Emergency agent, emergency processing system and method, medium and equipment

Abstract

The invention discloses an emergency agent, an emergency processing system and method, a medium and equipment. The emergency agent comprises a planning module used for generating an emergency task plan based on an obtained first emergency processing request; the tool calling module is used for calling a preset emergency processing tool for emergency processing based on the emergency task plan; the reflection module is used for responding to the second emergency processing request received by the planning module in the emergency processing process, evaluating whether a supplementary plan for coping with the second emergency processing request exists in the current emergency processing task plan or not, and responding to the situation that no supplementary plan exists in the emergency processing task plan; and the planning module is instructed to introduce knowledge of the supplementary plan to re-plan so as to update the emergency task plan. Through the emergency intelligent agent, the emergency processing request can be flexibly and rapidly responded, linkage processing is performed on the emergency situation through various emergency plans, the efficiency and accuracy of emergency processing are improved, and the life and property safety of people can be guaranteed.

Description

Technical Field

[0001] The present application relates to the field of emergency response technology, and in particular to an emergency intelligent entity, an emergency response system, a method, a medium and a device. Background Art

[0003] The existing subway emergency command and disposal process mainly relies on pre-prepared disposal plans. When an emergency occurs, emergency response personnel will handle it according to the guidance process of the plan. However, in actual operations, emergencies are often complex and uncertain. Different emergencies may occur simultaneously or intertwine with each other. For example, a large passenger flow event at a station occurs during the flooding disposal process, which poses a great challenge to emergency response personnel. The existing subway emergency command system has certain limitations in information collection, analysis and decision support. It is difficult to respond to such complex and changeable emergencies quickly and accurately, and it is impossible to provide flexible and effective decision-making suggestions for disposal personnel. It has a low degree of intelligence and low emergency handling efficiency. Summary of the invention

[0004] The present application mainly provides an emergency intelligent entity, an emergency handling system, a method, a medium and a device, aiming to solve the technical problem of low efficiency of subway emergency handling.

[0005] In order to solve the above technical problems, the technical solution adopted by the present application is: to provide an emergency agent. The emergency agent is constructed based on a large language model, and the emergency agent includes: a planning module, which is used to generate an emergency task plan based on the acquired first emergency processing request; a tool calling module, which is connected to the planning module in communication, and is used to call the preset emergency processing tool for emergency processing based on the emergency task plan; a reflection module, which is connected to the planning module and the tool calling module in communication, and is used to respond to the planning module receiving a second emergency processing request during the emergency processing, evaluate whether there is a supplementary plan for responding to the second emergency processing request in the current emergency processing task plan, and respond to the absence of the supplementary plan in the emergency task plan, instruct the planning module to introduce the knowledge of the supplementary plan for re-planning to update the emergency task plan.

[0006] In some embodiments, the planning module is also used to: obtain multiple knowledge text blocks from a preset knowledge base based on the first emergency processing request; recall and reorder the multiple knowledge text blocks to obtain multiple recalled knowledge blocks; summarize the multiple recalled knowledge blocks to obtain emergency knowledge text; perform planning semantic analysis on the emergency knowledge text based on a large language model to generate the emergency task plan; wherein the preset knowledge base includes at least one of the knowledge vector space built into the emergency agent and the external emergency plan knowledge base called by the tool calling module.

[0007] In some embodiments, the reflection module is also used to: perform comprehensive scoring and indicator analysis on the recalled knowledge block, the emergency knowledge text and the emergency task plan to obtain a first evaluation result of the emergency task plan and a corresponding first adjustment instruction; in response to determining that the first evaluation result does not meet the preset output conditions, instruct the planning module to adjust parameters based on the first adjustment instruction, and / or instruct the planning module to introduce new knowledge based on the first adjustment instruction, so that the planning module generates a new emergency task plan.

[0008] In some embodiments, the reflection module is also used to: evaluate the emergency treatment results of the emergency treatment tool called to obtain a second evaluation result and a corresponding second adjustment instruction; in response to determining that the second evaluation result does not meet the preset processing result conditions, instruct the planning module to adjust parameters based on the second adjustment instruction, and / or instruct the planning module to introduce new knowledge based on the second adjustment instruction, so that the planning module generates a new emergency task plan.

[0009] In some embodiments, the emergency handling tool includes at least one of a knowledge base retrieval tool, a large language model generation tool, an emergency handling contact tool, an emergency material access tool, a broadcast sending tool, a notification publishing tool, an emergency agency management tool, and a traffic dispatch management tool.

[0010] In some embodiments, the planning module obtains the first emergency handling request and / or the second emergency handling request based on natural language interaction, wherein the natural language interaction includes text interaction and / or speech-to-text interaction.

[0011] In order to solve the above-mentioned technical problems, another technical solution adopted in the present application is: to provide an emergency response system, which includes an emergency command platform with a communication connection and an emergency intelligent entity as mentioned above; the emergency command platform is used to input the acquired first emergency response request and the second emergency response request into the emergency intelligent entity, so as to perform emergency response through the emergency intelligent entity.

[0012] To solve the above technical problems, another technical solution adopted in the present application is: to provide an emergency handling method, which includes: obtaining an emergency intelligent agent as mentioned above; inputting a first emergency handling request into the emergency intelligent agent, and selectively inputting a second emergency handling request into the emergency intelligent agent to perform emergency handling through the emergency intelligent agent.

[0013] In order to solve the above technical problems, another technical solution adopted by the present application is: to provide a storage medium, on which program data is stored, characterized in that when the program data is executed by a processor, the steps of the emergency handling method as described above are implemented.

[0014] To solve the above technical problems, another technical solution adopted in the present application is: to provide a computer device, which includes a processor and a memory connected to each other, the memory stores a computer program, and when the processor executes the computer program, the steps of the emergency handling method as described above are implemented.

[0015] The beneficial effects of the present application are as follows: Different from the prior art, the present application discloses an emergency intelligent agent, an emergency handling system, a method, a medium and a device. The present application introduces an emergency intelligent agent built on a large language model, uses the planning module included therein to generate an emergency task plan, the tool calling module calls the corresponding emergency handling tool to support the execution of the emergency task, and the reflection module evaluates and optimizes the entire emergency handling process, thereby achieving collaborative work among the modules, being able to flexibly and quickly respond to emergency handling requests, and jointly handle emergency situations through various emergency plans, quickly respond to emergencies and make intelligent decisions, dynamically generate and adjust emergency task plans to adapt to the ever-changing emergency needs, improve the efficiency and accuracy of emergency handling, and help protect people's lives and property safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative work, among which:

[0017] Figure 1 It is a structural diagram of an embodiment of an emergency agent provided by the present application;

[0018] Figure 2 It is a structural schematic diagram of an embodiment of an emergency treatment system provided by the present application;

[0019] Figure 3 It is a flow chart of an embodiment of an emergency treatment method provided by the present application;

[0020] Figure 4 It is a structural schematic diagram of an embodiment of a storage medium provided by the present application;

[0021] Figure 5 It is a structural diagram of an embodiment of a computer device provided by the present application. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0023] The terms "first", "second", "third" in the embodiments of the present application are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Thus, the features defined as "first", "second", "third" can expressly or implicitly include at least one of the features. In the description of the present application, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or devices.

[0024] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0025] This application provides an emergency agent, see Figure 1 , Figure 1 It is a structural diagram of an embodiment of an emergency agent 10 provided by the present application. The emergency agent 10 is constructed based on a large language model, and the emergency agent 10 includes: a planning module 11, which is used to generate an emergency task plan based on the acquired first emergency processing request; a tool calling module 12, which is communicated with the planning module 11, and is used to call a preset emergency processing tool for emergency processing based on the emergency task plan; a reflection module 13, which is communicated with the planning module 11 and the tool calling module 12 respectively, and is used to respond to the planning module 11 receiving a second emergency processing request during the emergency processing, evaluate whether there is a supplementary plan to deal with the second emergency processing request in the current emergency processing task plan, and respond to the absence of a supplementary plan in the emergency task plan, instruct the planning module 11 to introduce the knowledge of the supplementary plan for re-planning to update the emergency task plan.

[0026] In this embodiment, the emergency agent 10 is constructed based on a large language model (LLM), which is a model that can process natural language. It can understand and generate human language through large-scale corpus training. The emergency agent 10 used in this application is applied in emergency scenarios, and preferably the large language model is trained and fine-tuned through the relevant corpus data of the emergency plan, so that the emergency agent 10 can understand complex emergency situation descriptions and perform corresponding emergency processing planning and decision control. Among them, the corpus data used for training and fine-tuning the emergency agent 10 may include historical emergency event records, emergency processing process records, emergency plan data, emergency response tool use cases, etc. These data can be technical precipitation in the development process of relevant emergency tools, and are used in emergency scenarios such as station flooding disposal, station large passenger flow disposal, and station fire disposal. In this embodiment, the emergency agent 10 can be obtained by constructing only one large language model, or by constructing multiple large language models to obtain a multi-emergency agent architecture, which can include emergency agents specifically for handling station flooding, emergency agents specifically for handling large passenger flows at stations, and other emergency agents corresponding to different emergency plan scenarios, thereby achieving faster and more accurate emergency handling.

[0027] In this embodiment, the first emergency handling request is a request information issued when an emergency situation occurs and emergency handling is required to obtain relevant resources and perform emergency operations. The first emergency handling request can be issued by the emergency commander through telephone communication, network communication or written report. The emergency agent 10 of this solution is used to process natural language. Therefore, the corresponding initial request issued by the emergency commander can be semantically parsed and textualized through natural language interaction to obtain the first emergency handling request, wherein the initial request is input by the relevant emergency commander through text, voice and action, or it can be a distress signal automatically recognized and generated by the monitoring alarm system. The natural language interaction includes text interaction and / or speech-to-text interaction. For example, a specific speech-to-text module can be used to convert the speech into text using automatic speech recognition (Automatic Speech Recognition). The speech-to-text module can use complex algorithms and machine learning techniques to achieve high-accuracy speech recognition, and can process speech data in real time even in a noisy environment, providing accurate input data for the emergency agent 10, thereby speeding up the emergency response. In addition, the module can also enhance the applicability and flexibility of the system by supporting multiple languages ​​and dialects. In addition, the acquisition of the first emergency processing request can also be combined with image processing, audio and video recognition, and encoding and decoding processing to process structured data to obtain the corresponding first emergency processing request.

[0028] In this embodiment, the second emergency handling request is an additional request generated during the emergency handling process due to changes in the situation, the emergence of new situations, or the need to revise the plan. Its form and acquisition method are similar to the first emergency handling request. For example, the second processing request can also be obtained by the planning module 11 through natural language interaction, which will not be repeated here. For example, the first emergency handling request is issued by the emergency commander to request flooding treatment. While the intelligent agent is handling the flooding, the emergency commander finds that an escalator failure has occurred. Then the second processing request can be a request to handle the escalator failure, so that the emergency intelligent agent 10 will simultaneously consider the situation of flooding treatment and the reasonable handling method of the escalator failure under this situation, thereby updating the corresponding emergency task plan for emergency handling, and realizing rapid response processing of the first emergency handling request and the second emergency handling request.

[0029] In this embodiment, the planning module 11 is supported by a large language model, and can generate a detailed emergency task plan based on real-time data and a preset emergency response process, and perform planning and execution by calling tools. Among them, the preset emergency response process is a process in the relevant plan content compiled in the built-in knowledge vector space of the emergency agent 10, such as "Emergency Plan for Fire Emergency Response of Station Equipment" and "Escalator Failure Response Plan of Station", etc. The preset emergency response process is used as the corresponding sample data in the training and fine-tuning process of the emergency agent 10, so that the emergency agent 10 obtained by training and fine-tuning can handle the relevant emergency situations according to the standardized process. At the same time, the planning module 11 can also monitor the progress of the execution of the plan to ensure that all tasks can be completed on time and effectively. In addition, the planning module 11 can also cooperate with the subsequent tool calling module 12 and the reflection module 13, so that it can update the emergency task plan in real time according to the changes in the on-site situation to deal with sudden emergency events.

[0030] In this embodiment, the planning module 11 is responsible for receiving the first emergency handling request on the one hand, and generating a corresponding emergency task plan according to the request content, and on the other hand, it is responsible for receiving the second emergency handling request, and dynamically adjusting based on the feedback of the reflection module 13, so as to realize continuous iterative optimization of the plan. This module can analyze the key information in the request, such as the type, scale and scope of impact of the emergency event, and build in, record or access the available resources, callable tools and related plan knowledge to formulate a detailed emergency operation process, namely the emergency task plan, which can ensure the timeliness and effectiveness of the emergency response, and at the same time take into account the reasonable allocation of resources and the minimization of risks, and generate an effective and reliable emergency task plan for subsequent planning execution to handle emergency situations.

[0031] Optionally, in one embodiment, the planning module 11 is specifically used to: obtain multiple knowledge text blocks from a preset knowledge base based on the first emergency processing request; recall and reorder the multiple knowledge text blocks to obtain multiple recalled knowledge blocks; summarize the multiple recalled knowledge blocks to obtain emergency knowledge text; perform planning semantic analysis on the emergency knowledge text based on a large language model to generate an emergency task plan.

[0032] In this optional embodiment, the preset knowledge base is a database containing rich emergency knowledge. Specifically, the preset knowledge base includes at least one of the knowledge vector space built into the emergency agent 10 and the external emergency plan knowledge base called by the tool calling module 12. Among them, the knowledge vector space built into the emergency agent 10 stores the emergency knowledge internalized by the emergency agent 10. These emergency knowledge are generated and represented by vectors during the training and fine-tuning of the large language model, and can be automatically called when the emergency agent 10 performs task planning and plan execution. The external emergency plan knowledge base contains more extensive and detailed emergency handling solutions, historical case analysis, expert advice, etc. These knowledge are stored in a database outside the emergency agent 10 and can be accessed and retrieved in real time by the tool calling module 12 as needed.

[0033] In this optional embodiment, multiple knowledge text blocks can be obtained from the preset knowledge base based on the first emergency handling request. These knowledge text blocks refer to text fragments in the preset knowledge base related to the first emergency handling request, such as emergency event type definitions, relevant plans, historical cases, expert advice and safety regulations. The obtained knowledge text blocks are usually scattered knowledge with weak relevance and difficult to integrate into easy-to-understand semantic content. Therefore, they can be further processed by recalling and reordering the text blocks.

[0034] In this optional embodiment, text block recall refers to retrieving relevant information and knowledge from a preset knowledge base, further parsing the relevant knowledge text blocks, and providing more comprehensive background information and data support through recalling text blocks, which helps the intelligent agent understand and make judgments faster and more accurately. Reordering refers to reordering the recalled text blocks or generated processing steps to ensure that they are ordered according to logic and practicality, thereby improving the logic, readability and operability of the output. For example, the first emergency handling request is "There is a fire in the exhibition hall of the subway station. Please give disposal suggestions immediately." Then the knowledge text block may be some relevant knowledge about fire handling in subway stations. After the text block is recalled, the corresponding handling measures will be obtained, such as "Immediately start the emergency evacuation procedure", "Notify the fire department and provide the specific location of the fire", "Use the fire-fighting facilities in the station for preliminary fire extinguishing" and other text blocks, and reordering is to sort out these text blocks. For example, the order after reordering is "Notify the fire department and provide the specific location of the fire", "Immediately start the emergency evacuation procedure" and "Use the fire-fighting facilities in the station for preliminary fire extinguishing", that is, when handling, you first need to ensure that the fire department knows that the fire has occurred, and evacuate the passengers immediately after notifying the fire department, and carry out preliminary fire extinguishing while evacuating.

[0035] In this optional embodiment, multiple recalled knowledge blocks can be obtained by recalling and reordering the multiple knowledge text blocks. Since the recalled knowledge blocks have been optimized and integrated, a more coherent and logical emergency knowledge text can be formed by summarizing the recalled knowledge blocks to further extract key information and form an emergency knowledge text. This process involves in-depth analysis and understanding of the recalled knowledge blocks in order to extract the most core emergency handling information from them. The summary method can be based on semantic understanding of a large language model, a rule-based text summary, an expert system-based reasoning mechanism, or a comprehensive analysis combining multiple methods.

[0036] In this optional embodiment, the summarized emergency knowledge text will be used as the input of planning semantic analysis to ensure that the generated emergency task plan is both comprehensive and accurate. Planning semantic analysis is a key step in converting the summarized emergency knowledge text into an emergency task plan. Through the advanced semantic understanding ability of the large language model, the planning module 11 can parse the intentions and instructions in the text to formulate a specific action plan. This process requires not only understanding the literal meaning of the text, but also grasping the emergency logic and priority behind it to ensure the rationality and executability of the plan. Ultimately, the emergency task plan generated by it will guide the emergency agent 10 and related equipment to perform specific emergency operations. The plan will detail the operation content, execution order, required resources and expected results of each step to ensure that it can respond quickly and effectively in the face of complex and changeable emergency situations. In addition, the planning module 11 also has a dynamic adjustment function, which can update the emergency task plan in real time according to changes in the on-site situation to adapt to the changing emergency needs.

[0037] In this embodiment, the tool calling module 12 works closely with the planning module 11 and is responsible for calling the corresponding emergency handling tools to perform emergency handling according to the emergency tasks generated by the planning module 11. These tools may include various software applications, hardware devices or professional services, etc., which are used to perform specific emergency operations to ensure that the emergency agent 10 can quickly respond and execute the planned tasks, thereby improving the efficiency and effectiveness of the overall emergency handling.

[0038] Optionally, in one embodiment, the emergency handling tool includes at least one of a knowledge base retrieval tool, a large language model generation tool, an emergency handling contact tool, an emergency material access tool, a broadcast sending tool, a notification publishing tool, an emergency agency management tool, and a traffic dispatch management tool.

[0039] In this optional embodiment, the knowledge base retrieval tool is a software tool for quickly retrieving and acquiring emergency response related knowledge. The tool is connected to an external knowledge base through communication, and retrieves the required knowledge information, such as emergency plans, historical cases and expert advice, from the external knowledge base through specific signaling, to provide data support for the planning generation of the planning module 11 or in the subsequent planning execution; the large language model generation tool uses the capabilities of the large language model to generate text content for emergency handling, such as generating planning task details, planning execution progress, escape guidance information and broadcast information, etc. These text contents can be used by relevant on-site personnel and related systems for emergency operation guidance and reporting necessary information; the emergency response contact tool is used to quickly contact relevant The emergency response personnel or organization ensures that the emergency situation can be communicated to the relevant personnel performing the emergency tasks in a timely manner; the emergency material access tool is responsible for managing the inventory and distribution of emergency materials, ensuring that materials can be quickly dispatched when needed, and providing material support for emergency response; the broadcast sending tool and the notification publishing tool are used to send emergency notifications and warning information to personnel in the area and external relevant personnel to guide personnel to take emergency actions; the emergency agency management tool is used to coordinate cooperation and communication between different emergency agencies to ensure the unity and efficiency of emergency response; the traffic dispatch management tool is used to adjust the transportation system in emergency situations, such as changing the train operation plan, adjusting the train speed and route, etc., to avoid or reduce the impact of accidents.

[0040] In this optional embodiment, the emergency handling tool can be connected to the tool calling module 12 through the tool interface provided by the backend service, so that the tool calling module 12 can pass parameters through the tool interface and the corresponding protocol to call the corresponding emergency handling tool. These passed parameters can be generated by the planning module 11 according to the current handling situation to accurately and effectively call the tool. For example, the parameters for calling the emergency handling contact tool include station name, date and other parameters. These parameters can be preset in advance by the planning module 11 according to its specific deployment site and obtained by clock synchronization of the network. For example, the station name is Changping Station and the date is November 21. After the emergency handling contact tool obtains these parameters through the interface, it will query the background data to output the position, name, contact number, work number, etc. of the person on duty at the site on that day, so as to get in touch with the person on duty.

[0041] In this optional embodiment, the emergency handling tool is dynamically selected and called by the tool calling module 12 according to the needs of the emergency task planning. By selecting the appropriate tool for calling, the smooth progress of the emergency handling can be ensured. The calling process can be a call of multiple tools one by one or at the same time. Through the integration and application of these tools, the rapid circulation of emergency-related information and resource sharing are ensured. In addition, when managing and maintaining the emergency intelligent body 10, it is also possible to use the tool calling module 12 to perform self-checks on each emergency handling tool to ensure that all tools are in the latest state and can be quickly deployed as needed, so that the emergency intelligent body 10 can flexibly respond to various complex emergency situations and achieve rapid and accurate emergency response.

[0042] In this embodiment, the emergency agent 10 may still encounter problems such as output not meeting the user's intention, irregular output format, and the tool not returning correct knowledge when processing complex tasks, and it is necessary to improve the performance and accuracy of the model output through continuous iteration. The emergency agent 10 in this embodiment also specifically uses the reflection module 13 to perform reflection agents (ReflectionAgents), and establishes a mechanism for evaluating and optimizing the emergency response process to improve the self-learning and adaptability of the emergency agent 10. The reflection module 13 is responsible for evaluating the emergency treatment process, identifying and analyzing errors or deficiencies in the treatment. For example, when the emergency agent 10 is dealing with a flooding accident, if it is found that the evacuation route recommended by it is not ideal, the reflection module 13 will record this situation and adjust the algorithm in subsequent processing to avoid similar problems from happening again. Through such an iterative process, the emergency agent 10 can continuously learn and improve, thereby providing a more accurate and effective solution in emergency response. The reflection module 13 can adopt different self-reflection algorithms, such as feedback-based adaptive algorithms, case-based reasoning algorithms, and rule-based adjustment algorithms. These algorithms can help the emergency agent 10 learn from historical emergency handling experience or corrective suggestions during the execution process, and automatically adjust and optimize the emergency response strategy.

[0043] In the present embodiment, the self-reflection algorithm preferably adopts LangGraph to connect the planning module 11 with the reflection module 13, uses the planning module 11 as the execution generator (Generator), and uses the reflection module 13 as the reflector (Reflector), and forms a reflection algorithm logic diagram for implementation. By connecting the planning module 11 and the reflection module 13, they can work together, the planning module 11 is responsible for generating preliminary outputs, and the reflection module 13 is responsible for evaluating and optimizing these outputs. Through this connection, an iterative process can be realized, that is, the emergency task planning generated by the planning module 11 is evaluated by the reflection module 13, adjusted according to the evaluation results, and then generated again to form a cycle until the output meets certain quality standards. Through continuous iteration and optimization, the performance and accuracy of the model output are improved, so that the model can better meet the user's intention and specification requirements. The reflection algorithm logic diagram formed reflects the interactive process between the planning module 11 and the reflection module 13, including the steps of generation, evaluation, feedback and adjustment, providing a clear guide, which helps relevant personnel understand and implement the self-reflection algorithm and ensures that each step can be correctly executed. The reflection algorithm logic diagram can improve the output quality by ensuring that the goals of the planning module 11 and the reflection module 13 are consistent, continuously monitor the output quality during the iteration process, ensure that each iteration can bring improvements, and establish an effective feedback mechanism to ensure that the evaluation results of the reflection module 13 can be correctly understood and applied by the planning module 11. Among them, the evaluation method of the reflection module 13 can be specifically implemented by scoring indicators such as accuracy and compliance with specifications to evaluate the quality of planning output and tool call processing, and can determine whether the score meets the requirements by setting a threshold of the indicator or a threshold of the number of iterations. When the reflection module 13 confirms that the score is higher than the threshold or the number of iterations exceeds a certain number, it stops the iteration and performs planning output and adjusts the tool call.

[0044] In this embodiment, the reflection module 13 performs linkage treatment on various emergency plans based on the current treatment situation, the planning module 11 and the tool calling module 12. When the emergency agent 10 receives the second emergency handling request, the reflection module 13 first evaluates whether the current emergency task plan contains the corresponding supplementary plan to deal with the new request. If there is no corresponding supplementary plan in the plan, the reflection module 13 will guide the planning module 11 to obtain the knowledge related to the supplementary plan and re-plan to ensure that the emergency task plan can be updated in time and adapt to new emergency situations. This process involves rapid analysis of existing plans and integration of new knowledge to generate a more comprehensive and adaptable emergency response plan. The reflection module 13 is not only responsible for supervising the output of the planning module 11, but also for ensuring that the tool calling module 12 can perform correct operations according to the latest plan. Through this linkage treatment, the emergency agent 10 can flexibly respond to changing emergency needs, thereby improving the efficiency and effectiveness of the overall emergency response. The evaluation criteria of reflection module 13 are mainly whether the current task execution and the called tools can meet the current disposal execution, evaluate the effectiveness of planning and tool calling, and put forward possible improvement suggestions, so that when other emergency situations occur in the middle, each module can work together to handle emergencies and respond to emergencies in various situations more quickly, effectively and reliably.

[0045] In this embodiment, the above linkage treatment is explained by taking an example: for example, a station flooding event is reported to the emergency agent 10 through the first emergency treatment request. The reporting method can be specifically that the station staff reports it through text or voice, or the agent monitors the water level sensor and other means. After the agent receives the first emergency treatment request, it first obtains the relevant plan documents through the planning module 11 and generates an emergency task plan. Through the emergency task plan, the emergency treatment tool is called to perform emergency treatment, such as first calling the broadcast sending tool to notify the line staff to close the entrance, and then calling the emergency treatment contact tool to obtain the treatment personnel information and calling the SMS notification tool to notify the corresponding treatment personnel to close the entrance and exit. In this process, further, the station staff can initiate a second emergency treatment request to specifically report that the current flooding situation has reached the situation of level 1 flooding. The agent determines through the reflection module 13 that it has a plan for handling level 1 flooding in the task plan, so it can continue to execute the plan, or call more knowledge of handling level 1 flooding to optimize the plan. Furthermore, the station staff can continue to supplement the second emergency handling request, such as reporting an escalator failure. The emergency agent 10 will then evaluate the current emergency task plan through the reflection module 13. For example, it may be found that there is no supplementary plan for escalator failure in the event of flooding. Then the reflection module 13 will feedback the situation to the planning module 11 to instruct the planning module 11 to introduce knowledge related to the supplementary plan for re-planning. The planning module 11 obtains relevant plans for escalator failure handling by calling the knowledge base retrieval tool based on the feedback, thereby continuing to iterate the planning and reflection of the next step to update the emergency task plan and continue to perform corresponding emergency handling and reflection optimization.

[0046] Optionally, in one embodiment, the reflection module 13 is specifically used to: perform comprehensive scoring and indicator analysis on the recalled knowledge blocks, emergency knowledge texts and emergency task plans to obtain a first evaluation result of the emergency task plan and a corresponding first adjustment instruction; in response to determining that the first evaluation result does not meet the preset output conditions, instruct the planning module 11 to adjust parameters based on the first adjustment instruction, and / or instruct the planning module 11 to introduce new knowledge based on the first adjustment instruction, so that the planning module 11 generates a new emergency task plan.

[0047] In this optional embodiment, the reflection module 13 is specifically provided with the function of evaluating the emergency task planning process, and the first evaluation result obtained can accurately reflect the quality and efficiency of the emergency task planning. Through comprehensive scoring and indicator analysis, the reflection module 13 can identify deficiencies in the planning, such as incomplete planning, inaccurate execution, or irrational resource allocation. These evaluation results and adjustment instructions provide a clear direction for improvement for the planning module 11, ensuring that the emergency agent 10 can dynamically adjust the emergency response strategy according to the actual situation, thereby improving the adaptability and flexibility of the emergency planning.

[0048] In this optional embodiment, the preset output conditions are used to ensure that the output of the emergency task planning can meet specific accuracy and timeliness standards, while ensuring the rationality of resource allocation and the efficiency of emergency response. These conditions include, but are not limited to: the completeness of the plan, the standardization of task execution, the optimization of resource utilization, and the minimization of emergency response time. In addition, the output conditions may also involve performance evaluation of the emergency agent 10, such as response speed, processing capacity, and adaptability in different emergency scenarios. Through these preset conditions, it can be ensured that the emergency agent 10 can achieve the expected performance standards in actual operation, providing reliable and effective support for emergency handling. When it is determined that the first evaluation result does not meet the preset output conditions, the reflection module 13 will instruct the planning module 11 to adjust the parameters based on the first adjustment instruction, or instruct the planning module 11 to introduce new knowledge based on the first adjustment instruction to generate a new emergency task plan, ensuring that the emergency agent 10 can continuously adjust and optimize the emergency response strategy according to the feedback and evaluation results, thereby improving the efficiency and effect of the overall emergency handling.

[0049] Optionally, in some embodiments, the reflection module 13 is specifically used to: evaluate the emergency treatment results of the called emergency treatment tool to obtain a second evaluation result and a corresponding second adjustment instruction; in response to determining that the second evaluation result does not meet the preset processing result conditions, instruct the planning module 11 to adjust parameters based on the second adjustment instruction, and / or instruct the planning module 11 to introduce new knowledge based on the second adjustment instruction, so that the planning module 11 generates a new emergency task plan.

[0050] In this optional embodiment, the reflection module 13 is specifically provided with the function of evaluating the effect of calling the emergency handling tool, and the second evaluation result obtained can accurately reflect the efficiency and effect of calling the emergency handling tool. By evaluating the execution results of the emergency handling tool, the reflection module 13 can identify problems in the tool call, such as too long response time, improper resource allocation or inaccurate execution. These evaluation results and adjustment instructions provide the planning module 11 with a direction for improvement, ensuring that the emergency agent 10 can adjust the emergency response strategy according to the actual situation, thereby improving the effectiveness and reliability of the emergency handling tool call.

[0051] In this optional embodiment, the preset processing structure conditions are used to ensure that the output of the emergency handling tool can meet specific efficiency and accuracy standards. These conditions may include minimization of tool response time, optimization of resource allocation, and correctness of execution results. In addition, the preset conditions may also involve the evaluation of the performance of the emergency handling tool, such as the reliability, stability and adaptability of the tool in different emergency situations. Through these preset conditions, it can be ensured that the emergency handling tool can meet the expected performance standards in actual operation, providing reliable and effective support for emergency handling. When it is determined that the second evaluation result does not meet the preset processing result conditions, the reflection module 13 will instruct the planning module 11 to adjust the parameters based on the second adjustment instruction, or instruct the planning module 11 to introduce new knowledge based on the second adjustment instruction to generate a new emergency task plan, so as to ensure that the emergency agent 10 can continuously adjust and optimize the emergency response strategy according to the feedback and evaluation results, thereby improving the efficiency and effectiveness of the overall emergency handling.

[0052] In this embodiment, the reflection module 13 can perform a multi-dimensional evaluation of emergency task planning and the calling of emergency handling tools. Through reflection and optimization, it can ensure that the corresponding emergency handling is more logically consistent, feasible, safe and standardized, identify problems such as non-compliance with user intentions, non-standard formats and tool calling errors in the task planning and tool calling process, and propose specific improvement measures or suggestions based on the evaluation results, instructing the planning module 11 to make adjustments and optimizations, so that emergency handling is more accurate, reasonable and effective, and can effectively improve the quality of decision-making, so that the emergency intelligent agent 10 can better adapt to the ever-changing emergency situations, improve the ability to handle complex situations, reduce risks in the emergency handling process, and significantly improve the performance of the emergency intelligent agent 10.

[0053] In this embodiment, the reflection module 13 can not only evaluate the emergency task planning and emergency handling tool calls, but also monitor and analyze the entire emergency response process to ensure that the emergency agent 10 can make quick and accurate decisions when facing complex and changeable emergency situations. Through continuous self-learning and optimization, the emergency agent 10 can adapt to the ever-changing environment and needs, thereby providing more efficient and reliable emergency response services in emergency situations. In addition, the reflection module 13 can also record and analyze historical emergency events, extract lessons from them, provide references for future emergency planning and response, and further enhance the adaptability and flexibility of the emergency agent 10. Through this mechanism, the emergency agent 10 can continuously optimize, flexibly and quickly respond to emergency handling requests, and deal with emergency situations in a coordinated manner through various emergency plans, thereby improving the efficiency and accuracy of emergency handling, and helping to protect people's lives and property safety.

[0054] See also Figure 2 , Figure 2 It is a structural schematic diagram of an embodiment of an emergency response system provided by the present application.

[0055] The emergency handling system 20 includes an emergency command platform 21 and a communication connection. Figure 1 The described emergency agent 10 and the emergency command platform 21 are used to input the acquired first emergency processing request and the second emergency processing request into the emergency agent 10 so as to perform emergency processing through the emergency agent 10.

[0056] The emergency command platform 21 can specifically be an interactive platform provided by organizations such as the subway emergency command center, mobile command center, and remote command center. These platforms are connected to the emergency intelligent body 10 through different communication methods to ensure that the emergency intelligent body 10 can receive and process emergency requests in a timely manner under various circumstances, and ensure the timely reporting and processing of emergency situations. At the same time, the emergency command platform 21 also optionally has data collection and analysis functions, which can collect real-time data from the scene and provide it to the emergency intelligent body 10 for more accurate emergency planning and decision-making. In addition, the emergency command platform 21 can also exchange data with external systems such as the Meteorological Bureau, the Traffic Management Center, etc. to obtain more comprehensive emergency information.

[0057] See also Figure 3 , Figure 3 1 is a flow chart of an embodiment of an emergency treatment method provided by the present application, and the emergency treatment method includes:

[0058] Step 31: Get Figure 1 The emergency agent 10 described.

[0059] Step 32: Input the first emergency processing request into the emergency agent 10, and selectively input the second emergency processing request into the emergency agent 10, so as to perform emergency processing through the emergency agent 10.

[0060] In the emergency handling method, the input of the first emergency handling request is used to construct the relevant emergency handling task, and the second emergency handling request is a supplementary request in the emergency handling process, so it can be input selectively, that is, input when the supplementary request exists, and no input is required when it does not exist. The emergency handling request can be input into the emergency agent 10 in a specific way, such as actively inputting through text or voice-to-text, or passively triggered by the emergency agent 10 through sensors or detection mechanisms. By inputting the emergency handling request into the emergency agent 10, the emergency agent 10 can perform emergency handling according to its corresponding emergency response processing mechanism.

[0061] See also Figure 4 , Figure 4It is a structural diagram of an embodiment of the storage medium provided by the present application.

[0062] The storage medium 40 stores program data 41. When the program data 41 is executed by the processor, the following is achieved: Figure 3 Described emergency response methods.

[0063] The program data 41 is stored in a storage medium 40, and includes a number of instructions for enabling a network device (such as a router, a personal computer, a server, or other network device) or a processor to execute all or part of the steps of the methods of the various embodiments of the present application.

[0064] Optionally, the storage medium 40 may be a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, an optical disk, or other medium that can store the program data 41 .

[0065] See also Figure 5 , Figure 5 It is a structural schematic diagram of an embodiment of a computer device provided by the present application.

[0066] The computer device 50 includes a processor 52 and a memory 51 connected to each other. The memory 51 stores a computer program. When the processor 52 executes the computer program, the following is achieved: Figure 3 The described emergency handling method. The memory 51 may include the storage medium 40, or may be other independently developed memory.

[0067] Different from the prior art, the present application discloses an emergency agent, an emergency handling system, a method, a medium and a device. By introducing an emergency agent built on a large language model, using the planning module included therein to generate an emergency task plan, the tool calling module calling the corresponding emergency handling tool to support the execution of the emergency task, and the reflection module evaluating and feedback optimizing the entire emergency handling process, the collaborative work between the modules is realized, and the emergency handling request can be responded to flexibly and quickly, and emergency situations can be handled in a linked manner through various emergency plans, and sudden incidents can be responded to quickly and intelligently. The emergency task plan is dynamically generated and adjusted to adapt to the ever-changing emergency needs, thereby improving the efficiency and accuracy of emergency handling, and being conducive to protecting people's lives and property safety.

[0068] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for system embodiments, method embodiments, medium embodiments, and device embodiments, since they are basically similar to the emergency agent embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the emergency agent embodiments.

[0069] The present application can be used in many general or special computing system environments or configurations, such as personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0070] In the several embodiments provided in this application, it should be understood that the disclosed emergency agent, emergency handling system, method, storage medium and computer device can be implemented in other ways. For example, the emergency agent implementation described above is only illustrative, for example, the division of modules or units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0071] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0072] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0073] The above are merely embodiments of the present application and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. An emergency agent, characterized in that: The emergency agent is constructed based on a large language model, and the emergency agent includes: A planning module, used for generating an emergency task plan based on the acquired first emergency processing request; A tool calling module, which is in communication with the planning module and is used to call a preset emergency processing tool to perform emergency processing based on the emergency task planning; The reflection module is communicated with the planning module and the tool calling module respectively, and is used to respond to the planning module receiving a second emergency processing request during the emergency processing process, evaluate whether there is a supplementary plan to deal with the second emergency processing request in the current emergency processing task plan, and in response to the absence of the supplementary plan in the emergency task plan, instruct the planning module to introduce the knowledge of the supplementary plan for re-planning to update the emergency task plan.

2. The emergency agent according to claim 1, characterized in that: The planning module is also used to: Acquire multiple knowledge text blocks from a preset knowledge base based on the first emergency processing request; Recalling and reordering the multiple knowledge text blocks to obtain multiple recalled knowledge blocks; Summarizing the multiple recalled knowledge blocks to obtain emergency knowledge text; Performing planning semantic analysis on the emergency knowledge text based on a large language model to generate the emergency task plan; Among them, the preset knowledge base includes at least one of the knowledge vector space built into the emergency agent and the external emergency plan knowledge base called by the tool calling module.

3. The emergency agent according to claim 2, characterized in that: The reflection module is also used to: Performing comprehensive scoring and index analysis on the recall knowledge block, the emergency knowledge text, and the emergency task plan to obtain a first evaluation result of the emergency task plan and a corresponding first adjustment instruction; In response to determining that the first evaluation result does not meet the preset output condition, the planning module is instructing to adjust parameters based on the first adjustment instruction, and / or the planning module is instructing to introduce new knowledge based on the first adjustment instruction, so that the planning module generates a new emergency task plan.

4. The emergency agent according to claim 1, characterized in that: The reflection module is also used to: Evaluate the emergency processing result of the emergency processing tool called to obtain a second evaluation result and a corresponding second adjustment instruction; In response to determining that the second evaluation result does not meet the preset processing result condition, the planning module is instructed to adjust parameters based on the second adjustment instruction, and / or the planning module is instructed to introduce new knowledge based on the second adjustment instruction, so that the planning module generates a new emergency task plan.

5. The emergency agent according to claim 1, characterized in that: The emergency handling tool includes at least one of a knowledge base retrieval tool, a large language model generation tool, an emergency handling contact tool, an emergency material access tool, a broadcast sending tool, a notification publishing tool, an emergency agency management tool and a traffic dispatch management tool.

6. The emergency agent according to claim 1, characterized in that: The planning module obtains the first emergency handling request and / or the second emergency handling request based on natural language interaction, wherein the natural language interaction includes text interaction and / or speech-to-text interaction.

7. An emergency handling system, characterized in that: The emergency handling system comprises an emergency command platform and an emergency intelligent agent as claimed in any one of claims 1 to 6, which are connected in communication; The emergency command platform is used to input the acquired first emergency processing request and the second emergency processing request into the emergency intelligent agent so as to perform emergency processing through the emergency intelligent agent.

8. An emergency treatment method, characterized in that: include: Obtaining the emergency agent according to any one of claims 1 to 6; A first emergency processing request is input into the emergency agent, and a second emergency processing request is selectively input into the emergency agent, so that emergency processing is performed through the emergency agent.

9. A storage medium having program data stored thereon, characterized in that: When the program data is executed by a processor, the steps of the emergency handling method according to claim 8 are implemented.

10. A computer device, characterized in that: It comprises a processor and a memory connected to each other, the memory stores a computer program, and when the processor executes the computer program, the steps of the emergency handling method as claimed in claim 8 are implemented.

Images