Intelligent security control system and method based on simulation environment
By building an intelligent security control system with simulation environment and multi-agent reinforcement learning, the existing security system has solved the problems of poor environmental adaptability and isolation of sensing information, and achieved intelligent and dynamic security strategy optimization, improving the accuracy and defense capabilities of intrusion detection.
Patent Information
- Application Number
- CN202510347500.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-09
- Publication Date
- 2025-08-22
- Estimated Expiration
- 2045-05-09
AI Technical Summary
The existing intelligent security system has poor adaptability in the environment, lacks dynamic learning ability, isolated sensing information, and lacks intelligent simulation testing environment, resulting in insufficient detection accuracy and effectiveness of defense measures.
Build an intelligent security control system based on a simulation environment, and use random generation modules, virtual sensing modules and multi-agent reinforcement learning to simulate intrusion behavior and defense strategies, and dynamically optimize them based on actual monitoring data.
It improves the intelligence level and adaptability of the security system, enhances the accuracy of intrusion detection and the effectiveness of defense measures, and is suitable for intelligent security, unmanned monitoring and other scenarios.
Smart Images

Figure CN120526516A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of intelligent security technology, and in particular to an intelligent security control system and method based on a simulation environment. Background Art
[0002] With the development of artificial intelligence, the Internet of Things, and intelligent monitoring technologies, security systems are increasingly being used in scenarios such as smart cities, corporate campuses, and residential communities. Current intelligent security systems typically rely on fixed monitoring equipment (such as cameras and sensors) for intrusion detection and provide alerts and defenses using pre-set rules or traditional machine learning algorithms. However, these traditional security systems have the following technical limitations: 1. Poor environmental adaptability: Existing security systems typically analyze data based on fixed rules or static data, making them difficult to adapt to complex and changing intrusion behaviors and environmental changes. For example, factors such as camera monitoring angle, lighting conditions, and sensor sensitivity can affect detection accuracy.
[0003] 2. Lack of dynamic learning capabilities: Most security systems are unable to self-optimize based on real-world feedback, making them difficult to respond to new or subtle intrusions. Traditional systems are often slow to react to attackers' evasive tactics and lack the ability to continuously learn and dynamically adjust.
[0004] 3. Isolated sensor information: In current security systems, different types of sensors (such as infrared detectors, radars, cameras, etc.) usually operate independently, with low data fusion. This makes the detection results susceptible to false positives or omissions from a single sensor, reducing the overall reliability of the system.
[0005] 4. Lack of intelligent simulation testing environments: In practical applications, security system testing and optimization often rely on real-world intrusion testing, which is costly and risky. Existing simulation testing environments often lack realism and cannot accurately reflect complex intrusion behaviors and abnormal situations.
[0006] To address the above problems, the present invention proposes an intelligent security control system based on a simulation environment. By constructing a simulation environment, a virtual sensing system, and combining multi-agent reinforcement learning, it realizes dynamic simulation of intrusion behavior and defense optimization. Summary of the Invention
[0007] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide an intelligent security control system and method based on a simulation environment, which is used to dynamically optimize security strategies, improve the accuracy of intrusion detection and the effectiveness of defense measures.
[0008] To achieve the above objectives, the present invention provides the following technical solutions: an intelligent security control system based on a simulation environment, comprising: A random generation module is used to build a simulation environment for the intelligent security system and randomly generate a plurality of change parameters in the simulation environment, wherein the change parameters include simulation environment parameters and sensor abnormality parameters; A virtual sensing module is used to model multiple types of sensors to obtain multiple virtual sensing models, wherein the virtual sensing models are used to output simulated sensing data, and to fuse the simulated sensing data to obtain simulated fusion data; a collaborative simulation module, connecting the random generation module and the virtual sensing module, and configured to construct an intrusion agent, a defense agent, and a control agent in the simulation environment according to a pre-introduced multi-agent reinforcement learning framework, wherein the intrusion agent is configured to simulate and generate a plurality of simulated intrusion behaviors and simulated response strategies, the control agent is configured to generate a defense strategy based on the simulation fusion data when faced with each of the simulated intrusion behaviors and simulated response strategies of the intrusion agent, and the defense agent is configured to adjust the camera angle, alarm threshold, and lighting status according to the defense strategy, the simulation environment parameters, and the sensor anomaly parameters; The actual monitoring module is used to obtain actual environmental parameters, actual abnormal parameters, actual intrusion behavior, actual sensor fusion data, and actual response strategies; a parameter comparison module, connected to the collaborative simulation module and the actual monitoring module, for comparing the actual environmental parameters, the actual abnormal parameters, the actual intrusion behavior, the actual sensor fusion data, and the actual strain strategy with the simulated environmental parameters, the sensor abnormal parameters, the simulated intrusion behavior, the simulated sensor data, and the simulated strain strategy to obtain a plurality of learning parameters, the learning parameters including environmental deviation data, abnormal deviation data, intrusion deviation data, sensor fusion deviation data, and strain deviation data; An optimization module is connected to the parameter comparison module and the collaborative simulation module, and is used to perform dynamic correction optimization on the intrusion agent, the defense agent and the control agent according to each of the learning parameters.
[0009] Furthermore, the random generation module includes: An environment generation unit is configured to randomly match and generate a plurality of simulated environment parameters from a preset environment database based on the positioning data and time data of the location of the intelligent security control system based on the simulated environment, wherein the simulated environment parameters include simulated temperature, simulated humidity, simulated air quality, simulated lighting parameters, simulated electromagnetic interference data, and simulated biological interference data; The abnormality introduction unit is used to introduce the sensor abnormality parameters under the abnormal state of the sensor in the simulation environment, and the sensor abnormality parameters include simulated sensor noise data, simulated signal interruption data, simulated sensor conflict data, and simulated link abnormality data.
[0010] Furthermore, the virtual sensing module includes: a sensor modeling unit, configured to construct a plurality of virtual sensor models, wherein the virtual sensor models are configured to simulate response curves, delays, noise, and errors of various sensors, and output simulated sensor data, wherein the simulated sensor data includes simulated video data, simulated infrared data, simulated audio data, and simulated vibration data; The preliminary fusion unit is connected to the sensor modeling unit and is used to construct a multi-sensor data fusion algorithm in the simulation environment to sequentially perform spatiotemporal alignment, information extraction and data fusion on the simulated sensor data from different sources to obtain simulated fusion data.
[0011] Furthermore, the data fusion calculation formula of the multi-sensor data fusion algorithm is configured as:
[0012]
[0013]
[0014] in, represents the simulation fusion data, Indicates the total time of fusion data calculation, represents a discrete series index, express The set of dimensional real vectors, Represents the Gamma function, which is used to control the scale change of data. Represents the Riemann Zeta function, which is used to enhance the nonlinear mapping of data. represents the first kind of Bessel function, which is used to control the smoothness of the data. Represents a complex information filtering function, which is used to optimize data fusion by combining error function, logarithmic function and elliptic integral. Represents different types of sensor data, where To simulate video data, To simulate infrared data, To simulate audio data, To simulate vibration data, Represents the weight of various sensor data, which is used to distribute according to the confidence and contribution of the data. Represents the index of sensor data in the spatial domain, is the filter input variable, is the integration variable, is the first kind of elliptic integral, which is used to correct the nonlinear error of the data.
[0015] Furthermore, a framework construction module is included, connected to the collaborative simulation module, and the framework construction module is used to construct a hierarchical reinforcement learning framework and includes: an outer global strategy unit, configured to generate security strategy sub-goals based on hierarchical deep reinforcement learning, and transmit the security strategy sub-goals to the control agent, wherein the control agent generates the defense strategy based on the security strategy sub-goals; The inner local policy unit is used to perform fine-grained adjustments to the camera, alarm device, and lighting, and train the adjustment method based on the policy gradient method to optimize the local control tasks of the defense agent on the camera angle, the alarm threshold, and the lighting status.
[0016] Furthermore, the hierarchical reinforcement learning framework adopts a hierarchical reward structure, which includes an overall reward, an outer reward, and an inner reward. The conversion relationship between the overall reward, the outer reward, and the inner reward is configured as follows:
[0017] in, represents the overall reward, represents the outer reward, represents the inner layer reward, represents the outer layer weight, represents the inner layer weight, Indicates the total amount of inner layer rewards.
[0018] Furthermore, the actual environmental parameters include actual temperature, actual humidity, actual air quality, actual lighting parameters, actual electromagnetic interference data, and actual biological interference data. The framework construction module also includes an adaptive weight adjustment unit, which is used to establish a conversion relationship with the outer layer weight and the inner layer weight according to the actual environmental parameters, and dynamically adjust the outer layer weight and the inner layer weight according to the actual environmental parameters, wherein the conversion relationship between the actual environmental parameters and the outer layer weight and the inner layer weight is configured as follows:
[0019] in, represents the actual temperature, represents the actual humidity, represents the actual air quality, represents the actual lighting parameters, represents the actual electromagnetic interference data, represents the actual biological interference data, represents the average value of the actual temperature, represents the average value of the actual humidity, represents the average value of the actual air quality, represents the average value of the actual lighting parameter, represents the average value of the actual electromagnetic interference data, represents the average value of the actual biological interference data, 、 、 、 、 、 They respectively represent the value ranges of the actual temperature, the actual humidity, the actual air quality, the actual lighting parameters, the actual electromagnetic interference data, and the actual biological interference data.
[0020] Furthermore, the actual abnormality parameters include actual video data, actual infrared data, actual audio data, and actual vibration data. The framework construction module also includes a reward adjustment unit, which is used to calculate a sensor comprehensive abnormality value based on the actual video data, actual infrared data, actual audio data, and actual vibration data, and dynamically adjust the overall reward based on the sensor comprehensive abnormality value. The conversion relationship between the sensor comprehensive abnormality value and the overall reward is configured as follows:
[0021]
[0022] in, represents the sensor comprehensive abnormal value, represents the actual video data abnormal value, represents the abnormal value of the actual infrared data, represents the actual audio data abnormal value, represents the abnormal value of the actual vibration data, 、 、 、 Represent the mean of each sensor data, 、 、 、 Respectively represent the value range of each sensor data, Represents the preset basic overall reward, Represents the abnormality suppression coefficient, which is used to control the degree of attenuation of the overall reward by abnormal values. Indicates the preset abnormal compensation coefficient, which is used to prevent the reward value from being excessively reduced. Used to provide smooth abnormal adjustment effects and prevent sudden changes.
[0023] Furthermore, the hierarchical reinforcement learning framework includes an edge computing device and a cloud server, wherein the edge computing device is used to execute the local control task in real time, and the cloud server is used to continuously update the defense strategy.
[0024] An intelligent security control method based on a simulation environment, applied to the above-mentioned intelligent security control system based on a simulation environment, comprises: Step S1, a random generation module constructs a simulation environment of an intelligent security system and randomly generates a plurality of change parameters in the simulation environment, wherein the change parameters include simulation environment parameters and sensor abnormality parameters; Step S2: The virtual sensing module models multiple types of sensors to obtain multiple virtual sensing models, wherein the virtual sensing models are used to output simulated sensing data, and the simulated sensing data are fused to obtain simulated fusion data. In step S3, the collaborative simulation module constructs an intrusion agent, a defense agent, and a control agent in the simulation environment according to a pre-introduced multi-agent reinforcement learning framework. The intrusion agent is used to simulate and generate a variety of simulated intrusion behaviors and simulated response strategies. The control agent is used to generate a defense strategy based on the simulation fusion data when facing each of the simulated intrusion behaviors and simulated response strategies of the intrusion agent. The defense agent is used to adjust the camera angle, alarm threshold, and lighting status according to the defense strategy, the simulation environment parameters, and the sensor abnormality parameters. Step S4, the actual monitoring module obtains actual environmental parameters, actual abnormal parameters, actual intrusion behavior, and actual response strategies; Step S5: a parameter comparison module compares the actual environment parameters, the actual abnormal parameters, the actual intrusion behavior, and the actual strain strategy with the simulated environment parameters, the sensed abnormal parameters, the simulated intrusion behavior, and the simulated strain strategy to obtain a plurality of learning parameters, wherein the learning parameters include environment deviation data, abnormality deviation data, intrusion deviation data, and strain deviation data. Step S6: The optimization module performs dynamic correction optimization on the intrusion agent, the defense agent and the control agent according to each of the learning parameters.
[0025] Beneficial effects of the present invention: By introducing simulation environment, virtual sensing technology and multi-agent reinforcement learning, this invention realizes intelligent and dynamic security strategy optimization, breaks through the limitations of traditional security systems, significantly improves the intelligence level, adaptability and defense capabilities of security systems, improves the accuracy of intrusion detection and the effectiveness of defense measures, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] Figure 1 It is a structural diagram of the intelligent security control system of the present invention; Figure 2 It is a flowchart of the steps of the intelligent security control method in the present invention.
[0027] Figure numerals: 1. Random generation module; 11. Environment generation unit; 12. Abnormal introduction unit; 2. Virtual sensing module; 21. Sensor modeling unit; 22. Preliminary fusion unit; 3. Collaborative simulation module; 4. Actual monitoring module; 5. Parameter comparison module; 6. Optimization module; 7. Framework construction module; 71. Outer global strategy unit; 72. Inner local strategy unit; 73. Adaptive weight adjustment unit; 74. Reward adjustment unit. DETAILED DESCRIPTION
[0028] The present invention will be described in further detail below with reference to the accompanying drawings and embodiments. Identical components are denoted by the same reference numerals. It should be noted that the terms "front," "rear," "left," "right," "upper," and "lower" used in the following description refer to directions in the accompanying drawings, and the terms "bottom," "top," "inner," and "outer" refer to directions toward or away from the geometric center of a particular component, respectively.
[0029] Example 1, reference Figure 1 , which is the first embodiment of the present invention, provides an intelligent security control system based on a simulation environment, which can improve the accuracy of intrusion detection and the effectiveness of defense measures, including: Random generation module 1 is used to build a simulation environment for the intelligent security system and randomly generate multiple changing parameters in the simulation environment, the changing parameters including simulation environment parameters and sensor abnormality parameters; The virtual sensing module 2 is used to model multiple types of sensors to obtain multiple virtual sensing models, the virtual sensing models are used to output simulated sensing data, and the simulated sensing data are fused to obtain simulated fusion data; Co-simulation module 3, connecting random generation module 1 and virtual sensing module 2, is used to construct an intrusion agent, a defense agent, and a control agent in the simulation environment based on the pre-introduced multi-agent reinforcement learning framework. The intrusion agent is used to simulate and generate a variety of simulated intrusion behaviors and simulated response strategies. The control agent is used to generate a defense strategy based on the simulation fusion data when faced with the simulated intrusion behaviors and simulated response strategies of the intrusion agent. The defense agent is used to adjust the camera angle, alarm threshold, and lighting status according to the defense strategy, simulation environment parameters, and sensor anomaly parameters. Actual monitoring module 4, used to obtain actual environmental parameters, actual abnormal parameters, actual intrusion behavior, and actual response strategies; The parameter comparison module 5 is connected to the collaborative simulation module 3 and the actual monitoring module 4, and is used to compare the actual environmental parameters, actual abnormal parameters, actual intrusion behavior and actual strain strategy with the simulated environmental parameters, sensor abnormal parameters, simulated intrusion behavior and simulated strain strategy to obtain multiple learning parameters, the learning parameters including environmental deviation data, abnormal deviation data, intrusion deviation data and strain deviation data; The optimization module 6 is connected to the parameter comparison module 5 and the collaborative simulation module 3, and is used to dynamically correct and optimize the intrusion agent, the defense agent and the control agent according to each learning parameter.
[0030] Working principle of embodiment 1: Random Generation Module 1 is used to build an intelligent security simulation environment and randomly generate multiple simulation environment parameters and sensor anomaly parameters within the environment to simulate different external conditions and sensor anomalies. Simulation environment parameters include temperature, humidity, air quality, lighting parameters, electromagnetic interference data, and biological interference data. Sensor anomaly parameters include sensor noise data, signal interruption data, sensor conflict data, and link anomaly data, such as increased signal noise for cameras in low-light environments, false detections for infrared sensors in high-temperature environments, and radar data drift due to electromagnetic interference.
[0031] The goal of Random Generation Module 1 is to generate diverse simulation data to ensure the security system can adapt to complex environments. Virtual Sensing Module 2 models various sensors (such as cameras, infrared sensors, pressure sensors, vibration sensors, and radar) and outputs simulated sensor data. This data is derived from Random Generation Module 1 and simulates the sensors' operating conditions in various environments. For example, for cameras, the simulation targets recognition capabilities in daytime, nighttime, low light, and bright light conditions; for infrared sensors, the simulation simulates the infrared signatures of high and low temperatures, and dynamic targets; and for radar, the simulation simulates the echo changes of microwave radars under signal interference.
[0032] All sensor data is fused and processed to form simulated fusion data, improving the system's data accuracy. For example, a camera detects the movement of a suspicious target; an infrared sensor detects abnormal target temperature; and a radar analyzes changes in target distance and speed. This fused data serves as input for subsequent intelligent agent learning and optimization.
[0033] In this embodiment, a multi-agent reinforcement learning framework is used to construct an intrusion agent, a defense agent, and a control agent in a simulation environment to form a dynamic defense mechanism: Intrusion agent: simulates illegal intrusion behaviors, such as climbing over walls, blocking cameras, signal interference, password cracking, etc., and optimizes its attack strategy to enhance the system's defense capabilities.
[0034] Defense agent: Based on simulation fusion data and defense strategies, it adjusts camera angles, increases alarm sensitivity, activates strong lighting, etc. to respond to intrusions in real time.
[0035] Control agents: Based on the attack methods of the intrusion agents and the response strategies of the defense agents, optimize alarm thresholds, sensor parameters, camera patrol paths, etc., to reduce false alarms and missed alarms and improve system adaptability.
[0036] The actual monitoring module 4 is deployed in a real environment and is responsible for collecting real-time data, including: actual environmental parameters, actual abnormal parameters, actual intrusion behavior, and actual response strategies; These data are used to compare with simulation data to optimize system defense strategies.
[0037] Parameter comparison module 5 compares simulated data with real data and calculates key deviations. These include environmental deviation (the difference between simulated and real-world environmental parameters); anomaly deviation (the difference between simulated sensor anomalies and actual sensor false alarms); intrusion deviation (the degree of match between simulated intrusion behavior and real-world intrusion events); and strain deviation (the difference between the defense agent's performance in simulation and real-world scenarios, as well as sensor fusion deviation data).
[0038] If the simulation system predicts an 80% probability of nighttime intrusion, but actual monitoring data shows only 50%, the simulation environment needs to be adjusted to make it closer to the real situation; If a certain type of false alarm occurs frequently, the system's anomaly detection algorithm is optimized to improve the response accuracy of the defense agent.
[0039] Optimization Module 6 dynamically optimizes the agent based on the data provided by Parameter Comparison Module 5. Reinforcement Learning Optimization: Continuously adjust the agent's strategy to improve the accuracy of identifying intrusions. For example, after repeated training, the defense agent may notice a high incidence of nighttime intrusions and automatically increase the scanning frequency of patrol cameras. By adjusting the alarm threshold, false alarms are reduced, improving the system's practicality.
[0040] When there is a large difference between the abnormal situation in the real environment (such as foggy weather) and the simulation environment, the optimization module 6 will update the simulation model to be closer to the reality.
[0041] Application of this embodiment in intelligent residential communities: A community has installed this intelligent security system. One evening, a suspicious person attempts to break in by climbing over the wall. The system's workflow is as follows: The infrared sensor detected abnormal temperatures near the wall, and the camera simultaneously captured suspicious targets; The simulation system analyzed the behavior and found that it matched well with previous simulated intrusion behaviors; The defense agent automatically adjusts the camera angle to track the target and improves the alarm sensitivity; The control agent analyzes the current environment, determines whether to turn on strong lighting to deter the target, and activates the alarm system; The actual monitoring module 4 records the event and inputs its data into the parameter comparison module 5 for optimizing subsequent defense strategies.
[0042] This example demonstrates how a simulation-based intelligent security control system can optimize defense strategies within the simulation and implement intelligent defenses in real-world environments. Through reinforcement learning and dynamic adjustments, the system improves the accuracy of intrusion detection and enhances the security system's adaptability and defense capabilities. It is suitable for a variety of scenarios, including intelligent security and unmanned surveillance.
[0043] Preferably, the random generation module 1 includes: An environment generation unit 11 is configured to randomly match and generate a plurality of simulated environment parameters from a preset environment database based on the location data and time data of the intelligent security control system based on the simulated environment, wherein the simulated environment parameters include simulated temperature, simulated humidity, simulated air quality, simulated lighting parameters, simulated electromagnetic interference data, and simulated biological interference data; The abnormality introduction unit 12 is used to introduce sensor abnormality parameters under the abnormal state of the sensor in the simulation environment. The sensor abnormality parameters include simulated sensor noise data, simulated signal interruption data, simulated sensor conflict data, and simulated link abnormality data.
[0044] Specifically, in this embodiment, the environment generation unit 11 matches multiple simulation environment parameters from a preset environment database based on the positioning data (GPS coordinates) and time data (current time, season, day and night information) of the location of the intelligent security system to simulate different weather, geographical environments and environmental interference factors.
[0045] Among them, the simulated temperature is based on the geographic location and time, and historical temperature data is queried from the meteorological database, and fluctuation factors are randomly added to enhance the diversity of the simulated environment; the simulated humidity is randomly set according to the temperature and weather conditions to ensure that it matches the temperature; the simulated air quality is generated by historical air quality index (AQI) data, and the time period is taken into account (such as higher AQI during peak hours in the morning and evening); the simulated lighting parameters are obtained by combining the sunshine angle and weather (sunny / cloudy / rainy days); the simulated electromagnetic interference data is formed by simulating the random noise interference generated by the urban electromagnetic environment (such as WiFi, 5G base stations, and industrial equipment); the simulated biological interference data is used to simulate the interference of small animals, tree swings, etc. on the sensor to test the false alarm rate.
[0046] In order to improve the intelligent security system's ability to respond to abnormal situations, the abnormality introduction unit 12 actively introduces sensor abnormal states in the simulation environment to simulate sensor failures or interference situations that may occur in actual deployment.
[0047] Among them, simulated sensor noise data is used to simulate noise from sensors such as cameras, infrared sensors, and sound sensors, such as image noise, audio background noise, and temperature drift noise. This data is simulated using a Gaussian noise model. Simulated signal interruption data simulates signal loss and transmission delays to ensure the system's resilience to network fluctuations. This data uses a Poisson process to simulate the probability of signal loss. Simulated sensor conflict data: When multiple sensors simultaneously operate in the same area, signal interference can lead to data conflicts. For example, data from cameras and infrared sensors may not match, or multiple radar signals may interfere with each other. This data uses covariance analysis to calculate the degree of conflict. Simulated link anomaly data simulates network fluctuations, packet loss, bandwidth limitations, and other conditions, enabling the security system to cope with actual network failures. This data is calculated by multiplying the network packet loss rate and bandwidth fluctuation.
[0048] This embodiment demonstrates how the environment generation unit 11 and the anomaly introduction unit 12 work together to dynamically build a high-realism simulation environment and actively introduce various sensor anomalies to improve the adaptability, anti-interference ability and fault tolerance of the intelligent security system, ensuring stable operation in complex environments.
[0049] Preferably, the virtual sensing module 2 includes: The sensor modeling unit 21 is used to construct multiple virtual sensor models, which are used to simulate the response curves, delays, noise and errors of various sensors and output simulated sensor data, including simulated video data, simulated infrared data, simulated audio data and simulated vibration data; The preliminary fusion unit 22 is connected to the sensor modeling unit 21 and is used to construct a multi-sensor data fusion algorithm in a simulation environment to sequentially perform spatiotemporal alignment, information extraction and data fusion on various simulated sensor data from different sources to obtain simulated fusion data.
[0050] Specifically, in this embodiment, the sensor modeling unit 21 constructs multiple virtual sensor models based on the physical characteristics and response mechanisms of real sensors to simulate the response curves, delays, noise, and errors of different sensors. The following sensor types are mainly involved in this system: Video sensor (camera): used for visible light environment monitoring, responding to lighting changes and generating simulated video data.
[0051] Infrared sensor: used to detect heat sources, significantly affected by ambient temperature, and outputs simulated infrared data.
[0052] Audio sensor (microphone): used to monitor ambient sound, which may be affected by background noise, and output simulated audio data.
[0053] Vibration sensor: used to detect ground or object vibrations, identify abnormal movements, and output simulated vibration data.
[0054] Sensor response curve modeling: The response characteristics of each sensor are described by a response function. The standard response curve of the sensor needs to be modeled based on its physical characteristics: The video sensor uses a linear Gamma correction model, the infrared sensor uses a nonlinear temperature response model, the audio sensor uses a spectrum analysis model, and the vibration sensor uses an elastic damping model.
[0055] Noise and error terms are used to simulate the noise interference of real sensors. Noise is simulated by Gaussian noise, and error is simulated by signal drift.
[0056] Generation of simulated sensor data: Run the sensor model in a simulation environment to generate simulated video data, simulated infrared data, simulated audio data, and simulated vibration data to provide complete multimodal perception information.
[0057] Implementation of preliminary fusion unit 22: (1) Data spatiotemporal alignment Since different sensors have different sampling frequencies and time delays, spatiotemporal alignment is necessary. Dynamic Time Warping (DTW) is used to adjust the timestamps of different sensor data.
[0058] (2) Information extraction Extract key features from various sensor data: For video data, target detection (YOLO) and motion trajectory analysis are used to extract key features; for infrared data, heat source distribution and target temperature changes are used to extract key features; for audio data, voiceprint recognition and abnormal sound source location are used to extract key features; and for vibration data, vibration amplitude and frequency pattern are used to extract key features.
[0059] (3) Data fusion calculation Utilize the data fusion calculation formula and combine multi-sensor data to calculate the simulated fusion data.
[0060] This embodiment uses the virtual sensor module 2 to build a simulation environment, and utilizes the sensor modeling unit 21 and preliminary fusion unit 22 to complete data modeling and fusion. This effectively improves the realism of the simulation environment and enhances its adaptability to complex environments. It also reduces sensor false alarm rates by eliminating single sensor errors through data fusion. It also optimizes system decision-making capabilities and provides more accurate input data for intelligent security control strategies.
[0061] This embodiment realizes the modeling, spatiotemporal alignment and fusion of multi-sensor data, ensuring that the intelligent security system can accurately simulate the real situation in the simulation environment and optimize the defense strategy.
[0062] Preferably, the data fusion calculation formula of the multi-sensor data fusion algorithm is configured as:
[0063]
[0064]
[0065] in, represents the simulation fusion data, Indicates the total time of fusion data calculation, represents a discrete series index, express The set of dimensional real vectors, Represents the Gamma function, which is used to control the scale change of data. Represents the Riemann Zeta function, which is used to enhance the nonlinear mapping of data. represents the first kind of Bessel function, which is used to control the smoothness of the data. Represents a complex information filtering function, which is used to optimize data fusion by combining error function, logarithmic function and elliptic integral. Represents different types of sensor data, where To simulate video data, To simulate infrared data, To simulate audio data, To simulate vibration data, Represents the weight of various sensor data, which is used to distribute according to the confidence and contribution of the data. Represents the index of sensor data in the spatial domain, is the filter input variable, is the integration variable, is the first kind of elliptic integral, which is used to correct the nonlinear error of the data.
[0066] Specifically, in this embodiment, The value range of the integral interval And the distribution impact of each sensor data: When the sensor data signal is strong and consistent, A larger value indicates higher data fusion quality; when the sensor data is noisy or the information does not match, Small values indicate high uncertainty in the fused data. This formula ensures data fusion stability and nonlinear mapping through a combination of gamma, Riemann Zeta, and Bessel functions. It also optimizes information using error functions and elliptic integrals to minimize data fusion errors. This formula accurately describes the multi-sensor data fusion process and combines multiple complex mathematical functions to achieve nonlinear mapping, noise suppression, and optimal data fusion strategies. It can be used for simulation data fusion tasks in complex scenarios such as intelligent security, autonomous driving, and multimodal recognition.
[0067] Embodiment 2 is the second embodiment of the present invention. Unlike the previous embodiment, this embodiment provides a framework construction module 7, which can realize efficient strategy optimization and intelligent control of the intelligent security system. The framework construction module 7 is connected to the collaborative simulation module 3. The framework construction module 7 is used to build a hierarchical reinforcement learning framework and includes: The outer global strategy unit 71 is used to generate security strategy sub-goals based on hierarchical deep reinforcement learning, and transmit the security strategy sub-goals to the control agent, which generates a defense strategy based on the security strategy sub-goals; The inner local policy unit 72 is used to perform fine-grained adjustments to cameras, alarm devices, and lights, and train the adjustment methods based on the policy gradient method to optimize the defense agent's local control tasks of camera angles, alarm thresholds, and lighting status.
[0068] Preferably, the hierarchical reinforcement learning framework adopts a hierarchical reward structure, which includes an overall reward, an outer reward, and an inner reward. The conversion relationship between the overall reward, the outer reward, and the inner reward is configured as follows: in, It represents the overall reward, representing the defense effect of the entire intelligent security system; Represents the outer layer reward, which is used to evaluate the long-term security and resource optimization effects of the overall security strategy; Represents the inner reward, which is used to evaluate the real-time effect of local control strategies (camera, alarm, lighting) in a short period of time; represents the outer layer weight, represents the inner layer weight, Indicates the total amount of inner layer rewards Working principle of embodiment 2: The framework construction module 7 uses hierarchical deep reinforcement learning and includes an outer global policy unit 71 and an inner local policy unit 72. The outer global policy unit 71 is responsible for formulating the overall security strategy, generating sub-goals, and passing them to the control agent. The outer global policy unit 71 is based on hierarchical deep reinforcement learning (HDRL) and adopts an options framework to enable the outer agent to generate optimal sub-goals in different scenarios. Task examples: High alert mode (increases monitoring intensity when a threat is detected), energy-saving monitoring mode (optimizes energy consumption when there is no threat), and full deployment mode (increases monitoring intensity in the entire area).
[0069] The outer global strategy unit 71 performs reinforcement learning training based on sensor data fusion results, historical alarm data, security events, etc. The outer agent passes the generated security strategy sub-goals to the control agent, which optimizes the defense strategy based on these sub-goals.
[0070] The inner local policy unit 72 is responsible for fine-grained adjustments to cameras, alarms, and lighting to improve local defense efficiency. This unit is trained using policy gradient methods and uses agent-based reinforcement learning to optimize local tasks such as camera angles, alarm thresholds, and lighting control.
[0071] Task examples: Camera control: Adjust camera angles to maximize surveillance coverage. Alarm optimization: Optimize alarm sensitivity based on sensor data to reduce false alarms. Lighting control: Optimize lighting based on nighttime safety needs to ensure visibility.
[0072] Preferably, the actual environmental parameters include actual temperature, actual humidity, actual air quality, actual lighting parameters, actual electromagnetic interference data, and actual biological interference data. The framework construction module 7 further includes an adaptive weight adjustment unit 73, which is used to establish a conversion relationship with the outer layer weight and the inner layer weight according to the actual environmental parameters, and dynamically adjust the outer layer weight and the inner layer weight according to the actual environmental parameters, wherein the conversion relationship between the actual environmental parameters and the outer layer weight and the inner layer weight is configured as follows:
[0073] in, Indicates the actual temperature, Indicates the actual humidity. Indicates the actual air quality, Represents the actual lighting parameters, Indicates actual electromagnetic interference data, represents the actual biological disturbance data, Indicates the average value of the actual temperature, Indicates the average value of actual humidity, Indicates the average value of actual air quality, Represents the average value of actual lighting parameters, Represents the average value of actual electromagnetic interference data, represents the average value of actual biological disturbance data, 、 、 、 、 、 They respectively represent the value ranges of actual temperature, actual humidity, actual air quality, actual lighting parameters, actual electromagnetic interference data, and actual biological interference data.
[0074] Specifically, in this embodiment, the formula uses a standardized method to normalize various environmental parameters to ensure that different physical quantities have the same dimension so that they can be processed within the same calculation framework.
[0075] Controlled by Sigmoid function The value range is between (0,1), which allows it to smoothly adjust the global weight and ensure stability. Depend on Make sure that the sum of the global and local policy weights is 1, so that it can be adjusted adaptively.
[0076] When the external environment deviates from the average value (that is, the external disturbance is more severe), Approaching 1, the global strategy is more important to ensure system stability; on the contrary, when the environment is stable, As it approaches 1, local strategies are more important to optimize resource efficiency.
[0077] When the environment is completely stable (all parameters are equal to their mean values): the sum of the normalized terms is 0, , =0.5, at this time and balance.
[0078] When the environment fluctuates violently (a parameter is far from the mean): the absolute value of the normalized term increases, and the exponential term tends to be extremely large or extremely small. Approaching 1 or 0 makes the global or local strategy dominant.
[0079] This formula can be used in intelligent security systems, enabling the system to adaptively adjust decision weights according to environmental changes and optimize the balance between global and local strategies.
[0080] Preferably, the actual abnormality parameters include actual video data, actual infrared data, actual audio data, and actual vibration data. The framework construction module 7 further includes a reward adjustment unit 74, which is used to calculate a sensor comprehensive abnormality value based on the actual video data, actual infrared data, actual audio data, and actual vibration data, and dynamically adjust the overall reward based on the sensor comprehensive abnormality value. The conversion relationship between the sensor comprehensive abnormality value and the overall reward is configured as follows:
[0081]
[0082] in, represents the sensor comprehensive abnormal value, Indicates the actual video data abnormal value, Indicates the actual infrared data abnormal value, Indicates the actual audio data abnormal value, Indicates abnormal values of actual vibration data, 、 、 、 Represent the mean of each sensor data, 、 、 、 Respectively represent the value range of each sensor data, Represents the preset basic overall reward, Represents the abnormal suppression coefficient, which is used to control the attenuation degree of abnormal values on the overall reward. Indicates the preset abnormal compensation coefficient, which is used to prevent the reward value from being excessively reduced. Used to provide smooth abnormal adjustment effects and prevent sudden changes.
[0083] Specifically, in this embodiment, the sensor comprehensive abnormal value is calculated The abnormal data of each sensor are normalized by using the standardization method and the average is taken to ensure that different physical quantities are calculated under the same dimension. If the abnormal value of a sensor deviates greatly from the average, the overall abnormal value Increase.
[0084] Adjusting overall rewards : Using exponential decay function ,When the anomaly increases, the overall reward is gradually reduced, ensuring that the system reduces incentives and enhances safety strategies when the anomaly increases.
[0085] Using the hyperbolic tangent function Ensure that the reward adjustment is relatively stable when the anomaly is small, and avoid the system being overly sensitive to small anomalies. and It can be adjusted according to actual needs. For example, in a security system, if the cost of false alarms is high, a larger To improve the ability to suppress abnormalities.
[0086] Range analysis: Normal environment (no abnormalities): ≈0, then Note: Under normal circumstances, system rewards remain at the default level.
[0087] Abnormal increase (sensor data deviates significantly from the mean): Increase, exponential term Rapid decay causes the overall reward to drop, indicating that the system has entered an abnormal state.
[0088] Extreme anomaly (high system anomaly): →∞, then This means that even when anomalies are significant, the system's rewards are still capped to prevent excessive penalties from impacting policy optimization. This formula is applicable to scenarios such as intelligent security systems, intelligent control systems, and multi-sensor data fusion, enabling the system to dynamically adjust rewards based on anomalies, ensuring maximum rewards in normal environments and appropriate reductions in abnormal ones, thereby enhancing system security and robustness.
[0089] Preferably, the hierarchical reinforcement learning framework includes edge computing devices and cloud servers, wherein the edge computing devices are used to execute local control tasks in real time, and the cloud servers are used to continuously update the defense strategies.
[0090] Specifically, in this embodiment, the system adopts an edge-cloud collaborative computing architecture, which is specifically composed as follows: Edge computing devices: Device types: smart cameras, edge servers (NVIDIA Jetson, Intel Movidius), embedded AI chips, etc.
[0091] Main tasks: Execute local control tasks in real time, including camera view adjustment, alarm triggering, and lighting control. Local data processing reduces transmission latency and avoids cloud computing bottlenecks. Preliminary reinforcement learning inference utilizes local models to quickly calculate the optimal local control strategy.
[0092] Cloud Server: Device type: High-performance GPU servers (such as AWS EC2 and Google Cloud TPU), AI computing clusters.
[0093] Main tasks: Optimize global defense strategies by training reinforcement learning models based on extensive historical data and generating global optimization strategies. Continuously update the models and distribute the newly learned optimization strategies to edge computing devices, enhancing intelligence. Large-scale data analysis processes massive amounts of video, infrared, audio, and vibration sensor data to provide global threat assessment capabilities.
[0094] Workflow: When the system is running, the edge computing device and the cloud server work together with the following division of labor: (1) Edge computing devices perform local control tasks in real time When an abnormal event is detected in the target area, the edge computing device first analyzes the current environmental data and uses the deployed reinforcement learning model to execute the optimal control strategy.
[0095] For example, cameras adjust to optimal monitoring angles to improve tracking of suspicious targets. Alarms adjust alarm volume or trigger emergency notifications based on the real-time threat level. Lighting systems dynamically adjust brightness based on the location of an intrusion to provide optimal monitoring conditions. Because computing is performed locally on the edge, the entire process typically responds in milliseconds, ensuring high real-time performance.
[0096] (2) Continuous optimization of cloud server defense strategies The cloud server collects data from multiple edge devices, analyzes the effectiveness of long-term defense strategies, and continuously optimizes the global strategy using reinforcement learning algorithms. The specific optimization process includes: Data synchronization: Edge computing devices regularly upload anomaly detection data, local execution results, and sensor information.
[0097] Model training: The cloud server uses historical data to train a global reinforcement learning model to optimize the long-term effectiveness of the strategy.
[0098] Strategy distribution: After training is completed, the cloud server sends the latest optimized defense strategy back to the edge device to update its control logic.
[0099] To improve the efficiency and stability of the system, this system adopts the following optimization strategies in the edge-cloud computing architecture: (1) Edge computing task optimization Lightweight AI model: Using technologies such as model pruning and quantization, reinforcement learning inference calculations can run efficiently on low-power devices (such as embedded AI chips).
[0100] Asynchronous reasoning mechanism: For urgent tasks (such as intrusion detection), the system prioritizes local decision-making to avoid delays caused by waiting for cloud computing.
[0101] (2) Cloud optimization strategy Federated Learning: Avoid uploading all data to the cloud. The cloud server only synchronizes the model weights learned by each edge device to improve privacy and security.
[0102] Adaptive training scheduling: The cloud server dynamically adjusts the frequency of policy updates based on the task load of edge devices to ensure that device computing resources are not over-occupied.
[0103] This embodiment effectively improves the real-time, stability, and adaptability of the intelligent security system through the layered reinforcement learning architecture of edge computing devices + cloud servers, ensuring that the system can maintain efficient and intelligent security control capabilities in various complex environments.
[0104] An intelligent security control method based on a simulation environment is applied to the above-mentioned intelligent security control system based on a simulation environment, referring to Figure 2 ,include: Step S1, the random generation module 1 constructs a simulation environment of the intelligent security system and randomly generates a plurality of variable parameters in the simulation environment, wherein the variable parameters include simulation environment parameters and sensor abnormality parameters; Step S2: the virtual sensing module 2 models multiple types of sensors to obtain multiple virtual sensing models, the virtual sensing models are used to output simulated sensing data, and the simulated sensing data are fused to obtain simulated fused data; In step S3, the collaborative simulation module 3 constructs an intrusion agent, a defense agent, and a control agent in the simulation environment according to the pre-introduced multi-agent reinforcement learning framework. The intrusion agent is used to simulate and generate a variety of simulated intrusion behaviors and simulated response strategies. The control agent is used to generate a defense strategy based on the simulation fusion data when facing the simulated intrusion behaviors and simulated response strategies of the intrusion agent. The defense agent is used to adjust the camera angle, alarm threshold, and lighting status according to the defense strategy, simulation environment parameters, and sensor anomaly parameters. Step S4, the actual monitoring module 4 obtains actual environmental parameters, actual abnormal parameters, actual intrusion behavior, and actual response strategies; Step S5: The parameter comparison module 5 compares the actual environmental parameters, the actual abnormal parameters, the actual intrusion behavior, and the actual strain strategy with the simulated environmental parameters, the sensed abnormal parameters, the simulated intrusion behavior, and the simulated strain strategy to obtain a plurality of learning parameters, the learning parameters including environmental deviation data, abnormal deviation data, intrusion deviation data, and strain deviation data. In step S6, the optimization module 6 performs dynamic correction optimization on the intrusion agent, the defense agent, and the control agent according to the learning parameters.
[0105] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that improvements and modifications that do not depart from the principles of the present invention are within the scope of protection of the present invention.
Claims
1. An intelligent security control system based on a simulation environment, characterized in that: include: A random generation module (1) is used to construct a simulation environment for an intelligent security system and randomly generate a plurality of variable parameters in the simulation environment, wherein the variable parameters include simulation environment parameters and sensor abnormality parameters; A virtual sensing module (2) is used to model multiple types of sensors to obtain multiple virtual sensing models, wherein the virtual sensing models are used to output simulated sensing data, and to fuse the simulated sensing data to obtain simulated fusion data; A collaborative simulation module (3) is connected to the random generation module (1) and the virtual sensing module (2), and is used to construct an intrusion agent, a defense agent, and a control agent in the simulation environment according to a pre-introduced multi-agent reinforcement learning framework, wherein the intrusion agent is used to simulate and generate a variety of simulated intrusion behaviors and simulated response strategies, the control agent is used to generate a defense strategy according to the simulation fusion data when facing each of the simulated intrusion behaviors and the simulated response strategies of the intrusion agent, and the defense agent is used to adjust the camera angle, alarm threshold, and light status according to the defense strategy, the simulation environment parameters, and the sensor abnormality parameters; The actual monitoring module (4) is used to obtain actual environmental parameters, actual abnormal parameters, actual intrusion behavior, actual sensor fusion data, and actual response strategies; a parameter comparison module (5), connected to the collaborative simulation module (3) and the actual monitoring module (4), for respectively comparing the actual environmental parameters, the actual abnormal parameters, the actual intrusion behavior, the actual sensor fusion data, and the actual strain strategy with the simulated environmental parameters, the sensor abnormal parameters, the simulated intrusion behavior, the simulated sensor data, and the simulated strain strategy to obtain a plurality of learning parameters, wherein the learning parameters include environmental deviation data, abnormal deviation data, intrusion deviation data, sensor fusion deviation data, and strain deviation data; An optimization module (6) is connected to the parameter comparison module (5) and the collaborative simulation module (3) and is used to perform dynamic correction optimization on the intrusion agent, the defense agent and the control agent according to each of the learning parameters.
2. The intelligent security control system based on a simulation environment according to claim 1, characterized in that: The random generation module (1) comprises: An environment generation unit (11) is used to randomly match and generate a plurality of simulation environment parameters from a preset environment database according to the location data and time data of the location of the intelligent security control system based on the simulation environment, wherein the simulation environment parameters include simulation temperature, simulation humidity, simulation air quality, simulation lighting parameters, simulation electromagnetic interference data, and simulation biological interference data; An abnormality introduction unit (12) is used to introduce the sensor abnormality parameters under the abnormal state of the sensor in the simulation environment, wherein the sensor abnormality parameters include simulated sensor noise data, simulated signal interruption data, simulated sensor conflict data, and simulated link abnormality data.
3. The intelligent security control system based on a simulation environment according to claim 1, characterized in that: The virtual sensing module (2) comprises: A sensor modeling unit (21) is used to construct a plurality of virtual sensor models, wherein the virtual sensor models are used to simulate the response curves, delays, noises and errors of various sensors and output the simulated sensor data, wherein the simulated sensor data includes simulated video data, simulated infrared data, simulated audio data and simulated vibration data; A preliminary fusion unit (22) is connected to the sensor modeling unit (21) and is used to construct a multi-sensor data fusion algorithm in the simulation environment to sequentially perform spatiotemporal alignment, information extraction and data fusion on each of the simulation sensor data from different sources to obtain simulation fusion data.
4. The intelligent security control system based on a simulation environment according to claim 3, characterized in that: The data fusion calculation formula of the multi-sensor data fusion algorithm is configured as: ; ; ; Function, used to control the scale change of data, Represents the Riemann Zeta function, which is used to enhance the nonlinear mapping of data. represents the first kind of Bessel function, which is used to control the smoothness of the data. Represents a complex information filtering function, which is used to optimize data fusion by combining error function, logarithmic function and elliptic integral. Represents different types of sensor data, where To simulate video data, To simulate infrared data, To simulate audio data, To simulate vibration data, Represents the weight of various sensor data, which is used to distribute according to the confidence and contribution of the data. Represents the index of sensor data in the spatial domain, is the filter input variable, is the integration variable, is the first kind of elliptic integral, which is used to correct the nonlinear error of the data.
5. The intelligent security control system based on simulation environment according to claim 1, characterized in that: The system further comprises a framework construction module (7) connected to the collaborative simulation module (3), wherein the framework construction module (7) is used to construct a hierarchical reinforcement learning framework and comprises: An outer global strategy unit (71) is used to generate security strategy sub-goals according to hierarchical deep reinforcement learning, and transmit the security strategy sub-goals to the control agent, and the control agent generates the defense strategy based on the security strategy sub-goals; The inner local policy unit (72) is used to perform fine-grained adjustments on the camera, alarm device and light, and train the adjustment method based on the policy gradient method to optimize the local control task of the defense agent on the camera angle, the alarm threshold and the light state.
6. The intelligent security control system based on a simulation environment according to claim 5, characterized in that: The hierarchical reinforcement learning framework adopts a hierarchical reward structure, which includes an overall reward, an outer reward, and an inner reward. The conversion relationship between the overall reward, the outer reward, and the inner reward is configured as follows: ; in, represents the overall reward, represents the outer reward, represents the inner layer reward, represents the outer layer weight, represents the inner layer weight, Indicates the total amount of inner layer rewards.
7. The intelligent security control system based on simulation environment according to claim 6, characterized in that: The actual environmental parameters include actual temperature, actual humidity, actual air quality, actual illumination parameters, actual electromagnetic interference data, and actual biological interference data. The framework construction module (7) further includes an adaptive weight adjustment unit (73), which is used to establish a conversion relationship with the outer layer weight and the inner layer weight according to the actual environmental parameters, and dynamically adjust the outer layer weight and the inner layer weight according to the actual environmental parameters, wherein the conversion relationship between the actual environmental parameters and the outer layer weight and the inner layer weight is configured as follows: ; in, represents the actual temperature, represents the actual humidity, represents the actual air quality, represents the actual lighting parameters, represents the actual electromagnetic interference data, represents the actual biological interference data, represents the average value of the actual temperature, represents the average value of the actual humidity, represents the average value of the actual air quality, represents the average value of the actual lighting parameter, represents the average value of the actual electromagnetic interference data, represents the average value of the actual biological interference data, 、 、 、 、 、 They respectively represent the value ranges of the actual temperature, the actual humidity, the actual air quality, the actual lighting parameters, the actual electromagnetic interference data, and the actual biological interference data.
8. The intelligent security control system based on a simulation environment according to claim 6, characterized in that: The actual abnormality parameters include actual video data, actual infrared data, actual audio data, and actual vibration data. The framework construction module (7) further includes a reward adjustment unit (74). The reward adjustment unit (74) is used to calculate a sensory integrated abnormality value based on the actual video data, actual infrared data, actual audio data, and actual vibration data, and dynamically adjust the overall reward based on the sensory integrated abnormality value. The conversion relationship between the sensory integrated abnormality value and the overall reward is configured as follows: ; ; in, represents the sensor comprehensive abnormal value, represents the actual video data abnormal value, represents the abnormal value of the actual infrared data, represents the actual audio data abnormal value, represents the abnormal value of the actual vibration data, 、 、 、 Represent the mean of each sensor data, 、 、 、 Respectively represent the value range of each sensor data, Represents the preset basic overall reward, Represents the abnormality suppression coefficient, which is used to control the degree of attenuation of the overall reward by abnormal values. Indicates the preset abnormal compensation coefficient, which is used to prevent the reward value from being excessively reduced. Used to provide smooth abnormal adjustment effects and prevent sudden changes.
9. The intelligent security control system based on simulation environment according to claim 5, characterized in that: The hierarchical reinforcement learning framework includes an edge computing device and a cloud server. The edge computing device is used to execute the local control task in real time, and the cloud server is used to continuously update the defense strategy.
10. An intelligent security control method based on a simulation environment, applied to the intelligent security control system based on a simulation environment according to any one of claims 1 to 9, characterized in that: include: Step S1, a random generation module (1) constructs a simulation environment of an intelligent security system and randomly generates a plurality of change parameters in the simulation environment, wherein the change parameters include simulation environment parameters and sensor abnormality parameters; Step S2, the virtual sensing module (2) models multiple types of sensors to obtain multiple virtual sensing models, wherein the virtual sensing models are used to output simulated sensing data, and the simulated sensing data are fused to obtain simulated fusion data; Step S3, the collaborative simulation module (3) constructs an intrusion agent, a defense agent, and a control agent in the simulation environment according to a pre-introduced multi-agent reinforcement learning framework, wherein the intrusion agent is used to simulate and generate a variety of simulated intrusion behaviors and simulated response strategies, the control agent is used to generate a defense strategy according to the simulation fusion data when facing each of the simulated intrusion behaviors and the simulated response strategies of the intrusion agent, and the defense agent is used to adjust the camera angle, alarm threshold, and light status according to the defense strategy, the simulation environment parameters, and the sensor abnormality parameters; Step S4, the actual monitoring module (4) obtains actual environmental parameters, actual abnormal parameters, actual intrusion behavior, and actual response strategies; Step S5, the parameter comparison module (5) compares the actual environment parameters, the actual abnormal parameters, the actual intrusion behavior and the actual strain strategy with the simulated environment parameters, the sensor abnormal parameters, the simulated intrusion behavior and the simulated strain strategy to obtain a plurality of learning parameters, wherein the learning parameters include environment deviation data, abnormality deviation data, intrusion deviation data and strain deviation data; In step S6, the optimization module (6) performs dynamic correction optimization on the intrusion agent, the defense agent and the control agent according to each of the learning parameters.
Citation Information
Patent Citations
Optical fiber sensing based borderline boundary anti-intrusion alarming system and method
CN108765814A
Security monitoring application test system
CN109410685A
Five-dimensional early-warning system based on artificial intelligence multi-factor panoramic monitoring detection
CN109640032A
Method and system for security warning of smart park
CN116935555A
Self-adaptive adjusting production method, device and system for safety shoes and safety shoes
CN118144225A
Cited By
Intelligent confrontation element deduction system based on dynamic environment and cognitive game
CN121615790A
An intelligent adversarial meta-inference system based on dynamic environment and cognitive game
CN121615790B
