A blockchain-based multi-agent collaborative risk prevention and control system

Abstract

This disclosure relates to the field of security risk prevention technology and proposes a blockchain-based multi-agent collaborative risk prevention and control system. The system includes: a risk map agent, used to construct a digital twin spatial model of the target environment and maintain a risk control blockchain for recording risk event reporting information, risk event voting information, risk response decision information, and risk response execution information; a risk screening agent, used to listen to various accessible real-time data stream points in the digital twin spatial model, and when a potential risk event is inferred at a real-time data stream point, report the potential risk event and corresponding decision suggestions to the risk control blockchain; and a personal guide agent, used to interact with the risk map agent and provide an operation entry point for the target user. The technical solutions provided by one or more embodiments of this disclosure can deeply integrate diverse information of the target environment, achieving comprehensive perception and timely early warning of both human and object risks.

Description

Technical Field

[0001] This disclosure relates to the field of security risk prevention technology, specifically to a blockchain-based multi-agent collaborative risk prevention and control system. Background Technology

[0002] As a semi-enclosed space with a high density of people and diverse activities, the safety management of schools is not only crucial to the physical and mental health of students but also directly affects the stability of teaching order and the improvement of educational quality. With the expansion of school size, changes in the social environment, and increasingly diverse student behaviors, campus safety risks are characterized by their complexity, high degree of concealment, and dynamic evolution. On the one hand, "unsafe human behaviors" at the level of teachers and students, such as violations, psychological abnormalities, and potential conflicts, are highly concealed and evolve rapidly, easily overlooked in their early stages. On the other hand, "unsafe physical conditions" at the level of physical facilities and the environment, such as aging equipment, blocked passageways, and sudden environmental changes, can also trigger safety accidents.

[0003] Traditional campus security management largely relies on passive response methods such as video surveillance, manual patrols, and access control systems, making it difficult to achieve comprehensive perception and early intervention of risks to both people and objects. With the advancement of educational informatization, some schools have begun to introduce digital means such as teaching management systems, social media platform data, and campus cards, attempting to build a risk identification system that integrates multi-source data.

[0004] However, due to the common barriers in data collection, storage, and interface aspects, existing campus security management systems still primarily rely on single data sources in practice, lacking in-depth integration of multi-dimensional information. Furthermore, risk assessment in existing campus security management systems often depends on the human experience of administrators, resulting in strong subjectivity and a high rate of misjudgment. Therefore, existing campus security management systems struggle to provide systematic and real-time early warnings for complex, dynamic, and multi-dimensional campus risks. Summary of the Invention

[0005] In view of this, one or more embodiments of this disclosure provide a blockchain-based multi-agent collaborative risk prevention and control system that can deeply integrate diverse information of the target environment to achieve comprehensive perception and timely early warning of both human and object risks.

[0006] This disclosure provides a blockchain-based multi-agent collaborative risk prevention and control system. The system includes a risk map agent, a risk screening agent, and a personal guide agent. The risk map agent constructs a digital twin space model of the target environment and maintains a risk control blockchain, which records risk event reporting information, risk event voting information, risk response decision information, and risk response execution information. The risk screening agent listens to various accessible real-time data stream points in the digital twin space model and, upon inferring the existence of potential risk events at these points, reports the potential risk events and corresponding decision suggestions to the risk control blockchain. The personal guide agent interacts with the risk map agent and provides an operation entry point for the target user, enabling the target user to report risk events and participate in voting within the risk control blockchain.

[0007] In one embodiment, constructing a digital twin spatial model of the target environment includes: acquiring environmental monitoring data of the target environment using environmental monitoring equipment; traversing the target environment using a mobile device to generate mobile observation data; and constructing the digital twin spatial model based on the environmental monitoring data and the mobile observation data.

[0008] In one implementation, maintaining the risk control blockchain includes at least one of the following: initializing the risk control blockchain; issuing risk control tokens to each real-time target user based on the risk control budget; broadcasting candidate risk events reported by the target user through the personal guide agent to the entire blockchain network; broadcasting potential risk events and corresponding decision suggestions reported by the risk screening agent to the entire blockchain network; broadcasting decision suggestions related to the candidate risk events or potential risk events reported by the target user through the personal guide agent to the entire blockchain network; determining the risk event voting information participated in by the target user based on the risk control tokens issued by the target user through the personal guide agent; summarizing the risk event voting information, determining the target risk event and the target response strategy, and distributing the risk task sheet containing the target risk event and the target response strategy to the task leader.

[0009] In one implementation, the maintenance of the risk control blockchain further includes: adding evidence to the risk events submitted by the target user through the personal guide agent and confirming the evidence; if the additional evidence of the risk event is confirmed, then issuing a token reward to the target user.

[0010] In one implementation, the step of listening to each accessible real-time data stream point in the digital twin space model includes: determining each of the real-time data stream points in the digital twin space model; assigning corresponding data analysis processes to each of the real-time data stream points to infer whether the potential risk event exists at each of the real-time data stream points; and if the potential risk event exists at the real-time data stream point, then planning decision suggestions for the potential risk event.

[0011] In one implementation, the risk assessment agent is also used to: periodically analyze the potential risk events, discover observable patterns, and generate readable risk reports.

[0012] In one embodiment, the system further includes a risk control exercise agent; the risk control exercise agent is used to generate a risk exercise plan based on preset exercise items and save the risk exercise plan to the database of the risk map agent.

[0013] In one implementation, generating a risk exercise plan based on a preset exercise project includes: obtaining risk point and emergency resource data by interacting with the risk map intelligence agent; determining the personnel division of labor, action instructions, evacuation routes, and guidance scheme of the risk exercise plan; reporting the risk exercise plan to the risk control blockchain and obtaining first feedback suggestions from each of the target users regarding the risk exercise plan; and optimizing the algorithm logic of the risk exercise plan and the risk control exercise intelligence agent based on the first feedback suggestions.

[0014] In one implementation, the risk assessment agent is further configured to: collect actual exercise data of the risk exercise plan; generate exercise optimization suggestions based on the actual exercise data; report the exercise optimization suggestions to the risk control blockchain and obtain second feedback suggestions from each of the target users regarding the exercise optimization suggestions; and optimize the exercise optimization suggestions and the algorithm logic of the risk assessment agent based on the second feedback suggestions.

[0015] In one implementation, the interaction with the risk map agent and the provision of operation entry points for the target user includes at least one of the following: providing the target user with an event reporting interface; providing the target user with a suggestion upload interface; providing the target user with an evidence upload interface; providing the target user with a voting interface; providing the target user with surrounding risk warnings based on the digital twin space model; and providing the target user with map navigation based on the digital twin space model.

[0016] This disclosure provides a technical solution through one or more embodiments, constructing a distributed collaborative architecture integrating "perception-reasoning-participation." The risk map intelligent agent, serving as the digital foundation, utilizes digital twin technology to integrate physical and information spaces, solving the challenges of spatial alignment and visualization management of multi-source data, and ensuring the immutability and traceability of information through blockchain technology. The risk screening intelligent agent, as the core of automated reasoning, can proactively listen to and integrate real-time data streams from multiple sources such as access control, monitoring, and social information, overcoming the limitations of traditional manual inspections and the high misjudgment rate caused by the subjective experience of managers, achieving intelligent identification and early intervention of potential risks. The personal guide intelligent agent breaks down the barriers between the management end and the teachers and students end, delegating risk reporting and voting decision-making power to each user, enabling timely perception of behavioral risks (such as psychological abnormalities and interpersonal conflicts) that are highly concealed and difficult to detect by sensors.

[0017] This disclosure provides a technical solution with one or more embodiments, in which three intelligent agents have clearly defined roles and work closely together. The risk screening intelligent agent is mainly used to quantitatively analyze the state of objects and the collective behavior of people; the personal guide intelligent agent is mainly used to effectively perceive individual behavior and support democratic decision-making; and the risk map intelligent agent provides a unified spatiotemporal benchmark and trusted evidence storage. The three agents achieve information exchange and action synchronization through the risk control blockchain, forming a closed loop of "perception-judgment-decision-execution-feedback". This design breaks through the limitations of traditional risk prevention and control systems in terms of passive response, reliance on single-source data, and reliance on experience-based judgment, and promotes the evolution of safety management and risk prevention and control from a "human-to-human" mode to a "human-machine collaboration and all-person co-governance" mode, significantly improving the real-time performance, accuracy, and systematic nature of risk prevention and control. Attached Figure Description

[0018] The features and advantages of the embodiments of this disclosure will be more clearly understood by referring to the accompanying drawings, which are illustrative and should not be construed as limiting the present disclosure in any way. In the drawings: Figure 1 This diagram illustrates the structure of a blockchain-based multi-agent collaborative risk prevention and control system according to one embodiment of the present disclosure. Figure 2 A schematic diagram of another blockchain-based multi-agent collaborative risk prevention and control system is shown in one embodiment of this disclosure. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0020] Please see Figure 1 The present disclosure provides a blockchain-based multi-agent collaborative risk prevention and control system, which may include a risk map agent, a risk screening agent, and a personal guide agent.

[0021] In this embodiment, the risk map agent is used to construct a digital twin spatial model of the target environment and maintain a risk control blockchain. The risk control blockchain is used to record risk event reporting information, risk event voting information, risk response decision information, and risk response execution information.

[0022] Specifically, a digital twin spatial model is not simply a 3D map, but a virtual space that is mapped and dynamically updated in real time to the real physical space. The digital twin spatial model can integrate not only static spatial data such as BIM (Building Information Modeling) of all buildings within the target environment (e.g., campuses, exhibition centers, industrial parks, transportation hubs), indoor and outdoor navigation networks, camera locations, access control card reader locations, fire protection facility distribution, and hazardous area markings, but also dynamic spatial data such as real-time data streams from IoT sensors, including crowd heat maps, equipment operating status, and environmental parameters (temperature, humidity, smoke). This allows managers / users to view the safety status of the entire target environment on a unified visual interface, identifying areas showing signs of congestion, abnormal equipment operation, and areas with excessive temperatures.

[0023] Risk control blockchain is used to record risk event reporting information (e.g., who, when, where, and what type of risk was reported), risk event voting information (e.g., consensus votes by other users or agents on the authenticity of the event), risk response decision information (e.g., handling plans formulated for confirmed risks), and risk response execution information (e.g., who is executing, to what stage, and the effect). The immutability of blockchain ensures that these records cannot be unilaterally modified afterward, thus establishing a traceable and auditable trust mechanism. The decentralized nature of blockchain means that no single node (agent or user) can unilaterally control the entire risk control process, avoiding the risk of internal data tampering in traditional systems.

[0024] Furthermore, the risk map agent can also serve as a spatial index. When a risk event occurs, it can quickly locate the corresponding coordinates in the digital twin model and retrieve all sensor data, historical event records, and recent personnel distribution in the vicinity of that location, providing geospatial context for the reasoning of other agents.

[0025] In this embodiment, the risk screening agent is used to listen to each accessible real-time data stream point in the digital twin space model, and when it infers that there is a potential risk event at the real-time data stream point, it reports the potential risk event and the corresponding decision suggestion to the risk control blockchain.

[0026] Specifically, the design philosophy of the risk assessment intelligent agent is to overcome the fundamental shortcomings of traditional safety management systems, namely the limitations of human visual inspection and the subjective nature of experience-based judgment. The risk assessment intelligent agent can transform safety management from "post-event response and random inspection" to "real-time detection and proactive early warning," significantly improving the timeliness and coverage of risk discovery.

[0027] The risk assessment agent can continuously monitor all accessible real-time data stream points within the digital twin spatial model. These points can include frame-by-frame images from video surveillance systems, card swipe records, personnel movement trajectories captured by Wi-Fi probes, abnormal fluctuations in smart water and electricity meters, keyword-based public opinion monitoring on social media (such as school forums and class groups), and even anonymous sentiment indicators from psychological counseling appointment systems. Unlike traditional security management systems that use data sources in isolation, the risk assessment agent inherently possesses cross-data source correlation analysis capabilities. For example, when video surveillance analysis detects a surge in pedestrian density at a stairwell, while card swipe data shows an abnormal decrease in the card swipe rate in the same area during the same period, combined with complaints about "stairwell congestion" on social media, the assessment agent can infer with high confidence that "there is a risk of blockage in evacuation routes" and provide a decision recommendation of "immediately opening backup routes and dispatching security personnel to guide traffic."

[0028] Risk assessment agents can embed lightweight machine learning models (such as time-series anomaly detection, behavior recognition, graph neural network relationship reasoning, etc.). These machine learning models can be trained offline using historical event data (including past security incidents and daily management records) and can be continuously optimized online.

[0029] Optionally, when a potential risk event is identified, the risk assessment agent will not directly trigger an alarm (which could lead to false alarms and panic). Instead, it will package the event description, evidence fragments (such as screenshots and data curves), and corresponding decision recommendations together, and record it as a risk event to be verified by calling the reporting interface of the risk control blockchain. The agent is responsible for initial screening and evidence organization, but a final confirmation step by a human (through voting by a personal guide agent) is still retained, thus balancing automation efficiency and decision security. In addition, the risk assessment agent can also reverse query the historical events stored on the risk control blockchain for self-learning and model calibration, avoiding repeated errors on the same type of risk.

[0030] In this embodiment, the personal guide agent is used to interact with the risk map agent and provide an operation entry point for the target user, so that the target user can report risk events and participate in voting in the risk control blockchain.

[0031] Specifically, the smart body that guides you is a bridge connecting the system with every end user. It "compresses" professional-grade risk control capabilities into the mobile phones, ID cards or wearable devices that target users (such as teachers and students on campus, employees in industrial parks, and participants in exhibitions) carry every day, thus achieving full reach and participation in security management.

[0032] The portable intelligent agent provides users with two core operational entry points: risk reporting and voting. The risk reporting function allows each target user to become a special sensor unit. When a target user discovers a missing manhole cover, broken corridor lights, abnormal emotions in others, or signs of fighting (these are soft risks that traditional sensors struggle to cover or that the risk assessment intelligent agent cannot directly capture), they can simply activate the photo, voice description, or select standardized tags to automatically synchronize the event to the risk control blockchain.

[0033] To suppress malicious or false reporting, for a newly reported risk event, other users' personal AI assistant will randomly or based on location relevance push requests to "please assist in determining whether this risk is real." Other users can vote anonymously (e.g., "exists," "does not exist," "uncertain," etc.). Only when the voting results reach a preset threshold (e.g., more than 5 people and a majority believe it exists) will the risk event be marked as "verified" and formally enter the decision-making and execution process. This democratized verification mechanism avoids subjective misjudgment by a single administrator and leverages the collective wisdom of the community. In addition, the personal AI assistant will also push the final decision information on the risk control blockchain (e.g., a laboratory will conduct ventilation maintenance in 30 minutes; please evacuate nearby personnel) to all users in the affected area in real time, completing a closed loop from perception to response. Through the personal AI assistant, the system successfully transforms every campus member from a passively managed "safety object" into an actively participating "safety subject," significantly enhancing the intrinsic motivation of safety culture.

[0034] Optionally, by interacting with the risk map agent, the personal guide agent can provide the current user with a personalized safety interface. For example, when a user enters a laboratory area, the system can automatically push the laboratory's emergency escape routes and hazardous chemical handling procedures; when a user is walking in a remote area at night, the system can display the location of the nearest emergency alarm post and a "one-click protection" button. In this way, on the one hand, the barrier for users to actively learn safety knowledge is lowered, and on the other hand, safety guidance can be cleverly and flexibly integrated into users' daily behavior.

[0035] In some implementations, constructing a digital twin spatial model of the target environment includes: acquiring environmental monitoring data of the target environment using environmental monitoring equipment; traversing the target environment using a mobile device to generate mobile observation data; and constructing the digital twin spatial model based on the environmental monitoring data and the mobile observation data.

[0036] Specifically, various fixed environmental monitoring devices deployed in the target environment (such as cameras, temperature and humidity sensors, smoke detectors, air quality monitors, and noise meters) can be used to continuously collect environmental monitoring data, forming a comprehensive perception of the basic parameters of the physical space. Mobile devices (such as inspection robots, drones, or handheld scanning terminals) can be used for traversal mobile observation of the target environment. During movement, using sensors such as lidar, depth cameras, and inertial navigation, mobile devices can generate mobile observation data containing details of spatial geometry, obstacle distribution, facility status, and dynamic changes, effectively compensating for the monitoring blind spots and limited perspectives of fixed equipment. By aligning and fusing environmental monitoring data with mobile observation data in time and space, a high-precision, real-time-updable digital twin spatial model can be constructed. This digital twin spatial model not only possesses visualized and measurable geometric and physical attributes but also maps the spatiotemporal distribution patterns of various risk factors, providing a unified and complete digital foundation for the subsequent proactive listening and reasoning of risk-screening agents.

[0037] In some implementations, maintaining the risk control blockchain includes at least one of the following: initializing the risk control blockchain; issuing risk control tokens to each real-time target user based on the risk control budget; broadcasting candidate risk events reported by the target user through the personal guide agent to the entire blockchain network; broadcasting potential risk events and corresponding decision suggestions reported by the risk screening agent to the entire blockchain network; broadcasting decision suggestions related to the candidate or potential risk events reported by the target user through the personal guide agent to the entire blockchain network; determining the risk event voting information participated in by the target user based on the risk control tokens issued by the target user through the personal guide agent; summarizing the risk event voting information, determining the target risk event and target response strategy, and distributing a risk task sheet containing the target risk event and target response strategy to the task leader.

[0038] Specifically, based on a preset risk control budget, a certain number of risk control tokens can be issued to each target user (such as teachers, students, and employees) during the initialization of the risk control blockchain. These tokens serve as voting rights for participating in risk verification and decision-making. The risk control blockchain can broadcast all three types of key information (candidate risk events proactively reported by users through the personal guide agent, potential risk events automatically detected and reported by the risk screening agent along with their corresponding decision suggestions, and supplementary decision suggestions submitted by other users for the above two types of events). When a vote on the authenticity or priority of a risk event / decision suggestion is required, users can invest their held risk control tokens through the personal guide agent. The voting weight can be proportional to the number of tokens invested. After aggregating all voting information, the target risk event (i.e., risks recognized as real or urgent) and the corresponding target response strategy (i.e., response strategies collectively approved) can be automatically determined. Finally, a risk task sheet containing details of the risk event, response strategy, responsible party, and completion deadline can be distributed to designated task leaders (such as security personnel or maintenance teams) through the blockchain's smart contract. Once a risk task order is distributed, its execution status can be recorded on the blockchain throughout the entire process, thereby achieving fully traceable and tamper-proof automated governance from risk discovery to closed-loop handling.

[0039] In some implementations, maintaining the risk control blockchain further includes: adding evidence to the risk events submitted by the target user through the personal guide agent and confirming the evidence; if the additional evidence of the risk event is confirmed, then issuing a token reward to the target user.

[0040] Specifically, this supplementary mechanism aims to encourage users to continuously participate in the dynamic improvement of risk events. Once a candidate or potential risk event is on-chain, any target user can submit new evidence related to the event (such as supplementary photos, videos, text descriptions, or screenshots of environmental data) through the accompanying intelligent agent. The risk map intelligent agent will verify these supplementary evidence to determine their authenticity and validity. Once the evidence is verified, indicating that the information has practical value for risk identification or response, the risk map intelligent agent can issue a certain number of risk control tokens as a reward to the submitter. This not only enriches the contextual information of risk events and improves the accuracy of decision-making, but also builds a positive feedback loop through token incentives, driving users to shift from passive observation to active participation, further strengthening the coverage depth of collective perception and the credibility of risk data.

[0041] In some implementations, the step of listening to each accessible real-time data stream point in the digital twin space model includes: determining each of the real-time data stream points in the digital twin space model; assigning a corresponding data analysis process to each of the real-time data stream points to infer whether the potential risk event exists at each of the real-time data stream points; and if the potential risk event exists at the real-time data stream point, then planning decision suggestions for the potential risk event.

[0042] Specifically, after identifying all accessible real-time data stream points (e.g., the specific locations and data interfaces of cameras, access control card readers, temperature and humidity sensors, Wi-Fi probes, etc.) in the digital twin space model, a corresponding data analysis process can be independently allocated or dynamically scheduled for each point. These data analysis processes can run lightweight models or rule engines to continuously perform anomaly detection and pattern recognition on the real-time data output by that point to infer whether there are potential risk events (e.g., exceeding the threshold for crowd density, sudden rise in equipment temperature, abnormal card swiping frequency, etc.). Once a point is determined to be at risk, the corresponding data analysis process immediately and automatically plans preliminary decision suggestions based on the risk type, level, and surrounding environment (e.g., notifying nearby security personnel to check, or triggering an evacuation announcement broadcast in the area). Through this distributed monitoring architecture of "one point, one process, real-time inference, and immediate suggestions," massive data streams can be processed in parallel, enabling rapid capture and initial response to potential risks.

[0043] In some implementations, the risk assessment agent is also used to: periodically analyze the potential risk events, discover observable patterns, and generate readable risk reports.

[0044] The risk assessment intelligence agent can systematically analyze all detected and reported potential risk events within a set period (e.g., daily, weekly, or monthly), including multi-dimensional indicators such as event type distribution, spatiotemporal hotspots, high-frequency triggering periods, and the frequency of associated device anomalies. Based on this, the agent uses data mining methods (e.g., cluster analysis, time-series trend detection, and association rule discovery) to uncover insightful patterns from the statistical results. For example, the probability of excessive crowd density in the third-floor corridor of the teaching building increases significantly every Friday afternoon, and there is a strong correlation between nighttime illegal electricity use in the dormitory area and late-night return records. These patterns are transformed into easily understandable, illustrated risk briefings that can be automatically pushed to security management personnel and relevant users. These briefings not only help the management team grasp the risk evolution of the target scenario from a macro perspective, providing data support for resource allocation and system optimization, but also enhance the risk awareness of all employees.

[0045] Please see Figure 2In some implementations, the blockchain-based multi-agent collaborative risk prevention and control system further includes a risk control exercise agent. This risk control exercise agent is used to generate risk exercise plans based on preset exercise projects and save the risk exercise plans to the database of the risk map agent.

[0046] Specifically, the newly added risk control exercise intelligence agent can automatically generate risk exercise plans based on preset exercise projects (such as fire evacuation, earthquake avoidance, and response to sudden violent incidents), and save these plans to the risk map intelligence agent's database. This saving behavior is not only for the exercise itself, but more importantly, it can immediately retrieve the plan that best matches the current scenario when a real risk event suddenly occurs, directly serving as an action blueprint for emergency response, thereby greatly reducing the decision-making time between the occurrence of an event and an effective response.

[0047] In some implementations, generating a risk exercise plan based on preset exercise projects includes: obtaining risk point and emergency resource data by interacting with the risk map intelligence agent, determining the personnel division of labor, action instructions, evacuation routes, and guidance scheme of the risk exercise plan; reporting the risk exercise plan to the risk control blockchain, and obtaining first feedback suggestions from each of the target users regarding the risk exercise plan; and optimizing the algorithm logic of the risk exercise plan and the risk control exercise intelligence agent based on the first feedback suggestions.

[0048] Specifically, through deep interaction with the risk map intelligent agent, the distribution of risk points within the target environment (such as congested passages and hazardous chemical storage areas) and emergency resource data (such as fire extinguisher locations, first aid stations, and emergency broadcast coverage areas) can be obtained. Based on this, the division of labor among personnel in the exercise (clarifying the responsibilities of each role), action instructions (specifying what actions to take at what time and how to operate), evacuation routes (statically planned optimal and backup paths), and guidance plans (configuring guide locations and setting temporary guidance signs) can be determined. Subsequently, the plan is reported to the risk control blockchain for broadcast across the entire network, and the first feedback suggestions submitted by all target users through the personal guide intelligent agent are collected (e.g., potential obstacles on a certain evacuation route in actual walking). Based on this feedback, the algorithm logic within both the risk exercise plan itself and the risk control exercise intelligent agent can be optimized simultaneously.

[0049] In some implementations, the risk assessment agent is further configured to: collect actual exercise data of the risk exercise plan; generate exercise optimization suggestions based on the actual exercise data; report the exercise optimization suggestions to the risk control blockchain and obtain second feedback suggestions from each of the target users regarding the exercise optimization suggestions; and optimize the exercise optimization suggestions and the algorithm logic of the risk assessment agent based on the second feedback suggestions.

[0050] Specifically, during the actual execution of the exercise, the risk assessment agent can collect all actual exercise data (such as the time it takes for people to pass through each location, the location of congestion, guidance response delays, and records of violations). By analyzing this data post-exercise, the agent can automatically generate exercise optimization suggestions (for example, suggesting increasing the number of guides in the second evacuation route to two or adjusting the timing of broadcast instructions). These optimization suggestions are also reported to the risk control blockchain, and second feedback suggestions are solicited from relevant users. After summarizing user feedback, not only can the optimization suggestions themselves be calibrated, but the algorithmic logic of the risk assessment agent itself will also be further optimized, enabling the agent to have stronger insight and a lower false alarm rate when assessing real risk events or analyzing exercise data in the future.

[0051] In some implementations, the interaction with the risk map agent and the provision of operation entry points for the target user include at least one of the following: providing the target user with an event reporting interface; providing the target user with a suggestion upload interface; providing the target user with an evidence upload interface; providing the target user with a voting interface; providing the target user with surrounding risk warnings based on the digital twin space model; and providing the target user with map navigation based on the digital twin space model.

[0052] Specifically, the portable intelligent agent, serving as the terminal window for user interaction, integrates multiple operation entry points, ensuring convenient user participation in various risk control processes. The portable intelligent agent's event reporting interface allows users to quickly describe observed risks. Its suggestion upload interface allows users to submit their ideas for handling existing risk events on the blockchain. Its evidence upload interface supports users adding supporting materials such as photos and videos. Its voting interface allows users to reach a consensus on the authenticity of risks or risk response strategies by investing tokens. By accessing a digital twin space model, the portable intelligent agent can push real-time warnings of surrounding risks (such as congestion ahead or malfunctioning nearby devices) and provide map navigation to guide users to avoid dangerous areas or quickly reach emergency resource points. These various functional interfaces of the portable intelligent agent lower the barrier to user participation, enabling each user to become a special, intelligent "mobile sensing node," greatly enriching the breadth and timeliness of risk identification.

[0053] This disclosure provides a technical solution through one or more embodiments, constructing a distributed collaborative architecture integrating "perception-reasoning-participation." The risk map intelligent agent, serving as the digital foundation, utilizes digital twin technology to integrate physical and information spaces, solving the challenges of spatial alignment and visualization management of multi-source data, and ensuring the immutability and traceability of information through blockchain technology. The risk screening intelligent agent, as the core of automated reasoning, can proactively listen to and integrate real-time data streams from multiple sources such as access control, monitoring, and social information, overcoming the limitations of traditional manual inspections and the high misjudgment rate caused by the subjective experience of managers, achieving intelligent identification and early intervention of potential risks. The personal guide intelligent agent breaks down the barriers between the management end and the teachers and students end, delegating risk reporting and voting decision-making power to each user, enabling timely perception of behavioral risks (such as psychological abnormalities and interpersonal conflicts) that are highly concealed and difficult to detect by sensors.

[0054] This disclosure provides a technical solution with one or more embodiments, in which three intelligent agents have clearly defined roles and work closely together. The risk assessment intelligent agent is mainly used to quantitatively analyze the state of objects and the collective behavior of people; the personal guide intelligent agent is mainly used to effectively perceive individual behavior and support democratic decision-making; and the risk map intelligent agent provides a unified spatiotemporal benchmark and trusted evidence storage. The three agents achieve information exchange and action synchronization through a risk control blockchain, forming a closed loop of "perception-judgment-decision-execution-feedback". This design breaks through the limitations of traditional systems in passive response, reliance on single-source data, and reliance on experience-based judgment, promoting the evolution of security management from a "human-to-human" model to a "human-machine collaboration and all-person co-governance" model, significantly improving the real-time, accuracy, and systematic nature of campus security management.

[0055] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0056] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

[0057] Although embodiments of the present disclosure have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present disclosure, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A blockchain-based multi-agent collaborative risk prevention and control system, characterized in that, The system includes a risk map agent, a risk screening agent, and a personal guide agent; The risk map intelligent agent is used to construct a digital twin spatial model of the target environment and maintain the risk control blockchain. The risk control blockchain is used to record risk event reporting information, risk event voting information, risk response decision information, and risk response execution information. The risk screening intelligent agent is used to listen to each accessible real-time data stream point in the digital twin space model, and when it infers that there is a potential risk event at the real-time data stream point, it reports the potential risk event and the corresponding decision suggestions to the risk control blockchain. The personal guide agent is used to interact with the risk map agent and provide an operation entry point for the target user, so that the target user can report risk events and participate in voting in the risk control blockchain.

2. The system according to claim 1, characterized in that, The construction of the digital twin space model of the target environment includes: Environmental monitoring data of the target environment are obtained using environmental monitoring equipment; Using a mobile device, the target environment is traversed to generate mobile observation data; The digital twin spatial model is constructed based on the environmental monitoring data and the mobile observation data.

3. The system according to claim 1 or 2, characterized in that, The maintenance risk control blockchain includes at least one of the following: Initialize the risk control blockchain and issue risk control tokens for each real-time target user based on the risk control budget. For candidate risk events reported by the target user through the personal guide agent, broadcast them on the blockchain and across the entire network; The potential risk events and corresponding decision recommendations reported by the risk assessment agent are broadcast on the entire blockchain network. Decision suggestions related to the candidate risk event or the potential risk event reported by the target user through the personal guide agent are broadcast on the blockchain and across the entire network. Based on the risk control tokens issued by the target user through the personal guide agent, determine the risk event voting information in which the target user participates; The voting information on the risk events is summarized to determine the target risk events and target response strategies. Risk task sheets containing the target risk events and target response strategies are then distributed to the task leaders.

4. The system according to claim 3, characterized in that, The maintenance risk control blockchain also includes: Additional evidence is added to the risk events submitted by the target user through the personal guide agent to confirm the evidence; If additional evidence confirms the risk event, a token reward will be issued to the target user.

5. The system according to claim 1, characterized in that, The monitoring of various accessible real-time data stream points in the digital twin spatial model includes: In the digital twin spatial model, each of the real-time data stream points is determined; Assign corresponding data analysis processes to each of the real-time data stream points, and infer whether the potential risk events exist at each of the real-time data stream points; If the potential risk event exists at the real-time data stream location, then a planning decision suggestion is made for the potential risk event.

6. The system according to claim 5, characterized in that, The risk assessment intelligent agent is also used for: Periodically analyze the potential risk events, uncover discernible patterns, and generate readable risk reports.

7. The system according to claim 1, characterized in that, The system also includes a risk control exercise intelligent agent; The risk control exercise intelligent agent is used to generate risk exercise plans based on preset exercise projects and save the risk exercise plans to the database of the risk map intelligent agent.

8. The system according to claim 7, characterized in that, The step of generating a risk exercise plan based on preset exercise items includes: By interacting with the risk map agent, risk point and emergency resource data are obtained, and the personnel division, action instructions, evacuation routes and guidance schemes of the risk exercise plan are determined. The risk exercise plan is reported to the risk control blockchain, and the first feedback suggestions from each of the target users regarding the risk exercise plan are obtained. Based on the first feedback suggestion, the algorithm logic of the risk exercise plan and the risk control exercise agent is optimized.

9. The system according to claim 8, characterized in that, The risk assessment intelligent agent is also used for: Collect actual exercise data for the aforementioned risk exercise plan; Based on the actual exercise data, exercise optimization suggestions are generated; The exercise optimization suggestions are reported to the risk control blockchain, and second feedback suggestions from each of the target users are obtained regarding the exercise optimization suggestions; Based on the second feedback suggestion, the exercise optimization suggestion and the algorithm logic of the risk screening agent are optimized.

10. The system according to claim 1, characterized in that, The interaction with the risk map agent and the provision of an operation entry point for the target user include at least one of the following: Provide an event reporting interface for the target user; Provide a suggested upload interface for the target user; Provide an evidence upload interface for the target user; Provide a voting interface for the target users; Provide the target user with surrounding risk warnings based on the digital twin space model; Provide the target user with map navigation based on the digital twin spatial model.

Images