A park multi-system linkage security event intelligent dispatching method and system

Abstract

The application discloses a kind of park multi-system linkage's security event intelligent scheduling method and system, method includes: real-time acquisition multiple security subsystems deployed in park inside sensing data and alarm signal, generate security event instance with confidence label;According to the event type and confidence label of security event instance, generate multi-system linkage control strategy;Based on multi-system linkage control strategy, generate conflict-free timing scheduling instruction set;Based on timing scheduling instruction set, generate collision-free action sequence;According to action sequence and the maximum response time delay constraint of each subsystem, using static priority scheduling algorithm generates security event intelligent scheduling scheme containing the action instruction and execution time window of each subsystem. Utilize the embodiment of the application, can realize the efficient cooperation and conflict-free scheduling of multiple security subsystems, improve the linkage response speed and resource utilization efficiency of park security event.

Description

Technical Field

[0001] This invention belongs to the field of security technology, specifically a method and system for intelligent scheduling of security events involving multiple systems in a park. Background Technology

[0002] Currently, security management in industrial parks typically deploys multiple heterogeneous subsystems, including video surveillance, access control, perimeter alarms, and fire alarms. However, these subsystems usually operate independently, lacking effective information fusion and coordination mechanisms. When multiple alarms occur concurrently, existing systems often respond using preset fixed priorities or manual decision-making, making it difficult to dynamically adjust linkage strategies based on the real-time situation, spatial correlation, and equipment load. Furthermore, task execution between multiple subsystems often involves command conflicts and spatiotemporal resource competition. For example, simultaneous access control unlocking and fire alarm broadcasting may cause logical conflicts, and the lack of coordination in security equipment movement paths can easily lead to action interference. Traditional scheduling methods ignore real-time resource consumption and task load changes between subsystems, resulting in response delays, command collisions, or resource waste, making it difficult to meet the intelligent handling requirements of security events under high concurrency and strong time constraints. Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent scheduling method and system for security incidents involving multiple systems in a park, in order to overcome the shortcomings of the existing technology, achieve efficient collaboration and conflict-free scheduling of multiple security subsystems, and improve the linkage response speed and resource utilization efficiency of security incidents in the park.

[0004] One embodiment of this application provides an intelligent scheduling method for security events involving multiple systems in a park, the method comprising: The system collects sensor data and alarm signals from multiple security subsystems deployed within the park in real time, and performs fusion analysis on the sensor data and alarm signals based on spatiotemporal correlation rules to generate security event instances with confidence labels. Based on the event type and confidence level label of the security event instance, and combined with the equipment topology in the 3D map of the park, the linkage response priority of each security subsystem is dynamically calculated to generate a multi-system linkage control strategy. Based on the aforementioned multi-system linkage control strategy, and combined with the real-time resource occupancy status and task load information of each security subsystem, a distributed constraint satisfaction algorithm is used to resolve task conflicts and pre-allocate resources, generating a set of conflict-free time-series scheduling instructions. Based on the set of time-series scheduling instructions, spatiotemporal trajectory collision detection and dynamic avoidance correction are performed on the action instructions of each subsystem to generate a collision-free action sequence. Based on the action sequence and the maximum response delay constraints of each subsystem, a static priority scheduling algorithm is used to generate an intelligent security event scheduling scheme that includes action instructions and execution time windows for each subsystem.

[0005] Another embodiment of this application provides an intelligent dispatching system for security incidents involving multiple systems in a park, the system comprising: The data acquisition module is used to collect sensor data and alarm signals from multiple security subsystems deployed in the park in real time, and to perform fusion analysis on the sensor data and alarm signals based on spatiotemporal correlation rules to generate security event instances with confidence labels. The calculation module is used to dynamically calculate the linkage response priority of each security subsystem based on the event type and confidence label of the security event instance, combined with the device topology relationship in the 3D map of the park, and generate a multi-system linkage control strategy. The allocation module is used to resolve task conflicts and pre-allocate resources based on the multi-system linkage control strategy, combined with the real-time resource occupancy status and task load information of each security subsystem, and to generate a set of conflict-free time-series scheduling instructions by using a distributed constraint satisfaction algorithm. The correction module is used to perform spatiotemporal trajectory collision detection and dynamic avoidance correction on the action instructions of each subsystem based on the time-series scheduling instruction set, and generate a collision-free action sequence. The scheduling module is used to generate an intelligent security event scheduling scheme that includes action instructions and execution time windows for each subsystem based on the action sequence and the maximum response delay constraints of each subsystem using a static priority scheduling algorithm.

[0006] Another embodiment of this application provides a storage medium storing a computer program, wherein the computer program is configured to execute the method described in any of the preceding claims when running.

[0007] Another embodiment of this application provides an electronic device including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the method described in any of the preceding claims.

[0008] Compared with existing technologies, the present invention provides an intelligent scheduling method for security events involving multiple systems in a park, which can achieve efficient collaboration and conflict-free scheduling of multiple security subsystems, thereby improving the linkage response speed and resource utilization efficiency of security events in the park. Attached Figure Description

[0009] Figure 1 Hardware structure block diagram of a computer terminal for an intelligent scheduling method for security events involving multiple systems in a park, provided in an embodiment of the present invention; Figure 2 A flowchart illustrating an intelligent scheduling method for security events involving multiple systems in a park, provided as an embodiment of the present invention; Figure 3This is a schematic diagram of the structure of a smart dispatching system for security incidents involving multiple systems in a park, provided as an embodiment of the present invention. Detailed Implementation

[0010] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0011] This invention first provides an intelligent scheduling method for security events involving multiple systems in a park. This method can be applied to electronic devices, such as computer terminals, specifically ordinary computers.

[0012] The following detailed explanation uses a computer terminal as an example. Figure 1 This is a hardware structure block diagram of a computer terminal for an intelligent scheduling method for security events involving multiple systems in a park, provided as an embodiment of the present invention. (See diagram below.) Figure 1 As shown, the computer device includes a processor, memory, and network interface connected via a system bus, wherein the memory may include non-volatile storage media and internal memory.

[0013] See Figure 2 The embodiments of the present invention provide an intelligent scheduling method for security events involving multiple systems in a park, which may include the following steps: S201: Real-time collection of sensor data and alarm signals from multiple security subsystems deployed within the park, and fusion analysis of the sensor data and alarm signals based on spatiotemporal correlation rules to generate security event instances with confidence labels. Specifically, the raw sensor data streams and alarm signal streams can be collected in real time through video surveillance subsystems, access control subsystems, perimeter intrusion detection subsystems and fire alarm subsystems deployed in the park, generating a multi-source heterogeneous real-time data stream set; The core of this step is to rely on four types of core security subsystems deployed throughout the park to continuously capture environmental sensor data and abnormal alarm signals in the park environment, integrate raw data of different types and transmission formats to form a standardized multi-source real-time data set, and provide the raw data foundation for subsequent spatiotemporal alignment and fusion analysis. The specific implementation method is as follows: The park's security system comprises four subsystems: video surveillance, access control, perimeter intrusion detection, and fire alarm. These subsystems serve as comprehensive data acquisition terminals, each with distinct functions and data acquisition dimensions, providing all-encompassing coverage of the park's security status, including personnel movement, perimeter intrusion, equipment fires, and access control at entrances and exits. The video surveillance subsystem primarily collects real-time video frame sensor data from public areas, passageways, and building perimeters. The frame rate is set to 25 frames per second, ensuring the integrity of dynamic details and preventing the omission of rapid abnormal behavior. Simultaneously, it outputs alarm signals generated by abnormal movement detection, such as personnel loitering or crossing boundaries. The access control subsystem is deployed at the entrances and exits of buildings and equipment rooms within the park. It collects real-time sensor data on access control status, card swiping, and facial recognition verification. The data refresh rate is 100 milliseconds, meaning the access control status data is updated every 100 milliseconds. This allows for accurate detection of momentary abnormal access control openings, unauthorized tampering, and other such behaviors, and simultaneously generates access control anomaly alarm signals. The perimeter intrusion detection subsystem covers the entire park's perimeter wall and isolation boundaries. It collects environmental sensor data in the boundary areas through infrared and vibration detection methods. The detection response delay is controlled within 20 milliseconds, ensuring rapid data capture after boundary anomalies occur. It generates intrusion alarm signals for climbing, scaling, or damaging perimeter equipment. The fire alarm subsystem is distributed throughout the park's indoor and outdoor functional areas, collecting real-time data from temperature, smoke, and combustible gas sensors. The data collection thresholds are calibrated, and when environmental parameters exceed safety thresholds, it immediately outputs fire alarm signals indicating abnormal fire conditions or equipment malfunctions.

[0014] The four types of security subsystems independently complete data acquisition, and the output raw data exhibits typical heterogeneous characteristics, specifically in terms of data format, data dimension, acquisition frequency, and data volume. Video surveillance data is visualized frame image data, belonging to high-dimensional, large-capacity data streams; access control data is state Boolean-type structured data; perimeter detection data is environmental analog quantity data; and fire protection data is threshold-triggered pulse data. The raw sensor data streams acquired by each subsystem are continuous real-time monitoring data, without triggering conditions, and are output uninterrupted 24 / 7. Alarm signal streams are event-triggered data, actively outputting only when abnormal security events are detected. After all raw data is collected, the system aggregates the data through a unified data receiving bus. The bus supports temporary caching of various types of heterogeneous data, with a caching duration set to 5 seconds. This parameter is used to retain short-term historical data and avoid instantaneous data loss. Finally, the sensor data streams and alarm signal streams output by all subsystems are integrated and unified to form a complete set of multi-source heterogeneous real-time data streams. This set contains all the raw dynamic data of the park's overall security monitoring, covering four major security dimensions: space, personnel, equipment, and environment, providing complete input for subsequent data preprocessing.

[0015] The multi-source heterogeneous real-time data stream set is processed by timestamp alignment and spatial coordinate normalization, and the alarm signals of each subsystem are uniformly mapped to the grid coordinates of the park's 3D map to generate a spatiotemporally aligned event trigger record set. The core of this step is to address the issue of inconsistent spatiotemporal references for multi-source data. Through time dimension calibration and spatial dimension standardization, it eliminates time and spatial coordinate deviations in data acquisition from different subsystems, anchoring all abnormal alarm events in a unified three-dimensional spatial coordinate system of the park. This forms standardized event record data with unified spatiotemporal dimensions, laying the foundation for subsequent correlation mining. The specific implementation method is as follows: Timestamp alignment is a core operation for achieving time dimension unification of multi-source data. Due to inherent slight deviations in the acquisition frequency and device clocks of the four types of security subsystems, the original data streams exhibit time misalignment at the millisecond to second level, requiring calibration through a unified time reference. The system uses the global unified clock of the park's security platform as the reference clock, with a clock accuracy of 1 millisecond. This parameter represents the smallest timing unit of the global clock, meeting the requirements for high-precision time alignment. For each sensor data and alarm signal in the data stream set, the system automatically extracts the device's native timestamp. Through a time difference compensation algorithm, the deviation value between the native timestamp and the global reference clock is calculated. A positive deviation value indicates that the device's acquisition time lags behind the reference time, while a negative value indicates that the device's acquisition time is ahead of the reference time. Based on the deviation value, the timestamps of all data are corrected and uniformly calibrated to the global reference time. For continuously outputting sensor data streams, batch alignment is completed with a minimum time slice of 1 second. For instantaneously triggered alarm signal streams, single-point precise time calibration is performed to ensure that the time dimension of all data is completely unified, eliminating time misalignment issues.

[0016] Spatial coordinate normalization is the core operation for unifying the spatial dimensions of multi-source data. Each security subsystem comes with its own independent local spatial coordinate system, with different coordinate origins and axes, making direct spatial correlation analysis impossible. Therefore, it needs to be uniformly mapped to the park's 3D map coordinate system. The park's 3D map adopts a globally unified spatial grid division rule, dividing the entire park space into 5m x 5m x 3m three-dimensional grid units. The length, width, and height parameters of the grid unit correspond to the horizontal, vertical, and height dimensions of the park's plane, respectively, adapting to the spatial scale of building heights and road widths, and accurately dividing every physical space in the park. During spatial normalization, the installation calibration coordinates of each subsystem device are first read. The origin and axis parameters of the device's local coordinate system are matched with the global coordinate system of the park's 3D map to generate coordinate transformation coefficients. The transformation coefficients include four core parameters: horizontal offset, vertical offset, height offset, and angle correction, used to correct the deviation between local and global coordinates. By using coordinate transformation formulas, the monitoring location and anomaly location corresponding to each piece of sensor data and alarm signal are all converted into grid coordinate codes of the park's 3D map. Each set of grid coordinate codes uniquely corresponds to a fixed physical space in the park, achieving standardization and unification of the spatial location of all data.

[0017] After completing timestamp alignment and spatial coordinate normalization, the system binds each calibrated alarm signal to its corresponding standardized timestamp, 3D grid coordinates, subsystem type, and abnormal data parameters. This completes the structured reorganization of the data, eliminating empty data, abnormal garbled data, and duplicate redundant data generated during data acquisition, while retaining valid event trigger data. The result is a spatiotemporally aligned event trigger record set. All data in this record set has a unified time and spatial reference, and each record independently corresponds to a single security anomaly trigger behavior. The data dimensions are well-organized, the information is complete, and it can be directly used for subsequent event correlation pattern mining.

[0018] Based on the spatiotemporal association rule base, the sliding time window method is used to mine association patterns in the event trigger record set, calculate the spatial proximity and temporal continuity between alarm signals, and generate multi-source alarm association clusters. The core of this step is to rely on preset industry-adaptive spatiotemporal correlation rules, filter effective event data through a dynamic sliding time window, quantify the correlation degree of different alarm signals from both temporal and spatial dimensions, aggregate multi-source alarm signals with the same origin and correlation into correlation clusters, and realize the integration and fusion of scattered single-point alarm events. The specific implementation method is as follows: The spatiotemporal correlation rule base is a set of rules pre-stored in the security dispatch system database. Built based on the characteristics of security events occurring within the park, the rule base covers temporal correlation thresholds, spatial correlation thresholds, and cross-system linkage logic for alarm signals from different security subsystems. All rules have been trained and calibrated using massive amounts of historical security event samples from the park, adapting to various security scenarios such as personnel intrusion, fire hazards, equipment malfunctions, and unauthorized access. The rule base clearly defines the effective correlation range for different types of alarm signals, such as the maximum correlation time and maximum correlation spatial distance between perimeter intrusion alarms and video anomaly alarms, and the spatiotemporal correlation standards for fire smoke alarms and temperature anomaly alarms, providing standardized judgment criteria for data correlation mining.

[0019] The sliding time window method is the core algorithm for implementing time-series event filtering and correlation analysis. This algorithm dynamically traverses the time-series data of the event trigger record set by setting a fixed-size time window and a sliding step size, filtering out security anomalies occurring within the same time period. In this system, the sliding time window duration is set to 10 seconds, meaning that a single data filtering operation covers all event records within 10 seconds, adapting to the instantaneous linkage characteristics of park security events and avoiding invalid correlations across time periods. The sliding step size is set to 2 seconds, meaning that the time window slides forward by 2 seconds each time, achieving seamless traversal of time-series data without data omissions or redundant calculations. Each time the time window completes a slide, it captures all spatiotemporally aligned event trigger records within a time period, serving as the analysis sample for a single correlation mining operation.

[0020] For each time window segment of the event samples, the system calculates two core quantitative indicators: spatial proximity and temporal continuity between alarm signals. Spatial proximity characterizes the degree of spatial overlap between the abnormal locations corresponding to two or more alarm signals. It is calculated based on the difference in three-dimensional grid coordinates between different alarm signals. The smaller the coordinate difference, the higher the spatial proximity value, indicating that the event locations are closer and the spatial correlation is stronger. The effective threshold for spatial proximity is set at 8 meters; any two alarm signals with a spatial distance less than this value are considered to have spatial correlation characteristics. Temporal continuity characterizes the sequential connection of multiple alarm signals. It is calculated based on the difference in standardized timestamps of different alarm signals. The smaller the timestamp difference, the higher the temporal continuity, indicating that the event sequence is more coherent and the probability of belonging to a chain of events is higher. The effective threshold for temporal continuity is set at 5 seconds; signal trigger time differences less than this value are considered to have temporal correlation characteristics.

[0021] Based on the judgment criteria of the spatiotemporal correlation rule base, the spatial proximity and temporal continuity of all alarm signals within the time window are comprehensively verified. Multi-source alarm signals that simultaneously meet the spatiotemporal correlation threshold and conform to the correlation logic of the rule base are aggregated, and isolated alarm signals without any spatiotemporal correlation are eliminated. Finally, the scattered single-point alarm data are integrated into several groups of multi-source alarm correlation clusters. All alarm signals within each correlation cluster are cross-system linkage signals triggered by the same security vulnerability event. The clusters are independent of each other and correspond to different security anomalies.

[0022] Confidence scores are assigned to each cluster in the multi-source alarm association cluster. The scores are then matched and verified against real event patterns calibrated in the historical sample database, and security event instances with confidence labels are output.

[0023] The core of this step is to quantitatively evaluate the authenticity of events in each alarm cluster. Confidence is quantified through multi-dimensional indicator scoring, and pattern verification is performed by combining historical real event samples. False alarms and misleading data are eliminated, and finally, standardized security event instances labeled with their degree of authenticity and reliability are generated. The specific implementation method is as follows: The system conducts a comprehensive confidence score for each multi-source alarm association cluster generated through screening. The confidence score uses a percentage-based quantitative system, with scores ranging from 0 to 100. Higher scores indicate a higher probability of the corresponding security event actually occurring and stronger data credibility. The scoring system includes four core dimensions: signal integrity, spatiotemporal correlation matching, device status credibility, and abnormal data offset. Each dimension is assigned a fixed weight: signal integrity accounts for 30% and is used to assess the quantity integrity of alarm signals across subsystems within the association cluster; the more signals from the associated subsystems, the higher the score. Spatiotemporal correlation matching accounts for 40% and is used to assess the spatial proximity and temporal continuity of signals within the cluster, matching the standard rules; higher matching results in a higher score. Device status credibility accounts for 15% and is used to collect the real-time operating status of the devices that triggered the alarm; devices operating normally without faults receive a higher score. Abnormal data offset accounts for 15% and is used to assess the magnitude of abnormal sensor data; data anomalies within the standard range for real events receive a higher score.

[0024] By integrating the four dimensions of scoring results through a weighted summation algorithm, an initial confidence score is obtained for each group of related clusters. In the example, a group of related clusters containing perimeter intrusion alarms and video personnel movement alarms has a signal integrity score of 90, a spatiotemporal correlation matching degree of 95, a device status credibility of 92, and an abnormal data offset of 90. After weighted calculation, the final confidence score is 92.3, which accurately quantifies the true probability of the event.

[0025] The historical sample database stores standardized patterns of massive amounts of real security events in the park. For various security events such as intrusions, fires, access control anomalies, and equipment malfunctions, the database labels corresponding multi-source alarm association patterns, spatiotemporal characteristics, and data anomaly characteristics, forming standardized real event templates. The system performs a full-domain matching verification between the association features, data features, and spatiotemporal characteristics of each association cluster and the real event patterns in the historical sample database. A feature similarity comparison algorithm is used to calculate the matching similarity between the current association cluster and the historical real event templates. The similarity value ranges from 0 to 100%. When the similarity is greater than 85%, the current association cluster is determined to be a real security event; when the similarity is less than 85%, it is determined to be a false alarm event caused by equipment interference or environmental disturbances and is filtered out.

[0026] After scoring and matching verification, the system assigns a unique confidence label to each valid security event that passes verification. The labels are divided into three levels based on the scoring range: scores of 85-100 indicate a high confidence label, representing a genuine and reliable event with no possibility of false alarms; scores of 60-84 indicate a medium confidence label, representing a high probability of the event being genuine with a very small probability of interference; and scores of 0-59 indicate a low confidence label, which is directly judged as an invalid event and not output. Finally, by integrating complete information such as event type, spatiotemporal location, associated signals, confidence labels, and matching similarity, standardized security event instances that can be directly used for subsequent scheduling and analysis are generated, completing the data fusion and event generation work of this step.

[0027] S202, Based on the event type and confidence label of the security event instance, and combined with the device topology relationship in the 3D map of the park, dynamically calculate the linkage response priority of each security subsystem and generate a multi-system linkage control strategy. Specifically, the event type field and confidence label in the security event instance can be parsed, and the set of target subsystems that need to participate in the response can be determined through the event type-subsystem mapping table to generate an initial response subsystem list; The core of this step is to decompose the core attribute parameters of standardized security event instances. Based on preset event and subsystem association rules, it accurately selects the linkage subsystems that are suitable for the current security event, and initially delineates the device scope of multi-system linkage. This is the foundational preliminary step for generating precise linkage control strategies, which is used to identify the core processing objects for subsequent topology analysis and priority calculation. The specific implementation method is as follows: Security event instances are standardized event data carriers generated after multi-source data fusion and correlation analysis. They carry complete security event attribute information. Among them, the event type field and confidence label are the two core judgment parameters of the filtering and linkage subsystem. Other redundant fields such as spatiotemporal coordinates and device codes will be automatically filtered during the parsing process to avoid interfering with the filtering results. The event type field is a standardized coding parameter used to accurately define the specific scenario category of security events in the park, covering security events in all scenarios of the park, such as illegal perimeter intrusion, fire hazards in the park, abnormal access control opening, unauthorized personnel stay, and abnormal equipment alarms. Each code uniquely corresponds to a type of security event, without overlap or ambiguity, enabling the system to accurately identify event attributes. The confidence level label is a core parameter for quantifying the probability of a security event being genuine or false. Its value ranges from 0 to 100. The value of this parameter is obtained by mining the spatiotemporal correlation of the preceding events and verifying the historical samples. The closer the value is to 100, the higher the probability that the security event detected is a real and valid event. The closer the value is to 0, the more likely the alarm is a false alarm caused by ambient light and shadow interference, sensor noise, or weather disturbance. It is the core quantitative basis for determining the necessity of event response.

[0028] The event type-subsystem mapping table is a backend logical association rule system based on the park's security operation and maintenance logic and the preset functional attributes of each subsystem. It solidifies the adaptation and correspondence between various security events and security subsystems, clearly defining the core response subsystem and auxiliary response subsystem for each type of event. This eliminates the need for manual configuration and enables automatic and accurate matching of event types with linked devices. After the system completes the extraction and parsing of the event type field encoding and confidence label values, the parsing results are input into the mapping rule system for matching and retrieval. This filters out all security subsystems suitable for handling the current event, eliminating redundant subsystems with mismatched functions or those unnecessary to participate in the response, thus forming a target subsystem set specific to this security event.

[0029] To ensure the uniqueness and accuracy of the subsystem set, the system will automatically perform deduplication processing on the matching results to eliminate the problem of repeated matching of the same subsystem. Then, the deduplicated target subsystem set is standardized and regulated, basic attributes such as the exclusive system code, core function, and campus deployment area of each subsystem are recorded, and an initial response subsystem list with clear structure and complete information is integrated and generated. The list completely covers all security subsystems that need to participate in the current event linkage response, with no omissions and no redundancy, and delimits accurate data boundaries for the subsequent extraction of device topology relationships. In the example, if the parsed event type field is the perimeter illegal intrusion code EV_003, and the confidence label value is 93, it indicates that the intrusion event has extremely high authenticity. After matching through the mapping table, the perimeter intrusion detection subsystem, high-definition video monitoring subsystem, campus entrance and exit access control subsystem, and outdoor intelligent inspection subsystem are screened out as target response subsystems, and finally an initial response subsystem list including four types of subsystems is generated.

[0030] Based on the device topology relationship in the 3D campus map, extract the connection paths and dependency levels between each device node in the target subsystem set, and generate a device topology adjacency matrix; The core of this step is to rely on the 3D geographic information model of the global digital campus to excavate the spatial deployment correlation, communication connection relationship and functional linkage subordination relationship of all device nodes in the initial response subsystem, convert the abstract device topology correlation into quantifiable digital matrix data, and provide accurate topological dimension data support for the subsequent quantitative calculation of subsystem priority. The specific implementation method is as follows: The 3D campus map is a global space digital carrier that has completed refined modeling in advance, with a modeling accuracy of 0.1 meters. It completely reproduces all campus building structures, security equipment deployment points, equipment communication links, networking architecture and other global information. The map embeds a device topology relationship database, which continuously stores three types of core topology data of all security equipment: physical connection, data communication, and functional linkage, and can realize rapid retrieval and extraction of topology information for any security equipment node. The device topology relationship includes two core technical dimensions: connection path and dependency level, and the two dimensions jointly define the linkage correlation strength between devices.

[0031] The connection path refers to the real-time data communication link and physical linkage execution link between the target subsystem device nodes. Core parameters include link smoothness, data transmission latency, and link connectivity status, directly determining the efficiency and stability of inter-device linkage command transmission and data interaction. Higher link smoothness and lower transmission latency result in stronger device linkage and coordination. The dependency layer refers to the primary and secondary hierarchical levels of each device node in the security event response process. Based on device function positioning, it is divided into a first-level core detection layer, a second-level verification and monitoring layer, and a third-level execution and handling layer. First-level devices are the source detection devices that trigger events, with the highest priority and basic weight. Second-level devices are event status verification devices, playing a supporting role in evidence collection. Third-level devices are on-site handling and execution devices, relying on data from preceding devices to complete linkage actions.

[0032] The system uses the initial response subsystem list as the sole search scope, traversing the park's 3D map topology database to accurately extract all connection path parameters and dependency hierarchy parameters of all device nodes under each subsystem within the list. It masks the topology data of all irrelevant devices outside the list, eliminating interference from invalid data. After extracting the topology information, the system digitally quantifies and models the relationships between all device nodes, generating a device topology adjacency matrix. This matrix uses independent device nodes within the target subsystem as row and column bases, with each row and column number corresponding to a specific device node. Elements within the matrix range from 0 to 10, with values ​​representing the strength of the topological association between two device nodes. A value of 0 indicates no communication or functional association between the devices, while a larger value indicates a higher degree of device interoperability and stronger collaboration. In the example, the perimeter detection device and the video surveillance device are paired in a first-level linkage, with a link transmission latency of 15ms and an association strength of 9; the perimeter detection device and the access control device are paired in a cross-level linkage, with an association strength of 7; the inspection device and the front-end detection and monitoring device are paired in a remote collaborative manner, with an association strength of 6. Finally, the association values ​​of all device nodes are integrated to generate a complete and quantified device topology adjacency matrix.

[0033] Combining confidence labels and topological adjacency matrices, the analytic hierarchy process (AHP) is used to calculate the response urgency index of each subsystem. The closer the center of the subsystem is to the event occurrence area and the higher the confidence level, the higher its priority. This generates a subsystem priority score vector. The core of this step is to integrate quantitative parameters of event authenticity with quantitative parameters of device spatial topology. A weighted fusion calculation of multi-dimensional heterogeneous parameters is performed using the analytic hierarchy process (AHP) to eliminate the dimensional differences between parameters, uniformly quantify the urgency of event responses in each subsystem, and strictly adhere to the core rule that closer spatial distances correspond to higher event confidence and higher response priority. This generates a standardized quantitative scoring vector, providing accurate data for subsystem ranking. The specific implementation method is as follows: The Analytic Hierarchy Process (AHP) is a quantitative decision-making algorithm adapted to multi-criteria and multi-dimensional comprehensive evaluation. Its core function is to decompose, weight, and fusion qualitative linkages and quantitative numerical parameters in a hierarchical manner, solving the technical problem of inconsistent comparison of multi-dimensional parameters. The algorithm's calculation accuracy can reach 0.01, accurately distinguishing the differences in response urgency among different subsystems, making it suitable for the priority quantification calculation scenario of this security subsystem. The algorithm's calculation system includes two core evaluation dimensions: event confidence and device spatial topology. These two dimensions have preset fixed weight ratios. Combined with the core logic of park security scheduling, event confidence determines the necessity of event response, with a weight ratio set at 0.6; device spatial topology and distance determine response timeliness, with a weight ratio set at 0.4. The sum of the weights for both dimensions is fixed at 1, meeting the calculation specifications of the AHP.

[0034] The input parameters for the event confidence dimension directly use the confidence label values ​​of security event instances, ranging from 0 to 100, without secondary conversion. Higher values ​​indicate stronger event authenticity and a higher urgency for subsystem response. The device spatial topology dimension includes two core sub-parameters: the topology association mean and the spatial distance normalized score. The topology association mean is taken from the average association strength of all devices in the corresponding subsystem in the device topology adjacency matrix, ranging from 0 to 10, representing the tightness of the linkage between the subsystem as a whole and the core event detection device. The spatial distance normalized score is the score obtained by normalizing the straight-line spatial distance between the deployment point of the core subsystem device and the core area where the event occurred. The original distance unit is meters, and the normalized value ranges from 0 to 100, strictly following the conversion rule that the closer the distance, the higher the normalized score, accurately reflecting the spatial location advantage of the device response.

[0035] During the algorithm calculation process, the system first performs a weighted fusion of the topological association mean and spatial distance score of a single subsystem to obtain the comprehensive spatial topology score of the subsystem. Then, it combines the global confidence weight to complete the secondary weighted calculation and finally outputs the response urgency index of each subsystem. Continuing with the previous example of a perimeter intrusion incident, the event confidence score is 93, and the base score is fixed at 93. The perimeter intrusion detection subsystem is deployed in the core area of ​​the incident, at a spatial distance of 0 meters, with a normalized score of 100, a topological correlation average of 9.1, a comprehensive topological score of 99.1, and a final urgency score of 94.1. The video surveillance subsystem is 4 meters away from the incident area, with a normalized score of 96, a topological correlation average of 8.7, a comprehensive topological score of 91.5, and a final urgency score of 92.3. The access control subsystem is 28 meters away from the incident area, with a normalized score of 77, a topological correlation average of 7.2, a comprehensive topological score of 73.1, and a final urgency score of 79.5. The intelligent patrol subsystem is 45 meters away from the incident area, with a normalized score of 63, a topological correlation average of 6.4, a comprehensive topological score of 64.2, and a final urgency score of 70.8.

[0036] After the urgency index calculation of all subsystems is completed, the system maps the unique codes of each subsystem to the corresponding urgency scores one by one, integrates them into a one-dimensional ordered array according to the subsystem matching sequence, and generates a subsystem priority score vector. Each independent numerical element in the vector uniquely corresponds to one target response subsystem, which accurately and quantitatively presents the difference in response priority of each subsystem.

[0037] According to the subsystem priority score vector, each subsystem is sorted from the highest score to the lowest score, the linkage trigger delay time is set, and a linkage control strategy including subsystem number, response sequence and delay parameters is generated.

[0038] The core of this step is to complete the ranking of subsystem response levels based on the quantified priority score vector, match the adapted linkage trigger timing parameters with the priority levels, avoid the problems of signal congestion and task conflict caused by multi-device synchronous triggering, regularize all core linkage parameters, and form a standardized multi-system linkage control strategy that can be directly scheduled and executed. The specific implementation mode is as follows: The numerical value in the subsystem priority score vector directly corresponds to the response urgency level of each security subsystem. A higher score indicates that the subsystem has higher requirements for response necessity and timeliness for the current security event, and the linkage action needs to be started preferentially; a lower score indicates a lower response priority, and the linkage task can be started later. The system completely reads all data of the priority score vector, extracts the coding information and corresponding scores of all subsystems, strictly completes the global sorting according to the descending rule of scores from large to small, accurately divides the sequence of linkage responses of each subsystem, and forms a clearly hierarchical response sequence. Continuing with the foregoing example, the final sorting result is the perimeter intrusion detection subsystem, the video monitoring subsystem, the park access control subsystem, and the intelligent inspection subsystem, which completely conforms to the closed-loop response logic of "detection-verification-control-disposal" for security events.

[0039] The linkage trigger delay time is the core timing parameter to ensure the orderly coordinated linkage of multiple systems, with the unit of millisecond and the value range from 0 to 1000ms. The parameter setting follows the priority positive adaptation rule: the higher the priority of the subsystem, the shorter the delay time, so as to ensure the instantaneous response of the core equipment; the lower the priority of the subsystem, the longer the delay time, which reserves sufficient response buffer time for high-priority equipment, and effectively avoids channel congestion and task overlap conflict caused by synchronous transmission of multi-device instructions. Wherein, the delay time of the highest priority subsystem is fixed at 0ms to realize instantaneous trigger response; the delay time of the secondary priority subsystem is set to 100ms to 300ms; the delay time of the medium priority subsystem is set to 300ms to 600ms; the delay time of the low priority subsystem is set to 600ms to 1000ms.

[0040] In the corresponding example's sorting hierarchy, the perimeter intrusion detection subsystem has the highest priority, with a linkage trigger delay time set to 0ms. It continuously outputs detection alarm data immediately after event recognition is completed. The video surveillance subsystem has the second highest priority, with a delay time set to 200ms. It initiates real-time recording, image capture, and area focusing linkage actions after the detection system stably outputs data. The park access control subsystem has the middle priority, with a delay time set to 450ms. It executes area access control locking and passage interception actions after waiting for event image verification and confirmation. The intelligent patrol subsystem has the lowest priority, with a delay time set to 800ms. It initiates autonomous patrol, on-site verification, and anomaly handling linkage tasks after the preceding equipment completes the initial prevention and control actions.

[0041] After completing the subsystem sorting and delay parameter assignment, the system standardizes and integrates all core linkage parameters, fully recording the unique number of each linkage subsystem, the final response sequence, and the corresponding linkage trigger delay parameters. The parameters correspond one-to-one and form a logical closed loop, ultimately generating a standardized multi-system linkage control strategy. This strategy provides a precise basis for subsequent task conflict resolution, resource pre-allocation, and timing scheduling instruction generation, enabling layered, orderly, and efficient collaborative linkage response of multiple security subsystems in the park.

[0042] S203, based on the multi-system linkage control strategy, combined with the real-time resource occupancy status and task load information of each security subsystem, a distributed constraint satisfaction algorithm is used to resolve task conflicts and pre-allocate resources, generating a set of conflict-free time-series scheduling instructions; Specifically, the current real-time resource usage status and task load information, including CPU utilization, communication bandwidth usage and device busy / idle status, can be obtained from the device controllers of each security subsystem, and a snapshot of the subsystem resource status can be generated. The core of this step is to collect real-time data on the hardware resources and task load of all security subsystems in the park, capture the instantaneous system resource operating status, and form a standardized resource data snapshot. This provides a real and accurate data source for subsequent task scheduling, conflict resolution, and resource allocation. The specific implementation method is as follows: Each security subsystem in the park is equipped with an independent device controller, serving as the core terminal unit for resource data acquisition. This includes controllers for the video surveillance subsystem, access control subsystem, perimeter intrusion detection subsystem, and fire alarm subsystem. All controllers support millisecond-level real-time data acquisition and uploading, with a fixed data refresh rate of 100 milliseconds. This ensures accurate capture of instantaneous resource fluctuations in the system, preventing scheduling deviations caused by data lag. The core acquisition dimensions of the device controllers include three key indicators: CPU utilization, communication bandwidth usage, and device busy / idle status. These three indicators work together to comprehensively characterize the real-time load and resource availability of the subsystems. After data acquisition, all data undergoes standardized formatting and numerical calibration to ensure data consistency across subsystems and allow for horizontal comparison.

[0043] CPU utilization is a core parameter characterizing the degree of computing resource occupancy of a subsystem. Its value ranges from 0 to 100%. A higher value indicates a higher proportion of core computing resources being occupied, with less available computing power. 0% represents an idle state with no computing load, while 100% represents a fully loaded subsystem with no new tasks to be scheduled. In actual data acquisition scenarios, the video surveillance subsystem, due to continuous video parsing and recording, has a CPU utilization of approximately 42% under normal load. The perimeter intrusion detection subsystem, due to real-time monitoring of sensor data, has a CPU utilization of approximately 35%. The access control subsystem typically has a lower CPU utilization of approximately 18%, while the fire alarm subsystem has a CPU utilization of only 8% in standby mode. The differences in computing load among various subsystems can be accurately reflected by this parameter.

[0044] Communication bandwidth utilization is a key parameter characterizing the data transmission resource usage of a subsystem, measured in Mbps. The value ranges from 0 to 100 Mbps, based on the preset bandwidth limit of the park's security system. It represents the data traffic usage for real-time data uploads, command reception, and device linkage within the subsystem; a higher value indicates less remaining schedulable bandwidth. In the example, the video surveillance subsystem transmits high-definition video streams in real time, utilizing 38 Mbps of bandwidth; the perimeter intrusion detection subsystem continuously uploads sensor location data, utilizing 15 Mbps of bandwidth; and the access control and fire protection subsystems handle relatively small amounts of regular data transmission, utilizing 6 Mbps and 4 Mbps of bandwidth respectively. The remaining bandwidth resources can be used to support the data interaction needs of newly added security linkage tasks.

[0045] The busy / idle status of a device is a qualitative parameter characterizing the current task execution status of a subsystem. It is uniformly divided into three states: idle, semi-busy, and busy. Status determination relies on CPU utilization and bandwidth usage thresholds for automated identification. Specifically, the idle state corresponds to CPU utilization below 20% and bandwidth usage below 10Mbps, indicating that the subsystem has no core security tasks being executed; the semi-busy state corresponds to CPU utilization between 20% and 60% and bandwidth usage between 10Mbps and 50Mbps, indicating that the subsystem is only executing routine background monitoring tasks and can handle low-priority new tasks; the busy state corresponds to CPU utilization above 60% and bandwidth usage above 50Mbps, indicating that the subsystem is operating under high load and cannot handle new scheduling tasks. In the example, the fire alarm subsystem and access control subsystem are in the idle state, the perimeter intrusion detection subsystem is in the semi-busy state, and the video surveillance subsystem is in the busy state.

[0046] After collecting and determining the status of the three core parameters of all subsystems, the system integrates the device number, CPU utilization, communication bandwidth usage, and device busy / idle status of all subsystems. Abnormal fluctuations are removed, missing data is supplemented, and the system freezes the global resource data for the current 100 milliseconds, generating a snapshot of the subsystem resource status. This snapshot is an instantaneous static resource dataset that fully records the real-time load of all security subsystems in the park. The data precision is uniformly retained to two decimal places, and it can be directly used as a benchmark for subsequent task scheduling and resource allocation.

[0047] The multi-system linkage control strategy is parsed into a set of atomic tasks to be scheduled. Each atomic task includes the target subsystem, action type and required resource quota, and an atomic task queue is generated. The core of this step is to finely decompose the macroscopic multi-system linkage control strategy, breaking down the overall linkage scheduling scheme into the smallest indivisible execution units. The execution subject, execution actions, and resource requirements of each execution unit are standardized and defined, and then systematically integrated to form a schedulable task queue. This provides a standardized task carrier for subsequent algorithm scheduling. The specific implementation method is as follows: The multi-system linkage control strategy is a macro-level linkage scheme generated based on security event types and subsystem priorities. It includes core elements such as a list of participating subsystems, subsystem response order, and linkage delay parameters. As a structured set of linkage scheduling rules, it cannot be directly used for fine-grained resource scheduling. Therefore, it needs to be decomposed and parsed by the system's built-in strategy parsing module. This module has standardized rule decomposition logic, which can break down the entire linkage strategy into multiple independent, single, and indivisible atomic tasks based on the response requirements of security events. Each atomic task is the smallest execution unit for park security scheduling, and there is no possibility of secondary decomposition. Each atomic task independently corresponds to a specific security linkage operation.

[0048] Each atomic task uniformly includes three core attributes: target subsystem, action type, and required resource quota. These three attributes completely define the task's executing entity, execution content, and resource consumption standards; none can be omitted. The target subsystem attribute is used to clarify the system to which the current atomic task's executing device belongs, limiting the scope of the task's executing entity. Optional entities include video surveillance subsystems, access control subsystems, perimeter intrusion detection subsystems, and fire alarm subsystems. All tasks correspond to one or more linked subsystems. In the example, for a security event involving intrusion into a park area, the target subsystems are the perimeter intrusion detection subsystem, video surveillance subsystem, and access control subsystem.

[0049] The action type attribute defines the specific execution operation of an atomic task. Different security subsystems correspond to specific standardized actions. The perimeter intrusion detection subsystem includes actions such as enhanced area detection and anomaly point focusing monitoring; the video surveillance subsystem includes actions such as real-time image retrieval, anomaly area capture, and video trajectory tracking; the access control subsystem includes actions such as access control locking, area permission blocking, and entrance / exit status monitoring; and the fire alarm subsystem includes actions such as fire investigation, equipment inspection, and alarm signal verification. In the example, the action types corresponding to the above subsystems are anomaly point focusing monitoring, anomaly area capture, and access control locking, respectively.

[0050] The required resource quota is the minimum hardware resource standard required to execute the corresponding atomic task. It includes three types of quantitative parameters: computing power resource quota, bandwidth resource quota, and device occupancy time quota. All parameters are generated based on historical security scheduling samples to ensure that the values ​​are reasonably adapted to the performance of the park's equipment. Among them, the computing power resource quota is expressed in CPU utilization ratio, representing the minimum CPU utilization rate required for task execution. The computing power quota for the abnormal point focusing monitoring task is 20%, the computing power quota for the abnormal area capture task is 15%, and the computing power quota for the access control locking task is 10%. The bandwidth resource quota is expressed in Mbps, representing the minimum bandwidth resource required for task data interaction. The bandwidth quotas for the three types of tasks are 8Mbps, 12Mbps, and 3Mbps, respectively. The device occupancy time quota is expressed in milliseconds, representing the continuous device occupancy time required for task execution. The occupancy times for the three types of tasks are 300 milliseconds, 200 milliseconds, and 150 milliseconds, respectively.

[0051] After decomposing and assigning parameters to all atomic tasks, the system sorts all atomic tasks according to the subsystem response priorities in the multi-system linkage control strategy. Atomic tasks corresponding to higher-priority subsystems are ranked higher, while the linkage timing relationships of tasks with the same priority are preserved, resulting in an ordered linear queue of atomic tasks. This queue strictly follows the response logic of the linkage strategy, with all task parameters fully verifiable and the task order clearly defined, providing standardized task input for the subsequent scheduling operations of the distributed constraint satisfaction algorithm.

[0052] Based on the subsystem resource status snapshot and atomic task queue, a distributed constraint satisfaction algorithm is used to search for feasible execution time slots for each atomic task, and conflict resolution is performed for tasks with overlapping time slots to generate a preliminary resource pre-allocation scheme. The core of this step is to rely on real-time resource snapshots and standardized task queues, and use distributed constraint satisfaction algorithms to accurately match tasks with system resources, filter legal execution periods, resolve resource and timing conflicts, and initially complete the pre-allocation of global resources, achieving preliminary adaptation between tasks and resources. The specific implementation method is as follows: The distributed constraint satisfaction algorithm is an intelligent scheduling algorithm adapted to the distributed architecture of multiple subsystems in a park. Its core adaptability lies in constraint scheduling scenarios involving multiple nodes, multiple resources, and multiple tasks. It can address the characteristics of independent operation, dispersed resources, and parallel tasks in each security subsystem by distributively solving for the optimal resource matching scheme and execution time of each atomic task, thus resolving the problems of insufficient computing power and local scheduling imbalance in centralized scheduling. The algorithm uses the resource limits of each subsystem, task resource quotas, and equipment busy / idle status as core constraints, aiming to maximize resource utilization and minimize task response latency, iteratively completing the global task matching calculation.

[0053] The execution time slot is the system's pre-defined standardized minimum scheduling time unit. Each time slot has a fixed duration of 100 milliseconds and represents the smallest time granularity for the execution of all atomic tasks and resource usage. All scheduling sequences in the park security system are divided based on time slots to ensure the uniformity and accuracy of scheduling. After the algorithm starts, it iterates through the system's pre-defined continuous time slot intervals and, combined with real-time load data from subsystem resource status snapshots, searches for feasible time slots for each task in the atomic task queue, matching idle time slots that meet the task's resource quota requirements and where the device is in a schedulable state.

[0054] The specific matching logic is as follows: For each atomic task in the queue, the time slots corresponding to the idle or semi-busy states of the target subsystem are retrieved first. The remaining CPU computing power and communication bandwidth of the subsystem within that time slot are verified to meet the resource quota required by the task. If the remaining resources completely cover the task's resource requirements, the time slot is determined to be a feasible execution time slot, and the task is initially bound to the corresponding time slot and resources. If there are currently no idle time slots, time slots that can utilize idle resources in the semi-busy state are retrieved, completing resource matching without affecting the operation of the original core tasks. In the example, the access control locking task has low resource requirements and can be matched with the first idle time slot; the video capture task needs to be matched with the idle resource time slot in the semi-busy state of the video surveillance subsystem; and the perimeter monitoring task matches the idle time slot of the perimeter subsystem.

[0055] During multi-task parallel matching, multiple atomic tasks may compete for the same time slot resources in the same subsystem, resulting in time slot overlap and conflict. Specifically, within the same time slot, the cumulative computing power and bandwidth quotas of multiple tasks exceed the remaining available resources of the subsystem, preventing simultaneous execution. To address this conflict, the algorithm initiates a conflict resolution mechanism. Based on the confidence level of security events, subsystem response priority, and task urgency, tasks are prioritized. The resource requirements of high-priority and high-urgency tasks are guaranteed first, while the time slots of conflicting low-priority tasks are postponed and scheduled to adjacent available time slots.

[0056] After completing time slot matching and conflict resolution for all tasks, the system aggregates all information corresponding to all atomic tasks, including the target subsystem, execution time slot number, pre-allocated CPU resource ratio, pre-allocated bandwidth resource value, and device occupancy time period, and integrates them to form a preliminary resource pre-allocation scheme. This scheme achieves a one-to-one match between all scheduled atomic tasks and system resources and execution time periods, eliminating basic resource preemption and timing overlap issues, and laying the foundation for subsequent global verification and optimization.

[0057] A global consistency check is performed on the initial resource pre-allocation scheme, the task order caused by resource mutual exclusion is adjusted, and finally a set of conflict-free time-series scheduling instructions is generated.

[0058] The core of this step is to move beyond local, single-task matching logic and perform global verification of resource allocation and task timing from the perspective of overall security scheduling across the entire park. This resolves the implicit conflict between reasonable local matching and mutually exclusive global resources, optimizes task execution timing, and ultimately forms a set of standardized scheduling instructions that are completely conflict-free and can be executed directly. The specific implementation method is as follows: Global consistency verification is a compliance verification mechanism for the entire domain of the preliminary resource pre-allocation plan. Unlike the local conflict resolution of a single task, the verification dimensions cover four core dimensions: total resource load of all security subsystems, cross-subsystem linkage timing, resource mutual exclusion adaptability, and task latency compliance. The purpose is to identify hidden global scheduling problems in the preliminary allocation plan and ensure that the overall scheduling logic is unified, resource allocation is balanced, and timing is orderly.

[0059] Resource mutual exclusion is an inherent characteristic of hardware and communication resources in security systems. Specifically, it means that the same physical device, the same communication channel, and the same computing power cluster cannot carry two or more exclusive security tasks at the same time. Some core devices and high-frequency communication resources have strong mutual exclusion attributes. Even if different tasks are matched with different time slots, there may still be implicit mutual exclusion problems such as overlapping resource occupation across time periods and conflicts in background resource locking. For example, the high-definition capture and trajectory tracking tasks of the video surveillance subsystem have computing power resource mutual exclusion, and the multi-point monitoring and enhanced scanning tasks of the perimeter detection subsystem have sensor resource mutual exclusion. Such implicit conflicts cannot be investigated by matching local time slots and must rely on global verification and identification.

[0060] The specific execution logic of the global consistency check is as follows: the system iterates through the resource usage range, execution sequence, and device usage status of all tasks in the initial resource pre-allocation scheme, calculates the cumulative resource usage of each subsystem over the entire time period, and verifies whether there are issues such as excessive resource usage, parallel scheduling of mutually exclusive tasks, or misalignment of cross-subsystem linkage timing. For resource-mutually exclusive conflicting tasks discovered during the check, the system automatically fine-tunes the execution order of conflicting tasks by combining the priority of security event response, task latency constraints, and the sequential relationship of linkage logic. Under the premise of not changing the pre-allocated resource quota, not exceeding the device resource limit, and not violating the maximum response latency constraint, the mutually exclusive tasks are split into different consecutive time slots to achieve staggered execution of mutually exclusive tasks.

[0061] After completing global verification and timing adjustments, all atomic tasks achieve reasonable resource allocation, smooth timing connections, and no explicit or implicit conflicts. The system standardizes and encapsulates all optimized task information, uniformly labeling each task's execution subsystem, specific actions, resource quotas, precise execution time slots, and task execution order. This information is then systematically integrated according to the complete logic of security event linkage response, ultimately generating a conflict-free set of timing scheduling instructions. This instruction set possesses full compliance, timing uniqueness, and resource adaptability, and can be directly used for subsequent device action trajectory correction and final scheduling scheme generation. It serves as the core execution basis for the multi-system linkage scheduling of the park's security system.

[0062] S204, Based on the set of time-series scheduling instructions, perform spatiotemporal trajectory collision detection and dynamic avoidance correction on the action instructions of each subsystem to generate a collision-free action sequence; Specifically, it can parse each action instruction in the time-series scheduling instruction set, extract the position trajectory points and speed parameters of mobile devices such as inspection robots or drones, and generate a set of predicted motion trajectories for mobile devices; The core of this step is to accurately parse the pre-processed conflict-free time-series scheduling instructions, select the equipment operation instructions with dynamic movement capabilities within the park's security system, extract core motion parameters, and complete standardized integration to construct a dataset of equipment motion trajectories that can be used for collision detection. This provides complete raw data support for subsequent spatiotemporal collision risk verification. The specific implementation method is as follows: The time-series scheduling instruction set is a standardized collection of instructions formed after task conflict resolution and resource pre-allocation. It internally organizes and includes all pending atomic action instructions from all park security subsystems, covering two main types: static operation instructions for fixed equipment and dynamic operation instructions for mobile devices. All instructions are labeled with standardized fields such as a unique equipment code, equipment type, task, execution time period, and resource usage parameters. The data format is uniform and there are no task conflicts. Before parsing begins, the system defaults to locking instructions from fixed equipment without motion attributes, such as access control, monitoring, fire protection, and perimeter detection. It only performs in-depth parsing and parameter extraction on action instructions from two types of mobile devices commonly used in park security scenarios: ground inspection robots and aerial security drones, precisely defining the data processing scope.

[0063] Location trajectory points are the core parameters defining the spatial movement path of mobile devices. Specifically, they refer to the 3D grid coordinates of the park corresponding to each discrete sampling time node within a preset execution time range. The entire coordinate system is completely unified with the park's 3D map, including three dimensions: x (horizontal coordinate), y (vertical coordinate), and z (height coordinate). The coordinate unit is meters, and the data precision is fixed to 0.01 meters, enabling precise location of the device at any given time. The system extracts the complete trajectory coordinates of all sampling nodes within a single device movement command at a fixed sampling interval of 0.1 seconds, forming continuous spatial path point data. In the example, the security drone with park number M01 executes a patrol command for key areas of the park. The command execution time is 20 seconds. The system extracts 200 sets of trajectory point coordinates at 0.1-second intervals. The coordinates of the first starting trajectory point are (18.60, 22.30, 7.50), the coordinates of the trajectory point at the mid-range patrol turning point are (35.20, 40.80, 10.20), and the coordinates of the trajectory point at the end point are (52.70, 65.40, 8.00). All trajectory points are connected to form the drone's preset flight path.

[0064] Speed ​​parameters are core dynamic parameters characterizing the motion state of mobile devices, including two core indicators: instantaneous speed and average speed over the entire process. The unified unit of measurement for speed is meters per second, with an accuracy retained to 0.1 meters per second. Instantaneous speed is the real-time movement rate of the device at a single sampling time point, accurately reflecting the device's dynamic motion state. Average speed over the entire process is the overall movement rate of the device completing the entire preset movement path, used to verify the reasonableness of the device's operation time. The system also incorporates device speed threshold constraints: the normal movement speed range for ground inspection robots is 0.5 to 2.0 meters per second, and for aerial security drones, it is 2.0 to 5.0 meters per second. Threshold verification is automatically performed during parameter extraction, eliminating abnormally high-speed, low-speed, and invalid parameters. In the example, the instantaneous speed of the ground inspection robot (R02 in the park) is 1.3 meters per second, and the average speed over the entire inspection path is 1.1 meters per second, parameters that fully comply with the device's normal operating standards.

[0065] After extracting parameters from all mobile device instructions within the time-series scheduling instruction set, the system uses the device's unique code as the classification basis. It then associates and binds all trajectory point coordinates, corresponding sampling time, instantaneous speed, and average speed parameters for each device, eliminating invalid data such as null values, duplicate points, and abnormal speeds. The system unifies the storage format and sorting logic of all data, organizes the path data according to the chronological order of device operations, and ultimately generates a standardized mobile device motion trajectory prediction set. This prediction set comprehensively covers all motion information of all mobile devices participating in security-linked scheduling, accurately reproducing the preset operating trajectories and motion states of the devices, providing a comprehensive, accurate, and effective data foundation for subsequent high-precision spatiotemporal collision detection.

[0066] Based on the static obstacle layer of the 3D map of the park and the motion trajectory prediction set of mobile devices, a four-dimensional spatiotemporal collision detection algorithm is used to calculate the risk of any two mobile devices arriving at the same spatial location at the same time, and generate a collision risk pair list. The core of this step is to combine the fixed physical space environment of the park with the dynamic motion data of mobile devices, and rely on a four-dimensional spatiotemporal collision detection algorithm to complete the collision risk verification of all equipment from both spatial and temporal dimensions. This accurately identifies potential trajectory collision hazards in multi-device coordinated operations, locks down conflicting equipment combinations, and provides clear target objects and risk basis for subsequent trajectory avoidance correction. The specific implementation method is as follows: The static obstacle layer of the park's 3D map is a digital foundation layer that replicates the park's real physical space. It pre-compiles 3D modeling and input of all fixed facilities within the park, comprehensively including spatial data of all immovable obstacles such as building walls, columns, fences, streetlight poles, surveillance brackets, fixed fire-fighting equipment, and landscaping. The layer uses a 3D grid coordinate system consistent with the device trajectory coordinates, achieving a spatial positioning accuracy of up to 0.01 meters. It can accurately mark the occupied space and boundary coordinates of all fixed obstacles. This layer not only detects the collision risk between mobile devices and fixed obstacles in the park but also provides a unified spatial benchmark for verifying the relative spatial positions of multiple mobile devices. This ensures that the spatial position calculation, comparison, and verification standards of all devices are completely consistent, avoiding distortion of detection results due to coordinate system deviations.

[0067] The four-dimensional spatiotemporal collision detection algorithm is the core technology of this step. Unlike traditional detection methods that only detect overlap in three-dimensional space, this algorithm adds a time dimension (t) to the x, y, and z three-dimensional spatial coordinates, constructing a four-dimensional integrated detection model. This model can simultaneously verify the spatial and temporal overlap relationships of devices, completely avoiding the misjudgments and omissions caused by traditional three-dimensional detection methods that only judge spatial overlap and ignore temporal misalignment, thus significantly improving the accuracy of collision detection. The core operating logic of the algorithm is to perform pairwise full combination matching on all mobile devices participating in the scheduling, traversing all sampling time nodes and corresponding three-dimensional spatial coordinates in each group of device trajectory prediction sets, continuously comparing the spatial position difference between two devices in the same time dimension, and combining this with a preset safety distance threshold to determine whether there is a collision risk.

[0068] The algorithm operates based on two core controllable parameters: time sampling accuracy and device safety distance threshold. Both parameters are pre-calibrated according to the security operation scenario of the park. The time sampling accuracy is fixed at 0.1 seconds, meaning the algorithm completes a full spatiotemporal position comparison of all devices every 0.1 seconds. High-precision time sampling can capture the potential for instantaneous position overlap between devices, avoiding the omission of short-term collision risks due to excessively large sampling intervals. The device safety distance threshold is a differentiated judgment parameter, set according to the type of device. The safety distance threshold between ground inspection robots is 0.5 meters, the cross-boundary safety distance threshold between inspection robots and security drones is 1.0 meter, and the aerial safety distance threshold between security drones is 1.5 meters. The differentiated threshold adapts to the size and motion characteristics of different devices, ensuring that risk assessment is consistent with the actual operation scenario.

[0069] In the example, the system completed a full pairwise matching detection of three mobile devices in the park: drone M01, drone M02, and inspection robot R02. At time t=15.3 seconds, the spatial coordinates of drone M01 were (42.10, 55.30, 9.80), and the spatial coordinates of drone M02 were (42.90, 55.90, 10.10). The calculated real-time straight-line distance between the two devices was 0.95 meters, which is less than the 1.5-meter safe distance threshold between drones. The algorithm determined that this combination of devices had a collision risk, and the distance difference was relatively large, so it was labeled as a level 2 medium-risk collision hazard. The other combinations of devices were detected throughout the entire time period and there was no problem with overlapping spatiotemporal positions, so there was no collision risk.

[0070] After the algorithm completes full-dimensional, all-time detection of all device combinations, it will uniformly summarize all risky device combination information, synchronously recording the device code, precise collision time point, predicted collision spatial location, and corresponding risk level for each group of risky devices. The risk level is divided into three levels: Level 1 (high risk), Level 2 (medium risk), and Level 3 (low risk) according to the ratio of spatial distance difference to safety threshold. After all risk data is organized, it is integrated into a collision risk pair list according to a standardized format, completely preserving all device collision hazard information without omissions or misjudgments, providing accurate data support for subsequent targeted trajectory avoidance corrections.

[0071] For each pair of conflicting devices in the collision risk pair list, a priority-weighted adjustment method is used to fine-tune the trajectory or speed of the device with lower priority, generating a set of avoidance correction parameters. The core of this step is to use the priority of equipment linkage response as the core judgment criterion. A priority-weighted adjustment method is used to differentiate and correct conflicting equipment, ensuring that the operation trajectory and execution sequence of high-priority equipment remain completely unchanged, while only making minor parameter adjustments to low-priority equipment. This completely eliminates the risk of collisions while maximizing the integrity and execution efficiency of the original scheduling strategy. The specific implementation method is as follows: The equipment priority determination criteria follow the multi-system linkage response priority rules mentioned earlier, and are comprehensively generated by combining equipment task attributes, event correlation, confidence score, and spatial distance indicators. Equipment directly involved in the core investigation and early warning handling of security incidents has a higher priority. Equipment closer to the incident area and with a higher event confidence score is ranked higher. In conflicting equipment groups, the system automatically compares the priority score values ​​of two devices. The device with the higher score value is defined as a high-priority device, and its motion parameters are locked without any adjustments. The device with the lower score value is defined as a low-priority device and serves as the sole object for parameter correction, ensuring that the execution of core security handling tasks is not interfered with.

[0072] The priority-weighted adjustment method is the core algorithm for achieving dynamic collision avoidance correction. The core principle is to match corresponding weight coefficients based on the collision risk level, quantifying the parameter adjustment range of low-priority equipment to achieve standardized and intelligent collision avoidance correction. The algorithm has built-in fixed weight matching rules: Level 1 high-risk collision hazard corresponds to a weight coefficient of 0.3, representing the largest possible adjustment range for the equipment; Level 2 medium-risk collision hazard corresponds to a weight coefficient of 0.5, representing a moderate adjustment range; and Level 3 low-risk collision hazard corresponds to a weight coefficient of 0.7, representing only minor adjustments required. The weight coefficient values ​​are fixed between 0.1 and 0.9; the larger the value, the smaller the space for equipment modification, and the closer it is to the original operating trajectory.

[0073] The algorithm includes two core correction methods: spatial trajectory fine-tuning and dynamic speed adjustment. These two methods can be used individually or in combination depending on the collision scenario to adapt to different spatiotemporal collision risks. Trajectory fine-tuning is a spatial dimension correction, referring to a small offset of the trajectory coordinates of the lower-priority device in the direction perpendicular to the relative motion of the two conflicting devices. The offset magnitude is calculated jointly by the weighting coefficient and the safety distance gap value. The maximum offset of the device does not exceed 3 meters, strictly controlling the offset range to prevent the device from deviating from the preset working area and affecting the quality of security patrol and monitoring tasks. In the example, the aforementioned level-two medium-risk collision risk of drones M01 and M02 corresponds to a weighting coefficient of 0.5 and a safety distance gap of 0.55 meters. The system calculates that the lower-priority drone M02 needs to be offset laterally by 0.8 meters to spatially correct its trajectory coordinates during the conflict period.

[0074] Movement speed adjustment is a time-dimensional correction, referring to the fine-tuning of the instantaneous movement speed of low-priority equipment to avoid the spatiotemporal overlap of two devices, thus achieving temporal avoidance. Speed ​​adjustments are strictly limited to ±30% of the equipment's standard operating speed to prevent sudden speed changes that could cause equipment malfunctions or operational delays. In the example, if there is a low-risk (Level 3) collision hazard between the inspection robot and the drone, the system can fine-tune the robot's original movement speed of 1.2 meters per second to 0.9 meters per second, delaying the time it takes for the equipment to reach the conflict point, eliminating the collision risk without modifying its trajectory.

[0075] For each group of conflicting devices in the collision risk list, the system performs priority determination, risk level matching, and weight coefficient calculation. Based on the collision scenario, it selects an appropriate correction method and accurately calculates a complete set of correction parameters, including device trajectory offset, speed correction value, correction effective time, and correction duration. This ensures that each potential collision corresponds to a specific and precise correction plan. After all correction parameters are calculated, the system verifies their rationality, removing invalid parameters that exceed equipment operating thresholds or deviate from the park's operating range. Finally, following a standardized logic of equipment code, correction type, specific parameters, and effective time period, it generates a standardized set of avoidance correction parameters, providing accurate correction data for subsequent instruction updates.

[0076] The avoidance correction parameter set is written back to the original timing scheduling instruction set, replacing the relevant fields of the original motion instructions, and finally generating a collision-free motion sequence.

[0077] The core of this step is to complete the implementation and data update of the avoidance correction parameters, accurately synchronize the standardized correction parameters to the original timing scheduling instructions, replace the original motion parameters of the equipment that have potential collision risks, completely eliminate the risk of spatiotemporal collisions while retaining the original core scheduling logic, and generate a safe, compliant, and time-ordered sequence of equipment actions. The specific implementation method is as follows: Each mobile device motion instruction in the original time-series scheduling instruction set is equipped with modular editable fields. Only three types of dynamic motion fields, namely trajectory coordinates, motion speed, and short-term execution duration, are allowed to be modified. Core scheduling fields such as device task type, response priority, resource quota, overall execution time window, and linkage sequence are locked throughout the process and are not modified in any way. Following the principle of minimal modification, the scheduling results of previous task conflict resolution and resource pre-allocation are preserved to the greatest extent, ensuring the consistency and stability of the overall scheduling logic.

[0078] The system uses the unique device code and correction effective time in the avoidance correction parameter set as the basis for precise matching, achieving a one-to-one precise binding between correction parameters and original commands, eliminating problems such as parameter mismatch, omissions, and duplicate modifications. For correction scenarios involving fine-tuning of trajectory offsets, the system not only replaces the single-point trajectory coordinates at the conflict time node, but also smoothly updates all subsequent trajectory points associated with that node based on the principles of device motion inertia and trajectory continuity. This ensures a smooth and continuous overall trajectory for the device, avoiding unreasonable situations such as trajectory breaks or sudden angle changes, and closely reflects the actual motion characteristics of mobile devices. In the example, after correcting a 0.8-meter offset in the conflict trajectory of the M02 UAV, the system simultaneously updates the coordinate parameters of the subsequent eight consecutive trajectory nodes, ensuring a smooth and continuous flight trajectory.

[0079] In response to the correction scenario of movement speed adjustment, the system will simultaneously replace the instantaneous speed and average speed parameters in the instruction, and adaptively fine-tune the short-term execution duration of the corresponding work section. Through duration compensation, it ensures that the equipment's inspection range, monitoring duration, and work coverage fully meet the preset task requirements, and will not cause problems such as work omissions or task failures due to speed fine-tuning, thus ensuring that the quality of security operations is not affected by the correction operation.

[0080] After replacing all parameter fields, the system initiates a second full compliance check, again invoking the four-dimensional spatiotemporal collision detection algorithm to perform a full-domain, all-time re-check of the updated mobile device motion trajectories and execution sequences. This verifies that there are no spatiotemporal collision risks between devices or between devices and static obstacles in the park, completely eliminating all safety hazards. After passing the re-check, the system removes redundant verification data and invalid correction parameters from the instruction set. Following the original subsystem linkage sequence, task execution sequence, and priority sorting logic, it reorganizes all device action instructions, unifying the instruction data format and arrangement order.

[0081] The final standardized action instructions are integrated into a complete collision-free action sequence. This action sequence fully inherits the core logic of the previous multi-system linkage strategy, resource allocation scheme, and task scheduling sequence. At the same time, it completely solves the problem of spatiotemporal trajectory collision in multi-device linkage operations. The movement path, execution sequence, and operating status of all mobile devices meet the safety specifications of park security operations and can directly support the generation and implementation of subsequent intelligent scheduling solutions for security events.

[0082] S205, Based on the action sequence and the maximum response delay constraints of each subsystem, a security event intelligent scheduling scheme containing action instructions and execution time windows of each subsystem is generated using a static priority scheduling algorithm.

[0083] Specifically, the expected execution time and release time of each action instruction can be extracted from the action sequence, and the maximum response delay constraint value preset by each subsystem can be obtained to generate a task timing constraint matrix. The core of this step is to complete the collection and constraint definition of the time dimension parameters of all scheduling action instructions, integrate multiple types of time parameters to construct a standardized constraint matrix, provide a unified time benchmark for subsequent priority allocation and time scheduling, and eliminate the problem of conflict between instruction execution timing and subsystem latency specifications. The specific implementation method is as follows: The action sequence is a collision-free, standardized instruction sequence generated after spatiotemporal trajectory collision detection and dynamic avoidance correction. The sequence internally stores all action instructions that all security subsystems need to execute during this security incident response, including various practical instructions such as video surveillance capture, access control locking, perimeter alarm verification, fire equipment activation / deactivation, and robot patrol trajectory movement. All instructions have undergone trajectory conflict correction and possess the basic conditions for direct scheduling and execution. During parameter extraction, the system iterates through the complete action sequence line by line, accurately capturing two types of core time parameters corresponding to each independent action instruction: expected execution duration and release time. These two types of parameters are the basic timing parameters for instruction scheduling, possessing unique numerical definitions and physical meanings.

[0084] The expected execution time refers to the standard time required for a single action command to be fully executed from initiation, measured in milliseconds (ms) with a precision of 1ms. This parameter is generated by comprehensively calibrating the equipment operating performance, action execution logic, and equipment response rate of each security subsystem. Different types of action commands correspond to different fixed expected execution times. For example, the expected execution time for the panoramic capture command of the video surveillance subsystem is 200ms, the expected execution time for the remote locking command of the access control subsystem is 350ms, the expected execution time for the secondary signal verification command of the perimeter intrusion detection subsystem is 150ms, and the expected execution time for the fixed-point trajectory driving command of the inspection robot is 1200ms. The expected execution times of all commands are pre-stored in the system equipment parameter library and are directly matched and retrieved during extraction, ensuring the accuracy and consistency of the parameters.

[0085] Release time refers to the earliest time node at which a single action command can be scheduled and started, also measured in milliseconds (ms) with a precision of 1ms. This parameter is calibrated with the moment the security event instance is generated as the global time zero point, representing the start time at which the command can participate in system scheduling. Commands that have not reached their release time cannot be allocated execution resources and time windows by the system. Taking an illegal intrusion security event in a park as an example, the event generation time is defined as the global time zero point 0ms. The release time of the video surveillance tracking and capture command is 0ms, meaning that it can be scheduled and executed immediately after the event is triggered. The release time of the access control linkage blocking command is 100ms, meaning that it is necessary to wait 100ms after the event is triggered before scheduling can start. This parameter can avoid the instantaneous resource congestion problem caused by the simultaneous start of multiple commands.

[0086] The maximum response delay constraint is a pre-defined limit delay parameter for each security subsystem based on security level and emergency response specifications. It represents the maximum allowable time from receiving a dispatch command to completing the action, measured in milliseconds (ms) with a precision of 1ms. If the total execution time exceeds this value, it is considered a response timeout, leading to delays in security incident handling. The constraint values ​​for different subsystems are set differently based on their functional importance. Specifically, the emergency response delay constraint for the fire alarm subsystem is 1000ms, for the perimeter intrusion detection subsystem it is 800ms, for the video surveillance subsystem it is 1500ms, and for the access control subsystem it is 1200ms. This parameter is a hard constraint for subsequent timing scheduling; all scheduling results must meet this parameter requirement.

[0087] After extracting and retrieving all parameters, the system correlates the release time, expected execution duration, and maximum response latency constraint value of each action command with the corresponding subsystem. These parameters are then integrated and arranged according to the preset execution order to construct a task timing constraint matrix. This matrix uses a single action command as the smallest unit, fully encompassing all time constraints for each command. The data within the matrix is ​​mutually corresponding and without errors or omissions, clearly defining the schedulable start time, execution time, and maximum latency limit for each command. This provides precise constraints for subsequent static priority allocation and timing scheduling.

[0088] Assign static priorities to action instructions within each subsystem, with instructions directly related to the event having higher priority than auxiliary instructions, and those with lower response latency having higher priority, and generate a static priority allocation table; The core of this step is to establish standardized static priority determination rules based on event association attributes and latency constraint attributes, to hierarchically divide and prioritize action instructions of all subsystems, to clarify the order and weight of instruction scheduling, and to provide the core basis for time-series scheduling. The specific implementation method is as follows: Static priority is a pre-set, fixed command scheduling weight level for the current security event scheduling scenario. Unlike dynamically adjusted priority parameters, this priority is determined solely by two dimensions: the event correlation of the command and the subsystem response latency constraints. It remains unchanged during scheduling regardless of system resource status or task load, possessing stability and uniqueness. Priority is represented numerically; a smaller value indicates higher scheduling priority, allowing for priority access to system resources and execution time windows. A larger value indicates lower priority, requiring waiting for higher-priority commands to complete before scheduling.

[0089] Instructions are categorized into two types: directly related instructions and auxiliary instructions, which form the core basis for priority classification. Directly related instructions are those that directly affect security incident handling, quickly contain the incident, and collect core evidence. They directly determine the efficiency of security incident handling and are core scheduling tasks. Examples include perimeter signal verification, video trajectory tracking, and area access control blocking instructions in perimeter intrusion incidents. Auxiliary instructions are supporting instructions that assist core handling tasks, optimize the handling environment, and aid data recording. They do not directly participate in incident handling. Examples include full-process video recording and archiving, equipment operation status self-checks, and background log update instructions. In the priority rules, all directly related instructions have a default priority higher than auxiliary instructions, ensuring that core handling tasks are prioritized for scheduling.

[0090] Response latency is the core basis for prioritizing commands of the same type. Response latency corresponds to the maximum response latency constraint value of each subsystem. The smaller the maximum response latency constraint value, the higher the urgency of the corresponding subsystem's response, and the higher the command priority. For multiple directly related commands of the same type, the system compares the latency constraint parameters of the subsystems to which each command belongs to complete the sub-ranking. For example, the perimeter intrusion detection subsystem has a latency constraint of 800ms, and the corresponding intrusion signal verification command has a priority value set to 1; the fire protection subsystem has a latency constraint of 1000ms, and the corresponding fire alarm linkage command has a priority value set to 2; the access control subsystem has a latency constraint of 1200ms, and the corresponding area lockdown command has a priority value set to 3. The same rule applies to auxiliary commands: the video surveillance archiving command has a latency constraint of 1500ms and a priority value set to 10; the device self-test command has no hard latency constraint and a priority value set to 15.

[0091] After the system completes the type determination and priority assignment of all instructions, it integrates the instruction name, subsystem to which it belongs, instruction type, maximum response delay constraint, and priority value of each action instruction to generate a standardized static priority allocation table. This allocation table covers all instructions in this scheduling, and all priority assignments strictly follow preset rules without manual intervention or dynamic adjustment, ensuring the fairness and standardization of priority allocation. It clearly defines the scheduling weight level of each instruction, providing complete priority data support for the timing allocation of subsequent static priority scheduling algorithms.

[0092] A static priority scheduling algorithm is adopted to assign specific start and end times to each action instruction in descending order of priority, ensuring that each instruction meets the maximum response latency constraint and generating a task scheduling table with time windows; The core of this step relies on a static priority scheduling algorithm, using priority sorting as the core logic, combined with the time parameters of the task timing constraint matrix, to accurately allocate execution time intervals for all instructions, strictly avoid latency and timeout issues, and generate a standardized time-series scheduling form. The specific implementation method is as follows: The static priority scheduling algorithm is a time-series allocation algorithm for tasks with fixed priorities. Its core logic follows a pre-defined static priority ranking throughout the process. Instructions with lower priority values ​​are allocated execution time windows first, thus occupying system scheduling resources more frequently. The time-series allocation of high-priority instructions will not be preempted by lower-priority instructions. Simultaneously, the execution sequence of all instructions must strictly adhere to the maximum response latency constraints of their respective subsystems, ensuring the overall compliance of the scheduling. This algorithm is adapted to the fixed-priority characteristics of this security scheduling, enabling ordered and conflict-free time-series arrangement of multiple instructions. The algorithm's execution accuracy can reach 1ms, precisely controlling the instruction execution sequence at the millisecond level.

[0093] The first step in the algorithm execution is instruction sorting. The system reads all instruction data from the static priority allocation table and sorts them globally according to their priority values ​​from smallest to largest, forming an ordered instruction scheduling queue. High-priority core processing instructions are placed at the front of the queue, and low-priority auxiliary instructions are placed at the back of the queue. Subsequently, the algorithm retrieves the release time, expected execution duration, and maximum response delay constraint value of each instruction from the task timing constraint matrix as hard boundary conditions for timing allocation. The difference between the execution end time of any instruction and the zero point of the event must not exceed the corresponding maximum response delay constraint value.

[0094] During the scheduling process, the algorithm iterates through the queue starting with the highest priority instruction at the head of the queue, determining the earliest possible start time based on the instruction release time. If the current system scheduling time is greater than or equal to the instruction release time, a start time is immediately allocated to the instruction, and the end time is calculated by adding the expected execution duration to the start time. If the current system scheduling time is less than the instruction release time, the instruction release time is used as the start time to ensure that the instruction is not scheduled for execution prematurely. For example, the highest priority perimeter signal verification instruction has a release time of 0ms, an expected execution duration of 150ms, and a maximum delay constraint of 800ms. The algorithm allocates a start time of 0ms and an end time of 150ms, with the overall execution time being much less than the delay constraint value, thus meeting the scheduling requirements.

[0095] After allocating the timing for a single instruction, the algorithm automatically advances to the next higher priority instruction. Based on the end time of the previous instruction and the release time of the current instruction, the timing allocation is completed, and so on, until the time window allocation for all instructions is completed. After all instruction timing is allocated, the algorithm verifies the total execution time of each instruction, checking whether the total time from global time zero to the end of instruction execution is less than the maximum response delay constraint value of the corresponding subsystem. The timing parameters for critical states are fine-tuned to ensure that all instructions fully meet the delay constraints.

[0096] After all timing parameters are assigned and verified, the system integrates the subsystem to which all instructions belong, instruction content, priority level, execution start time, execution end time, effective time window range, and delay verification results to generate a task scheduling table with a time window. This table fully records the precise execution sequence of all action instructions and clarifies the legal execution time range of each instruction, serving as the core data carrier for subsequent schedulability analysis.

[0097] The task scheduling table is schedulable to verify that all instructions can be completed within the time window allowed by their respective subsystems. Finally, an intelligent scheduling scheme for security events is generated, which includes subsystem action instructions and execution time windows.

[0098] The core of this step is to perform a global compliance and feasibility check on the generated time-series scheduling results, identify issues such as timing conflicts, excessive latency, and abnormal window adaptation, and complete the compliance confirmation of the final scheduling results. This process integrates the results into a complete intelligent scheduling solution, and the specific implementation method is as follows: Scheduling analysis is a global verification mechanism for the task scheduling table. Its core objective is to verify whether the actual execution time interval of each action instruction falls entirely within the standard time window allowed by its subsystem, thus determining the feasibility of the overall scheduling scheme. The standard time window allowed by the subsystem is a legal time interval starting from the zero point of the security event trigger and ending at the subsystem's maximum response delay constraint value, measured in milliseconds (ms). Only when the complete execution time of an instruction is entirely encompassed within this interval can the scheduling be deemed compliant.

[0099] The analysis process employs a step-by-step verification method, covering all scheduling instructions in the task scheduling table without omission. The verification dimensions include two core aspects. The first is time window adaptability verification, which checks that the instruction execution start time is not earlier than the instruction release time and the execution end time is not later than the deadline corresponding to the subsystem's maximum response delay, ensuring that the instruction execution is within a legal time interval throughout. The second is timing compatibility verification, which checks that the execution timing of adjacent instructions does not overlap or intersect, relying on the previous trajectory collision correction results to ensure that there are no conflicts between the device action timing and spatial trajectory.

[0100] In the verification process, for conventional compliance instructions, the system directly marks them as schedulable and executable normally; for a very small number of instructions that are critically fitting to the time window, the system finely adjusts the execution start time by 1-5ms in combination with the idle resource status of the subsystems, ensuring that the entire execution of the instructions is within the legal window without affecting the overall scheduling sequence and event disposal efficiency. The whole analysis process is completed based on the system's automatic verification logic without manual intervention. The verification accuracy reaches 1ms, which can accurately identify subtle sequence deviations and window adaptation problems, and ensure the rigor of the scheduling result.

[0101] After completing the global schedulability analysis and all instructions pass the verification, the system integrates all compliant scheduling data, sorts out all action instructions corresponding to each security subsystem, the precise execution start time, execution end time, exclusive execution time window, priority level, and execution constraint conditions of each instruction, organizes and integrates them according to the logic of subsystem classification and sequence order, removes redundant verification data, retains core scheduling execution parameters, and finally generates a standardized intelligent security event scheduling scheme that can be directly implemented. The scheme completely covers all subsystem scheduling tasks for the current security event response, with accurate sequence, no conflict, compliance and controllability. It can directly drive each security subsystem to cooperatively complete linkage response and event disposal, and realize intelligent and precise sequence scheduling for park security events.

[0102] Another embodiment of the present invention provides an intelligent security event scheduling system for multi-system linkage in a park, see Figure 3 , the system may include: An acquisition module 301, configured to collect sensor data and alarm signals of a plurality of security subsystems deployed in the park in real time, perform fusion analysis on the sensor data and alarm signals based on spatial-temporal association rules, and generate security event instances with confidence labels; A calculation module 302, configured to dynamically calculate the linkage response priority of each security subsystem according to the event type and confidence label of the security event instance, in combination with the device topology relationship in the three-dimensional park map, and generate a multi-system linkage control strategy; An allocation module 303, configured to perform task conflict resolution and resource pre-allocation by adopting a distributed constraint satisfaction algorithm based on the multi-system linkage control strategy, in combination with the real-time resource occupation status and task load information of each security subsystem, and generate a conflict-free sequence scheduling instruction set; A correction module 304, configured to perform space-time trajectory collision detection and dynamic avoidance correction on the action instructions of each subsystem based on the sequence scheduling instruction set, and generate a collision-free action sequence; The scheduling module 305 is used to generate an intelligent security event scheduling scheme that includes action instructions and execution time windows of each subsystem based on the action sequence and the maximum response delay constraints of each subsystem using a static priority scheduling algorithm.

[0103] This invention also provides a storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above method embodiments when running.

[0104] This invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0105] Specifically, the aforementioned electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the aforementioned processor, and the input / output device is connected to the aforementioned processor.

[0106] The above description, based on the embodiments shown in the figures, details the structure, features, and effects of the present invention. The above description is only a preferred embodiment of the present invention, but the present invention is not limited to the scope of implementation shown in the figures. Any changes made in accordance with the concept of the present invention, or equivalent embodiments modified to have equivalent changes, that do not exceed the spirit covered by the specification and figures, should be within the protection scope of the present invention.

Claims

1. A method for intelligent scheduling of security incidents involving multiple systems in a park, characterized in that, The method includes: The system collects sensor data and alarm signals from multiple security subsystems deployed within the park in real time, and performs fusion analysis on the sensor data and alarm signals based on spatiotemporal correlation rules to generate security event instances with confidence labels. Based on the event type and confidence level label of the security event instance, and combined with the equipment topology in the 3D map of the park, the linkage response priority of each security subsystem is dynamically calculated to generate a multi-system linkage control strategy. Based on the aforementioned multi-system linkage control strategy, and combined with the real-time resource occupancy status and task load information of each security subsystem, a distributed constraint satisfaction algorithm is used to resolve task conflicts and pre-allocate resources, generating a set of conflict-free time-series scheduling instructions. Based on the set of time-series scheduling instructions, spatiotemporal trajectory collision detection and dynamic avoidance correction are performed on the action instructions of each subsystem to generate a collision-free action sequence. Based on the action sequence and the maximum response delay constraints of each subsystem, a static priority scheduling algorithm is used to generate an intelligent security event scheduling scheme that includes action instructions and execution time windows for each subsystem.

2. The method according to claim 1, characterized in that, The system collects real-time sensor data and alarm signals from multiple security subsystems deployed within the park, and performs fusion analysis on the sensor data and alarm signals based on spatiotemporal correlation rules to generate security event instances with confidence labels, including: By deploying video surveillance subsystems, access control subsystems, perimeter intrusion detection subsystems, and fire alarm subsystems within the park, raw sensor data streams and alarm signal streams are collected in real time to generate a multi-source heterogeneous real-time data stream set. The multi-source heterogeneous real-time data stream set is processed by timestamp alignment and spatial coordinate normalization, and the alarm signals of each subsystem are uniformly mapped to the grid coordinates of the park's 3D map to generate a spatiotemporally aligned event trigger record set. Based on the spatiotemporal association rule base, the sliding time window method is used to mine association patterns in the event trigger record set, calculate the spatial proximity and temporal continuity between alarm signals, and generate multi-source alarm association clusters. Confidence scores are assigned to each cluster in the multi-source alarm association cluster. The scores are then matched and verified against real event patterns calibrated in the historical sample database, and security event instances with confidence labels are output.

3. The method according to claim 2, characterized in that, The process involves dynamically calculating the linkage response priority of each security subsystem based on the event type and confidence label of the security event instance, combined with the device topology in the 3D map of the park, and generating a multi-system linkage control strategy, including: Parse the event type field and confidence label in the security event instance, determine the set of target subsystems that need to participate in the response through the event type-subsystem mapping table, and generate an initial response subsystem list; Based on the equipment topology relationships in the 3D map of the park, the connection paths and dependency levels between each equipment node in the target subsystem set are extracted to generate a equipment topology adjacency matrix. Combining confidence labels and topological adjacency matrices, the analytic hierarchy process (AHP) is used to calculate the response urgency index of each subsystem. The closer the center of the subsystem is to the event occurrence area and the higher the confidence level, the higher its priority. This generates a subsystem priority score vector. Based on the subsystem priority scoring vector, each subsystem is sorted from high to low according to its score and a linkage trigger delay time is set to generate a linkage control strategy that includes the subsystem number, response order and delay parameters.

4. The method according to claim 3, characterized in that, Based on the multi-system linkage control strategy, and combining the real-time resource occupancy status and task load information of each security subsystem, a distributed constraint satisfaction algorithm is used to resolve task conflicts and pre-allocate resources, generating a conflict-free time-series scheduling instruction set, including: Obtain the current real-time resource usage status and task load information from the device controllers of each security subsystem, including CPU utilization, communication bandwidth usage, and device busy / idle status, and generate a snapshot of the subsystem resource status. The multi-system linkage control strategy is parsed into a set of atomic tasks to be scheduled. Each atomic task includes the target subsystem, action type and required resource quota, and an atomic task queue is generated. Based on the subsystem resource status snapshot and atomic task queue, a distributed constraint satisfaction algorithm is used to search for feasible execution time slots for each atomic task, and conflict resolution is performed for tasks with overlapping time slots to generate a preliminary resource pre-allocation scheme. A global consistency check is performed on the initial resource pre-allocation scheme, the task order caused by resource mutual exclusion is adjusted, and finally a set of conflict-free time-series scheduling instructions is generated.

5. The method according to claim 4, characterized in that, Based on the set of time-series scheduling instructions, the action instructions of each subsystem are subjected to spatiotemporal trajectory collision detection and dynamic avoidance correction to generate a collision-free action sequence, including: Analyze each action instruction in the time-series scheduling instruction set, extract the position trajectory points and speed parameters of mobile devices such as inspection robots or drones, and generate a set of mobile device motion trajectory predictions. Based on the static obstacle layer of the 3D map of the park and the motion trajectory prediction set of mobile devices, a four-dimensional spatiotemporal collision detection algorithm is used to calculate the risk of any two mobile devices arriving at the same spatial location at the same time, and generate a collision risk pair list. For each pair of conflicting devices in the collision risk pair list, a priority-weighted adjustment method is used to fine-tune the trajectory or speed of the device with lower priority, generating a set of avoidance correction parameters. The avoidance correction parameter set is written back to the original timing scheduling instruction set, replacing the relevant fields of the original motion instructions, and finally generating a collision-free motion sequence.

6. The method according to claim 5, characterized in that, The method for generating an intelligent security event scheduling scheme, which includes action instructions and execution time windows for each subsystem, based on the action sequence and the maximum response delay constraints of each subsystem, using a static priority scheduling algorithm, includes: Extract the expected execution time and release time of each action instruction from the action sequence, and obtain the maximum response latency constraint value preset by each subsystem to generate a task timing constraint matrix; Assign static priorities to action instructions within each subsystem, with instructions directly related to the event having higher priority than auxiliary instructions, and those with lower response latency having higher priority, and generate a static priority allocation table; A static priority scheduling algorithm is adopted to assign specific start and end times to each action instruction in descending order of priority, ensuring that each instruction meets the maximum response latency constraint and generating a task scheduling table with time windows; The task scheduling table is schedulable to verify that all instructions can be completed within the time window allowed by their respective subsystems. Finally, an intelligent scheduling scheme for security events is generated, which includes subsystem action instructions and execution time windows.

7. A smart dispatching system for security incidents involving multiple systems in a park, characterized in that, The system includes: The data acquisition module is used to collect sensor data and alarm signals from multiple security subsystems deployed in the park in real time, and to perform fusion analysis on the sensor data and alarm signals based on spatiotemporal correlation rules to generate security event instances with confidence labels. The calculation module is used to dynamically calculate the linkage response priority of each security subsystem based on the event type and confidence label of the security event instance, combined with the device topology relationship in the 3D map of the park, and generate a multi-system linkage control strategy. The allocation module is used to resolve task conflicts and pre-allocate resources based on the multi-system linkage control strategy, combined with the real-time resource occupancy status and task load information of each security subsystem, and to generate a set of conflict-free time-series scheduling instructions by using a distributed constraint satisfaction algorithm. The correction module is used to perform spatiotemporal trajectory collision detection and dynamic avoidance correction on the action instructions of each subsystem based on the time-series scheduling instruction set, and generate a collision-free action sequence. The scheduling module is used to generate an intelligent security event scheduling scheme that includes action instructions and execution time windows for each subsystem based on the action sequence and the maximum response delay constraints of each subsystem using a static priority scheduling algorithm.

8. The system according to claim 7, characterized in that, The acquisition module is specifically used for: By deploying video surveillance subsystems, access control subsystems, perimeter intrusion detection subsystems, and fire alarm subsystems within the park, raw sensor data streams and alarm signal streams are collected in real time to generate a multi-source heterogeneous real-time data stream set. The multi-source heterogeneous real-time data stream set is processed by timestamp alignment and spatial coordinate normalization, and the alarm signals of each subsystem are uniformly mapped to the grid coordinates of the park's 3D map to generate a spatiotemporally aligned event trigger record set. Based on the spatiotemporal association rule base, the sliding time window method is used to mine association patterns in the event trigger record set, calculate the spatial proximity and temporal continuity between alarm signals, and generate multi-source alarm association clusters. Confidence scores are assigned to each cluster in the multi-source alarm association cluster. The scores are then matched and verified against real event patterns calibrated in the historical sample database, and security event instances with confidence labels are output.

9. A storage medium, characterized in that, The storage medium stores a computer program, wherein the computer program is configured to execute the method of any one of claims 1-6 when it is run.

10. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the method of any one of claims 1-6.

Images