status assessment unit and the context risk correction factor output by the social context analysis unit, calculates it through a weighted algorithm, and outputs the individual risk score of the target person;

[0028] The physical space of the monitored area is discretized into a two-dimensional and / or three-dimensional digital grid composed of multiple grid points, and the initial risk value of all grid points is set to zero. For each person in the monitored area, a risk distribution function is constructed using their current real-time location as the center and their individual risk score as the amplitude, applying a preset kernel function to characterize the range and intensity of the individual risk of that person in space. Each grid point in the digital grid is traversed, and the function value of the risk distribution function for all persons at that grid point is calculated and accumulated to obtain the cumulative aggregate risk value at that grid point. All grid points and their corresponding aggregate risk values ​​are rendered to generate an aggregate risk topology map, and different aggregate risk values ​​are visualized through different color gradients.

[0029] The aggregated risk topology map is analyzed in real time to identify risk peak areas with risk values ​​higher than a preset threshold. The risk cause type and dynamic evolution trend of the risk peak area are determined. Based on the risk peak area, risk level, risk type and dynamic evolution trend, a multi-objective optimization scheduling algorithm is used to match the optimal executor and generate a scheduling instruction containing a suggested response level.

[0030] Furthermore, the physiological data is heart rate data and skin conductance data collected by a wearable wristband; the behavioral data is data on abnormal running, falling, and arm-waving attack actions of personnel obtained by analyzing the camera at the top of the monitoring area using a posture recognition algorithm.

[0031] Furthermore, such as Figure 2 As shown, the context-related risk assessment model includes a data input and preprocessing unit, an individual state assessment unit, a social context analysis unit, and a risk fusion decision unit. The data input and preprocessing unit receives the multi-source data and cleans, normalizes, and extracts features from it to generate a standardized input data stream. The individual state assessment unit calculates the target person's basic risk level in an isolated state based on the standardized input data stream. The social context analysis unit queries the nodes and relationships of the target person and surrounding individuals in the personnel relationship knowledge graph based on the target person's real-time location. Through graph computation and reasoning, it analyzes and quantifies the potential risk impact of the current social environment and generates a context risk correction factor. The risk fusion decision unit fuses the basic risk level output by the individual state assessment unit and the context risk correction factor output by the social context analysis unit, calculates the individual risk score of the target person using a weighted algorithm, and outputs the individual risk score of the target person.

[0032] Specifically, the method for calculating the basic risk level first identifies key indicators strongly correlated with risk status from an individual's physiological and behavioral data. Then, one or more risk thresholds are set for each key indicator. When the real-time value of an indicator exceeds a preset threshold, a predefined risk score is triggered. The basic risk level is ultimately obtained by weighted summation of the risk scores corresponding to all triggered key indicators, thereby integrating discrete physiological and behavioral signals into a quantifiable benchmark score characterizing the degree of danger in an individual's isolated state.

[0033] Specifically, when any indicator is detected and triggered, its corresponding risk score is accumulated to obtain the basic risk level. For example, if a person's heart rate is 140 bpm (+10 points) and they make an arm-swinging attack motion (+40 points), their basic risk level is 50 points. An example of calculating the basic risk level can be found in Table 1 below.

[0034] Table 1. Example of Basic Risk Level Reference

[0035] Key Indicators Risk threshold Risk Score resting heart rate >120 bpm or <50 bpm 10 Behavior "Arm-swinging attack" posture detected 40 Behavior "Falling down" was detected for more than 10 seconds. 30 Location Entering unauthorized areas 20

[0036] The core of the quantification method for the situational risk correction factor lies in constructing and applying a dynamic personnel relationship knowledge graph. This graph defines each person within the monitored area as an entity node with attributes such as identity, skills, and real-time risk level. The edges between nodes are used to represent their interaction history; for example, the system continuously records "close contact" relationships based on real-time location data, and "conflict" relationships determined through risk event analysis.

[0037] The aforementioned knowledge graph of interpersonal relationships possesses dynamic learning and updating capabilities. When the system identifies new interactions or risk events, it updates the edge weights of relationships between involved interpersonal nodes in real time based on the nature of the event. During risk assessment, the social context analysis unit queries the target interperson's neighboring nodes and their relationship weights in the graph, accurately quantifying the potential risks posed by the surrounding social environment through graph computation and reasoning. This method transforms ambiguous "historical relationships" into structured, quantifiable social contexts, ultimately generating a contextual risk correction factor that accurately reflects the current situation.

[0038] The weighted algorithm described is a mathematical model that integrates a basic risk level and a situational risk correction factor. Specifically, the algorithm is: Individual Risk Score = w_1 × Basic Risk Level + w_2 × Situational Risk Correction Factor, where w_1 and w_2 are preset weights representing the degree of emphasis placed on individual and environmental factors. By employing this type of algorithm, the system can generate a comprehensive and dynamic individual risk assessment result.

[0039] The context-related risk assessment model, by integrating individual status, social context, and historical interaction relationships, elevates risk assessment from isolated, superficial monitoring to contextualized, in-depth insights. This model overcomes the limitations and lag of traditional methods that focus solely on individuals, enabling earlier and more accurate identification and quantification of potential conflicts arising from interpersonal relationships. It significantly improves the reliability and foresight of risk warnings, providing a solid basis for proactive intervention.

[0040] Further, the step of generating the aggregated risk topology map includes: discretizing the physical space of the monitoring area into a two-dimensional and / or three-dimensional digital grid composed of multiple grid points, and setting the initial risk value of all grid points to zero; for each person in the monitoring area, using their current real-time location as the center and their individual risk score as the amplitude, applying a preset kernel function to construct a risk distribution function that characterizes the range and intensity of the individual risk of that person in space; traversing each grid point in the digital grid, calculating and accumulating the function value of the risk distribution function of all persons at that grid point to obtain the accumulated aggregated risk value at that grid point; rendering all grid points and their corresponding aggregated risk values ​​to generate the aggregated risk topology map, with different aggregated risk values ​​visualized through different color gradients.

[0041] The preset kernel function is a mathematical tool used to expand the discrete risk points of each individual into a continuous distribution field with spatial influence. In this invention, a Gaussian kernel function is preferably used. This function can smoothly attenuate the influence of individual risks from near to far, which is consistent with the intuitive understanding of the diffusion of risks in physical space.

[0042] In the kernel function, the bandwidth parameter directly determines the spatial range and intensity of an individual's risk. This embodiment preferably uses a fixed bandwidth. The risk distribution for all personnel adopts a uniform, pre-set fixed value based on the physical environment of the monitored area (such as the typical size of the monitored area, corridor width, etc.). Preferably, in a typical 3-meter-wide corridor or a 4x5-meter monitored area, this fixed bandwidth value is set to 1.5 meters. This value ensures that individual risk effectively covers their immediate vicinity without creating an unreasonable superposition effect with individuals who are too far away.

[0043] The specific calculation steps for the aggregated risk value are as follows: First, for the first person within the monitoring area, calculate their risk impact on this specific location. This impact value calculation integrates three factors: first, the person's individual risk score; second, the straight-line physical distance between the person and the location; and third, the risk impact range set for this person, i.e., bandwidth. During calculation, the physical distance is first divided by the bandwidth to obtain a relative distance value, which is then converted into an impact coefficient using a preset kernel function. Finally, this impact coefficient is multiplied by the person's individual risk score to obtain the risk value contributed by that person to the specific location. The same calculation is repeated for the second, third, and so on, up to the last person within the monitoring area, to obtain their respective risk values ​​contributed to the same location. The calculated risk values ​​contributed by all individuals are summed, and the total is the final aggregated risk value for that specific location.

[0044] This generation method innovatively integrates massive, discrete individual risk data into a continuous, visual global risk topology map through spatial gridding and kernel function application. This overcomes the problems of information fragmentation and cognitive overload caused by traditional point-based monitoring, enabling managers to clearly understand the overall distribution, intensity, and spatial form of risks, providing intuitive and scientific decision support for forward-looking spatial control and resource allocation.

[0045] Furthermore, the method for determining the type of risk causation is based on a multi-dimensional feature matching engine. This engine analyzes the data feature combinations of all individuals within the risk peak region and compares them with preset risk event templates to identify the most likely causation type. These templates are constructed based on historical data and expert experience. For example:

[0046] When the system detects that there are multiple individuals in the area, and their behavioral data shows high-amplitude, vigorous exercise, and their physiological data shows a rapid and synchronized increase in heart rate, while historical interaction data shows that there are conflict records among them, the system will determine the cause of the event as "physical conflict".

[0047] If only one individual's behavioral data within the area shows "falling down" or "remaining still for a long time," and their physiological data shows extreme abnormalities, while the data of other people in the vicinity are basically normal, then the cause type is determined to be "individual health problem."

[0048] If multiple individuals are detected to be gathering in an unauthorized area for an extended period of time, but their physiological and behavioral data do not show significant abnormalities in the initial stage, the cause is determined to be "illegal gathering," which is a potential risk that requires early intervention.

[0049] The method for determining the dynamic evolution trend is achieved by analyzing the changing characteristics of the risk peak over time. The system continuously tracks key indicators of the risk peak, such as the peak height, the area covered, and the number of people involved, and calculates their rate of change per unit time.

[0050] If the height or area of ​​a peak shows a continuous and rapid increase in a short period of time (such as within the past 30 seconds), and the rate of change exceeds the preset "rapid growth" threshold, the trend is judged as an "upgrading trend".

[0051] Conversely, if both the height and coverage area of ​​the peak show a clear and continuous decline, the trend is judged as a "mitigation trend".

[0052] If the peak remains at a high level, but its value fluctuates within a certain range and the rate of change is below the threshold of "significant change", the trend is judged as "continuous state".

[0053] If the risk value jumps from a safe level to a high-risk level in a very short period of time, showing a steep upward curve, the trend is judged as "sudden formation," which usually requires the highest priority response.

[0054] Through the above method, this invention can analyze an abstract, high-dimensional risk data group into "cause types" (such as fighting, disease onset) with clear business meaning and clear "evolutionary trends" (such as escalating, mitigating), thereby providing accurate and executable decision inputs for subsequent intelligent scheduling.

[0055] Furthermore, the step of generating the scheduling instruction includes: extracting features from the identified risk peak region to obtain its location, risk level, risk type, and dynamic evolution trend, which serve as input parameters for scheduling decisions; for each executor, comprehensively evaluating it based on the input parameters using the multi-objective optimization scheduling algorithm to calculate sub-indicators; performing weighted calculations on multiple sub-indicators for each executor to determine the optimal executor in the current situation; and generating a scheduling instruction for the optimal executor, the scheduling instruction including intervention location, risk type summary, and suggested response level.

[0056] The sub-indicators include intervention time indicators, skill matching degree indicators, and task interruption cost indicators;

[0057] The weighted calculation method is as follows: the sub-indicators are normalized, unifying their dimensions to the range of 0 to 1; after normalization, the system calculates a comprehensive score for each candidate executor. The weights are not fixed, but are dynamically adjusted according to the cause type and dynamic evolution trend of the current risk event;

[0058] For events identified as physical conflicts with an escalating trend, the system assigns the highest weight to the intervention time metric to ensure the fastest possible intervention.

[0059] For individual health issues, the system will significantly increase the weight of the skills matching index and prioritize dispatching personnel with first aid skills.

[0060] If the risk level is not high, but all alternative executors are performing high-priority tasks, the system will appropriately increase the weight of the task interruption cost indicator to avoid excessive impact on existing work.

[0061] Ultimately, the system will select the executor with the highest overall score as the optimal dispatch target for the current situation.

[0062] This dispatch instruction generation process elevates dispatch decision-making from simple rule-driven to intelligent tactical decision-making based on multi-objective optimization. By comprehensively optimizing dynamic risk characteristics and multi-dimensional indicators of executors, the system transcends the limitations of "dispatching to the nearest location," generating optimal solutions that balance response speed, skill matching, and task cost. This ensures the precision of intervention and the rationality of resource allocation, significantly improving the professionalism of emergency response and overall operational efficiency.

[0063] Furthermore, the sub-indicators include intervention time, skill matching, and task interruption cost. The intervention time is determined based on the executor's current real-time location and the location of the risk peak area, using a path planning algorithm to calculate the estimated travel time of the shortest reachable path. The skill matching is determined by comparing and quantifying the preset required skills corresponding to the risk type of the risk peak area with the skill items and proficiency levels in the executor's pre-existing skill profile. The task interruption cost is a quantified cost value comprehensively evaluated based on the preset priority of the task currently being performed by the executor and whether the task can be safely interrupted.

[0064] By precisely quantifying decision-making dimensions such as time, skills, and task costs, the optimal choice is transformed from a vague management concept into a calculable, objective technical indicator. This makes optimization decisions no longer abstract trade-offs, but a precise simulation of real-world opportunity costs. Therefore, the system can make highly context-aware optimal decisions, ensuring that scheduling schemes achieve a scientific balance across multiple dimensions, significantly improving the rationality and precision of resource allocation.

[0065] Furthermore, the personnel relationship knowledge graph is generated through social network analysis methods, specifically including: based on the real-time location data, continuously recording close contact events of n personnel within a preset distance threshold within the monitoring area, and storing the contact duration and identity information; when a risk peak area appears on the aggregated risk topology map and / or an externally input security incident report is received, performing correlation analysis on the close contact events related to the incident before and after the incident time point; based on the results of the correlation analysis, quantitatively updating the relationship attributes between the personnel involved in the close contact events; and storing the updated relationship attributes in a relational database for use by the context-related risk assessment model when calculating individual risk scores.

[0066] This social network analysis method endows the system with the ability to dynamically learn and remember interpersonal relationships. It correlates objective location data with risk events, making relationship assessment no longer reliant on subjective judgment or static archives, but based on continuously evolving data evidence. This allows the system to automatically identify potential "conflict pairs" and imbues the risk model with "social memory," thereby significantly improving the depth and predictive accuracy of situational risk assessment and making decision-making more insightful.

[0067] This invention provides an intelligent method to comprehensively upgrade the management of monitored areas from post-event handling and passive response to pre-event early warning and proactive intervention. By constructing a risk assessment model that integrates individual status, surrounding social context, and historical relationships, it achieves deep insight and accurate prediction of potential conflict risks, overcoming the one-sidedness and lag of traditional methods that rely solely on individual appearances. The unique aggregated risk topology map elevates discrete and independent risk information into a global and visible overall situation, enabling managers to intuitively grasp the distribution, intensity, and evolution of risks, achieving a leap from point-based monitoring to area-based perception. Multi-objective optimized scheduling based on the dynamic characteristics of risks ensures that every intervention decision achieves an optimal balance between response time, personnel skills, and task costs, realizing precise allocation of scheduling resources and maximizing intervention effectiveness. This invention systematically solves the core pain points of traditional management models: lagging risk identification, fragmented situational awareness, and low scheduling efficiency.

[0068] Example 2

[0069] This embodiment provides a detailed description of an IoT-based intelligent personnel dispatch and management system based on a specific application scenario. For example... Figure 3 As shown, the system includes a data acquisition module, a risk assessment module, a topology map generation module, and a scheduling decision module.

[0070] An IoT-based intelligent personnel dispatch and management system includes:

[0071] Data acquisition module: acquires multi-source data of each person within the monitoring area in real time, including real-time location data, physiological data, and behavioral data of each person; Risk assessment module: dynamically calculates the current individual risk score for each person based on a context-related risk assessment model; the calculation of the individual risk score integrates the multi-source data collected by the person, real-time multi-source data of other people in the surrounding area, and a dynamically updated knowledge graph of personnel relationships; the context-related risk assessment model includes a data input and preprocessing unit, an individual state assessment unit, a social context analysis unit, and a risk fusion decision unit; the data input and preprocessing unit... The system is used to receive the multi-source data, clean, normalize, and extract features from the multi-source data to generate a standardized input data stream; the individual state assessment unit calculates the basic risk level of the target person in an isolated state based on the standardized input data stream of the target person; the social context analysis unit is used to query the nodes and relationships of the target person and surrounding people in the personnel relationship knowledge graph according to the real-time location of the target person, and analyze and quantify the potential risk impact of the current social environment through graph computation and reasoning to generate a context risk correction factor; the risk fusion decision unit is used to fuse the basic risk level output by the individual state assessment unit and the data. The social context analysis unit outputs the context risk correction factor, which is then calculated using a weighted algorithm to output the individual risk score of the target person. The topology generation module discretizes the physical space of the monitored area into a two-dimensional and / or three-dimensional digital grid composed of multiple grid points, and sets the initial risk value of all grid points to zero. For each person within the monitored area, using their current real-time location as the center and their individual risk score as the amplitude, a preset kernel function is applied to construct a risk distribution function characterizing the range and intensity of the individual risk's influence in space. The module iterates through each grid point in the digital grid, calculating and accumulating the risk scores of all individuals. The risk distribution function is defined at the grid point, yielding the cumulative aggregated risk value at that grid point. All grid points and their corresponding aggregated risk values ​​are rendered to generate an aggregated risk topology map, with different aggregated risk values ​​visualized using different color gradients. The scheduling decision module analyzes the aggregated risk topology map in real time, identifies risk peak areas where the risk value exceeds a preset threshold, determines the risk cause type and dynamic evolution trend of the risk peak areas, and, based on the risk peak areas, risk levels, risk types, and dynamic evolution trends, uses a multi-objective optimization scheduling algorithm to match the optimal executor and generate scheduling instructions containing suggested response levels.

[0072] Scenario: Time: Wednesday, 15:30, public activity time in the monitored area. Location: Multifunctional activity room and adjacent corridor in monitored area A. Personnel: Monitored Person A: Has a history of violent conflict and is currently emotionally unstable. Monitored Person B: Has a past conflict with Person A and is marked as a "high-risk conflict pair" by the system. Other Monitored Personnel: 15 people, engaged in normal activities in the activity room. Security Personnel A: Patrolling the corridor in area A, 20 meters from the activity room, possesses basic first aid skills. Security Personnel B: On duty in the office in area B, approximately 100 meters from the activity room, possesses advanced conflict mediator qualifications. Initial System State: The aggregated risk topology map in monitored area A presents "green," representing low risk.

[0073] System operation process:

[0074] Step 1: Data Collection and Risk Assessment

[0075] Monitored individuals A and B enter the multi-functional activity room. The data acquisition module obtains their location data in real time through their positioning wristbands. The risk assessment module activates the context-related risk assessment model. The social context analysis unit identifies that A and B have entered a preset 3-meter social distance threshold. It immediately calls the personnel relationship knowledge graph to confirm that the two have a "high conflict risk" record. Subsequently, the system generates a higher "context risk correction factor" for the two. At this time, their physiological and behavioral data are still within the normal range, and the "basic risk level" calculated by the individual status assessment unit is still at a low level. After the risk fusion decision unit calculates using a weighted algorithm (e.g., individual risk score = 0.6 × basic risk level + 0.4 × context risk correction factor), the individual risk scores of A and B slightly increase.

[0076] Step 2: Generation and Evolution of the Aggregated Risk Topology Graph

[0077] The topology generation module re-renders the aggregated risk topology map based on the increasing individual risk scores of A and B using a Gaussian kernel function. In the northeast corner of the activity room, a small "yellow" area representing "mild risk" appears on the topology map.

[0078] The data acquisition module detected that Person A's heart rate rapidly increased from 85 bpm to 130 bpm, exceeding the resting heart rate risk threshold of 120 bpm. At the same time, the top camera, through a posture recognition algorithm, captured Person A making a dangerous "arm-swinging attack" motion.

[0079] The individual status assessment unit within the risk assessment module raises Member A's "basic risk level" to 50 points based on preset rules (excessive heart rate +10 points, arm swing attack +40 points).

[0080] The topology map generation module responds in real time, and the risk at A's location accumulates rapidly. The previously "yellow" risk area expands rapidly on the topology map and transforms into a "red" peak area representing "high risk".

[0081] Step 3: Scheduling Decision and Instruction Generation

[0082] The scheduling decision module automatically identified this newly formed "risk peak area" where the risk value exceeded a preset threshold. The module then began analyzing the risk's causes and dynamic evolution trends.

[0083] Cause determination: Combining multi-dimensional data such as "close contact between two people", "historical conflict relationship", "heart rate spike" and "aggressive actions", the system determined the risk type as "physical conflict".

[0084] Trend determination: Since the risk peak rapidly increased from "yellow" to "red" within 30 seconds, the system determined its dynamic evolution trend to be an "escalation trend".

[0085] The system initiates a multi-objective optimization scheduling algorithm to evaluate the optimal executor:

[0086] Security Personnel A: Intervention Time Index: Excellent. Route planning indicates reachable within 20 seconds. Skill Matching Index: Average. The incident is a "physical conflict," and their "basic first aid" skill matching is not high. Mission Interruption Cost Index: Low. Their current mission is "routine patrol," which can be interrupted at any time.

[0087] Security Personnel B: Intervention Time Index: Poor. Path planning indicates a 90-second arrival time; Skill Matching Index: Excellent. The "Senior Conflict Mediator" qualification is highly matched to the "Physical Conflict" incident; Mission Interruption Cost Index: Low. Currently on "Office Duty," which can be interrupted.

[0088] Optimal executor determination: Because the event was judged to be "escalating," the system dynamically adjusted the weights of the weighted algorithm, setting the weight of the "intervention time indicator" to the highest. After comprehensive calculation, security personnel A scored higher than security personnel B and was determined to be the optimal executor.

[0089] Dispatch instructions are generated: The system immediately sends an instruction to Security Personnel A's mobile terminal: "[Emergency] A physical altercation has occurred in the northeast corner of the activity room in Area A. The risk is escalating. Please proceed to handle the situation immediately!" The instruction includes a snapshot of the risk topology map of the scene and a suggested "Level 1 Response" level. Simultaneously, the system sends a secondary alarm to the command center and Security Personnel B, recommending that Security Personnel B proceed as backup.

[0090] Step 4: Post-event processing and systematic learning

[0091] Security personnel A quickly arrived at the scene and successfully separated the two individuals before the conflict fully erupted. With the individuals separated, the risk source was eliminated, and the "red" peak area on the aggregated risk topology map gradually faded, returning to "green."

[0092] The system automatically recorded all the data related to this event. Based on this event, the conflict relationship weight between member A and member B in the personnel relationship knowledge graph was further strengthened, making the system more sensitive to close contact between these two individuals in the future, thus enabling earlier warnings.

[0093] This system acquires multi-source data in real time, including personnel location, physiological and behavioral data, and dynamically calculates individual risk scores by integrating surrounding personnel and historical interaction relationships. Subsequently, it uses a pre-defined kernel function to generate a continuous, visualized global risk topology map from discrete risk points. Finally, the system automatically analyzes the causes and trends of risk peaks in the map, matches the optimal executor through a multi-objective optimization algorithm, and generates precise dispatch instructions, thereby significantly improving risk warning capabilities and emergency response efficiency.

[0094] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent personnel scheduling and management based on the Internet of Things, characterized in that, include: Real-time acquisition of multi-source data of each person within the monitoring area, including each person's real-time location data, physiological data, and behavioral data; Based on the context-related risk assessment model, a current individual risk score is dynamically calculated for each person. The calculation of the individual risk score integrates multi-source data collected by the person, real-time multi-source data of other people in the surrounding area, and a dynamically updated knowledge graph of personnel relationships. The context-related risk assessment model includes a data input and preprocessing unit, an individual state assessment unit, a social context analysis unit, and a risk fusion decision unit. The data input and preprocessing unit is used to receive the multi-source data and clean, normalize, and extract features from the multi-source data to generate a standardized input data stream. The individual status assessment unit calculates the basic risk level of the target person in an isolated state based on the standardized input data stream of the target person. The social context analysis unit is used to query the nodes and relationships of the target person and surrounding people in the personnel relationship knowledge graph based on the real-time location of the target person. Through graph calculation and reasoning, it analyzes and quantifies the potential risk impact of the current social environment and generates a context risk correction factor. The risk fusion decision unit is used to fuse the basic risk level output by the individual status assessment unit and the situational risk correction factor output by the social situation analysis unit, and calculate and output the individual risk score of the target person through a weighted algorithm. The physical space of the monitored area is discretized into a two-dimensional and / or three-dimensional digital grid composed of multiple grid points, and the initial risk value of all grid points is set to zero. For each person within the monitoring area, with their current real-time location as the center and their individual risk score as the amplitude, a preset kernel function is applied to construct a risk distribution function that characterizes the spatial range and intensity of the individual risk of that person; traversing each grid point in the digital grid, the function value of the risk distribution function for all persons at that grid point is calculated and accumulated to obtain the aggregated risk value at that grid point; Render all grid points and their corresponding aggregated risk values ​​to generate an aggregated risk topology map. Different aggregated risk values ​​are visualized using different color gradients. The aggregated risk topology map is analyzed in real time to identify risk peak areas where the risk value is higher than a preset threshold, and the risk cause type and dynamic evolution trend of the risk peak area are determined. Based on the aforementioned risk peak area, risk level, risk type, and dynamic evolution trend, a multi-objective optimization scheduling algorithm is used to match the optimal executor and generate scheduling instructions containing suggested response levels.

2. The method for intelligent personnel scheduling and management based on the Internet of Things according to claim 1, characterized in that, The steps for generating the scheduling instruction include: extracting features from the identified risk peak areas to obtain their location, risk level, risk type, and dynamic evolution trend, which serve as input parameters for scheduling decisions; for each executor, comprehensively evaluating them based on the input parameters using the multi-objective optimization scheduling algorithm to calculate sub-indicators; performing weighted calculations on multiple sub-indicators for each executor to determine the optimal executor in the current situation; and generating a scheduling instruction for the optimal executor, the scheduling instruction including the intervention location, a summary of the risk type, and a suggested response level.

3. The method for intelligent personnel scheduling and management based on the Internet of Things according to claim 2, characterized in that: The sub-indicators include intervention time, skill matching, and task interruption cost. The intervention time is determined based on the executor's current real-time location and the location of the risk peak area, using a path planning algorithm to calculate the estimated travel time of the shortest reachable path. The skill matching is determined by comparing and quantifying the preset required skills corresponding to the risk type of the risk peak area with the skill items and proficiency levels in the executor's pre-existing skill profile. The task interruption cost is a quantified cost value comprehensively evaluated based on the preset priority of the task currently being performed by the executor and whether the task can be safely interrupted.

4. The method for intelligent personnel scheduling and management based on the Internet of Things according to claim 1, characterized in that, The personnel relationship knowledge graph is generated using social network analysis methods, specifically including: continuously recording close contact events of n individuals within a preset distance threshold within the monitored area based on the real-time location data, and storing the contact duration and identity information; when a risk peak area appears on the aggregated risk topology map and / or an externally input security incident report is received, performing correlation analysis on the close contact events related to the incident before and after the incident's occurrence time; quantitatively updating the relationship attributes between the individuals involved in the close contact events based on the results of the correlation analysis; and storing the updated relationship attributes in a relational database for use by the context-related risk assessment model when calculating individual risk scores.

5. A personnel intelligent dispatch and management system based on the Internet of Things, characterized in that, include: Data acquisition module: acquires multi-source data of each person in the monitoring area in real time, including each person's real-time location data, physiological data and behavioral data; Risk assessment module: dynamically calculates the current individual risk score for each person based on the context-related risk assessment model; The calculation of the individual risk score integrates multi-source data collected by the individual, real-time multi-source data from other individuals in the surrounding area, and a dynamically updated knowledge graph of interpersonal relationships. The context-related risk assessment model includes a data input and preprocessing unit, an individual state assessment unit, a social context analysis unit, and a risk fusion decision unit. The data input and preprocessing unit receives the multi-source data and cleans, normalizes, and extracts features from the multi-source data to generate a standardized input data stream. The individual status assessment unit calculates the basic risk level of the target person in an isolated state based on the standardized input data stream of the target person. The social context analysis unit is used to query the nodes and relationships of the target person and surrounding people in the personnel relationship knowledge graph based on the real-time location of the target person. Through graph calculation and reasoning, it analyzes and quantifies the potential risk impact of the current social environment and generates a context risk correction factor. The risk fusion decision unit is used to fuse the basic risk level output by the individual status assessment unit and the situational risk correction factor output by the social situation analysis unit, and calculate and output the individual risk score of the target person through a weighted algorithm. Topology map generation module: Discretizes the physical space of the monitored area into a two-dimensional and / or three-dimensional digital grid composed of multiple grid points, and sets the initial risk value of all grid points to zero; For each person within the monitoring area, a risk distribution function is constructed using their current real-time location as the center and their individual risk score as the amplitude, applying a preset kernel function to represent the spatial range and intensity of the individual risk of that person. Each grid point in the digital grid is traversed, and the function values ​​of the risk distribution function for all people at that grid point are calculated and accumulated to obtain the cumulative aggregated risk value at that grid point. All grid points and their corresponding aggregated risk values ​​are rendered to generate an aggregated risk topology map, with different aggregated risk values ​​visualized using different color gradients. Scheduling decision module: Analyzes the aggregated risk topology map in real time, identifies risk peak areas where the risk value is higher than a preset threshold, and determines the risk cause type and dynamic evolution trend of the risk peak area; Based on the aforementioned risk peak area, risk level, risk type, and dynamic evolution trend, a multi-objective optimization scheduling algorithm is used to match the optimal executor and generate scheduling instructions containing suggested response levels.

Images