A method and apparatus for intelligent agent simulation decision making based on a large language model

Abstract

This invention provides a method and apparatus for intelligent agent simulation decision-making based on a large language model. The method includes: acquiring basic attribute information of the intelligent agent and environmental information of a real-time disaster relief scenario; serializing this information and transmitting it to a decision-making module; the decision-making module using a large language model to convert this serialized data into natural language descriptions; and combining historical decision records and a behavior database to generate a decision result containing action plans and decision explanations; then transmitting the decision result to an execution module for execution and recording it in a message history database. This method enhances the flexibility and adaptability of intelligent agent decision-making, enabling it to flexibly respond to changing disaster relief scenarios and ensure accurate and efficient decision-making in complex or unknown environments. Simultaneously, it efficiently integrates multi-dimensional data to avoid information limitations and improve decision accuracy. The transparent and interpretable decision-making process enhances credibility and comprehensibility. The use of a large language model improves development efficiency, bringing greater timeliness to practical system development.

Description

Technical Field

[0001] This invention relates to the field of intelligent agent simulation decision-making technology, and in particular to an intelligent agent simulation decision-making method based on a large language model, an intelligent agent simulation decision-making device based on a large language model, an electronic device, and a computer-readable medium. Background Technology

[0002] In the field of intelligent agent simulation decision-making, existing technologies can be mainly divided into the following methods:

[0003] (1) Rule-based decision-making system

[0004] These systems rely on manually written decision-making rules, typically built using expert systems or knowledge bases. These rules reason based on different environmental states and agent attributes, providing corresponding decisions. For example, in some simple simulation environments, decisions can be made based on pre-set rules (such as "if the water depth exceeds 50cm, the agent should move to higher ground").

[0005] (2) Traditional machine learning methods

[0006] Methods such as decision trees, support vector machines (SVM), and linear regression predict the behavior of agents by training models. These methods typically rely on large amounts of training data and feature selection, and can model certain variables in the environment, but they may encounter performance bottlenecks in complex and ever-changing environments.

[0007] (3) Reinforcement Learning (RL)

[0008] Reinforcement learning methods train an agent through interaction with its environment. The agent learns the optimal policy based on feedback signals (such as reward or penalty values) from the environment. Reinforcement learning has achieved good results in many complex decision-making problems, but it typically requires a significant amount of training time, and for some high-dimensional and diverse environments, the training process can be very complex and its effectiveness limited.

[0009] (4) Deep learning (e.g., deep reinforcement learning)

[0010] Feature extraction and decision prediction are performed using neural network models (such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs). These methods can handle more complex high-dimensional input data and have strong generalization capabilities, but they also face problems such as strong data dependence, large computational cost, and unstable training.

[0011] The aforementioned existing technologies have the following drawbacks:

[0012] (1) Limitations of rule design

[0013] Rule-based systems rely on experts manually writing decision rules, which are typically static and limited. When faced with dynamic, unknown, or highly complex environments, the rules' coverage is insufficient, and the agent may fail to make reasonable decisions. Furthermore, the number of rules grows exponentially with increasing environmental complexity, limiting the system's scalability and flexibility.

[0014] (2) Data requirements and feature selection issues

[0015] Traditional machine learning methods typically require large amounts of labeled data to train models, and this data must fully represent all possible environmental states. However, in many practical applications, obtaining high-quality labeled data is very difficult and costly. Furthermore, extracting the most effective features from complex raw data relies on human experience and design, and existing features may fail to effectively capture crucial information when facing new environments.

[0016] (3) Low development efficiency

[0017] While methods like reinforcement learning can improve decision quality through multiple interactions, they typically require lengthy training processes, often necessitating large amounts of data and computational resources. This makes it difficult to provide rapid and flexible decision support for real-time decision-making or dynamically changing environments.

[0018] (4) The black box nature and lack of explanatory power in decision-making

[0019] Most existing machine learning and deep learning methods, such as deep neural networks and reinforcement learning, are often "black box" models. While they can generate relatively good decision results, they lack interpretability of the decision-making process. This is especially problematic in critical areas such as security and ethics, where users or decision-makers need to understand the agent's decision-making rationale for oversight or adjustments. Current technologies cannot clearly explain why an agent makes a specific decision, reducing user trust and acceptability.

[0020] (5) Inability to effectively handle multi-source information and complex relationships

[0021] In multi-agent systems or when dealing with complex, multi-dimensional, and spatiotemporally related information, traditional methods often struggle to comprehensively handle large amounts of heterogeneous data and changing decision factors. While deep learning-based systems can handle more complex input data, they still face significant challenges in dealing with the intricate temporal relationships and contextual dependencies between agents and their environment. Summary of the Invention

[0022] In view of the above problems, the present invention is proposed to provide an agent simulation decision-making method based on a large language model that overcomes or at least partially solves the above problems, and a corresponding agent simulation decision-making device based on a large language model, an electronic device, and a computer-readable medium.

[0023] This invention discloses an intelligent agent simulation decision-making method based on a large language model, the method comprising:

[0024] Acquire basic attribute information of the intelligent agent and collect real-time environmental information of the disaster relief scenario through multiple data sources;

[0025] The basic attribute information and current environmental information of the intelligent agent are serialized to obtain serialized data of basic attributes and serialized data of environmental information, and then the serialized data of basic attributes and serialized data of environmental information are transmitted to the decision module.

[0026] The decision module uses a large language model to transform serialized data of basic attributes and environmental information into natural language descriptions of basic attributes and environmental information, respectively. Based on these descriptions, and combined with historical decision records and a behavior database, decision results are generated, including action plans and decision explanations.

[0027] The decision results returned by the decision module are transmitted to the execution module for execution, and the decision results are recorded in the message history database.

[0028] Optionally,

[0029] The behavior library includes standard operating procedures and standard behavioral patterns;

[0030] Standard operating procedures include evacuation route selection, distribution of relief supplies, and search for emergency shelters;

[0031] Standard behavioral patterns include behavioral classification, behavioral pattern mapping for different environmental states, and behavioral decision priority rules.

[0032] Optionally, the method further includes:

[0033] The execution module returns the execution results to the decision module, which then performs dynamic learning and optimization of the large language model based on the execution results to ensure that the decision matches the current environment state.

[0034] Optionally,

[0035] The current environmental information in a disaster relief scenario includes one or more of the following: the occurrence of a natural disaster, the availability of relief resources, the availability of relief personnel, traffic conditions, geographic information, and weather and climate information.

[0036] If the disaster relief scenario is a storm surge scenario, the environmental information includes one or more of the following: water level and depth changes, traffic conditions, topographic map, population distribution, location of rescue resources, location of shelters, occupancy status of shelters, presence of rescue personnel, ambient temperature, and ambient humidity.

[0037] Optionally,

[0038] Historical decision records are stored in the message history database, including action plans and decision explanations.

[0039] Optionally, the method further includes:

[0040] The large model makes decisions periodically at preset decision time intervals, and / or determines whether to adjust the decisions based on changes in environmental information.

[0041] This invention also discloses an intelligent agent simulation decision-making device based on a large language model, the device comprising:

[0042] The real-time information collection module is used to acquire basic attribute information of the intelligent agent and collect current environmental information of the disaster relief scenario in real time through multiple data sources;

[0043] The serialization processing module is used to serialize the basic attribute information and current environmental information of the agent to obtain serialized basic attribute data and serialized environmental information data, and then transmit the serialized basic attribute data and serialized environmental information data to the decision module.

[0044] The decision result generation module is used to transform the serialized data of basic attributes and environmental information into natural language descriptions of basic attributes and environmental information through the large language model of the decision module. Based on the natural language descriptions of basic attributes and environmental information, and combined with historical decision records and behavior database, the module generates decision results. The decision results include action plans and decision explanations.

[0045] The decision result transmission and recording module is used to transmit the decision results returned by the decision module to the execution module for execution, and to record the decision results in the message history database.

[0046] Optionally,

[0047] The behavior library includes standard operating procedures and standard behavioral patterns;

[0048] Standard operating procedures include evacuation route selection, distribution of relief supplies, and search for emergency shelters;

[0049] Standard behavioral patterns include behavioral classification, behavioral pattern mapping for different environmental states, and behavioral decision priority rules.

[0050] Optionally, the device further includes:

[0051] The decision model dynamic update module is used by the execution module to return the execution results to the decision module. The decision module dynamically learns and optimizes the large language model based on the execution results to ensure that the decision matches the current environment state.

[0052] Optionally,

[0053] The current environmental information in a disaster relief scenario includes one or more of the following: the occurrence of a natural disaster, the availability of relief resources, the availability of relief personnel, traffic conditions, geographic information, and weather and climate information.

[0054] If the disaster relief scenario is a storm surge scenario, the environmental information includes one or more of the following: water level and depth changes, traffic conditions, topographic map, population distribution, location of rescue resources, location of shelters, occupancy status of shelters, presence of rescue personnel, ambient temperature, and ambient humidity.

[0055] Optionally,

[0056] Historical decision records are stored in the message history database, including action plans and decision explanations.

[0057] Optionally, the device further includes:

[0058] The timed dynamic decision-making module is used for the large model to make decisions periodically according to a preset decision time interval, and / or to determine whether to adjust the decisions based on changes in environmental information.

[0059] The present invention also discloses an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0060] The memory is used to store computer programs;

[0061] When the processor executes the program stored in the memory, it implements the intelligent agent simulation decision-making method based on a large language model as described in this invention.

[0062] The present invention also discloses one or more computer-readable media having instructions stored thereon that, when executed by one or more processors, cause the processors to perform the agent simulation decision-making method based on a large language model as described in the present invention.

[0063] This invention has the following advantages:

[0064] This invention presents an intelligent agent simulation decision-making method based on a large language model. It acquires basic attribute information of the intelligent agent and environmental information from real-time disaster relief scenarios, serializes this information, and transmits it to the decision-making module. The decision-making module uses the large language model to convert this serialized data into natural language descriptions and, combined with historical decision records and a behavior database, generates decision results containing action plans and decision explanations. These results are then transmitted to the execution module for execution and recorded in a message history database. This method significantly enhances the flexibility and adaptability of intelligent agent decision-making, enabling the agent to flexibly respond to changing disaster relief scenarios and task requirements, ensuring accurate and efficient decision-making in complex or unknown environments. Secondly, by efficiently integrating and processing multi-dimensional data, it avoids the limitations of insufficient information processing or information silos, improving the accuracy of decision-making. Furthermore, the transparency and interpretability of the decision-making process make the agent's decision-making basis clear and traceable, enhancing the credibility and comprehensibility of the decisions. Finally, leveraging the powerful computational capabilities of the large language model, this method significantly improves the development efficiency of intelligent agent decision-making, bringing greater timeliness to practical system development. Attached Figure Description

[0065] Figure 1 This is a flowchart illustrating the steps of an intelligent agent simulation decision-making method based on a large language model, as provided in an embodiment of the present invention.

[0066] Figure 2 This is a flowchart of the intelligent agent simulation decision-making process based on a large language model provided in an embodiment of the present invention;

[0067] Figure 3 This is a structural block diagram of an intelligent agent simulation decision-making device based on a large language model provided in an embodiment of the present invention;

[0068] Figure 4 This is a block diagram of an electronic device provided in an embodiment of the present invention;

[0069] Figure 5 This is a schematic diagram of a computer-readable medium provided in an embodiment of the present invention. Detailed Implementation

[0070] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0071] Reference Figure 1 The diagram illustrates a flowchart of a smart agent simulation decision-making method based on a large language model, as provided in an embodiment of the present invention. Specifically, it may include the following steps:

[0072] Step 101: Obtain the basic attribute information of the intelligent agent and collect the current environmental information of the disaster relief scenario in real time through multiple data sources;

[0073] Step 102: Serialize the basic attribute information and current environmental information of the agent to obtain serialized basic attribute data and serialized environmental information data, and then transmit the serialized basic attribute data and serialized environmental information data to the decision module.

[0074] Step 103: The serialized data of basic attributes and environmental information are transformed into natural language descriptions of basic attributes and environmental information through the large language model of the decision module. Based on the natural language descriptions of basic attributes and environmental information, and combined with historical decision records and behavior database, decision results are generated. The decision results include action plans and decision explanations.

[0075] Step 104: The decision result returned by the decision module is transmitted to the execution module for execution, and the decision result is recorded in the message history database.

[0076] This invention presents an intelligent agent simulation decision-making method based on a large language model, aiming to improve the decision-making efficiency and accuracy of intelligent agents in disaster relief scenarios through intelligent means. This method integrates the powerful natural language processing capabilities of a large language model, achieving effective integration, serialization processing, and intelligent decision generation of the agent's basic attributes and real-time environmental information.

[0077] Specifically, refer to Figure 2 The specific process of the intelligent agent simulation decision-making method based on a large language model in this invention is as follows:

[0078] 1. Agent decision-making begins

[0079] The decision-making process begins at this step, marking the start of the agent's preparation for making a decision. At this point, the agent prepares to collect relevant information from multiple data sources in order to formulate subsequent decisions.

[0080] 2. Basic information parameter serialization

[0081] Before making a decision, the agent needs to process its basic information, such as height, weight, gender, age, personality, health status, and education level. This information is serialized into JSON format so that the system can perform subsequent processing and analysis.

[0082] 3. Environmental information parameter serialization

[0083] The agent's environmental perception module provides various information about the current environment, including external conditions such as the occurrence of natural disasters (e.g., flood depth), the presence of rescue personnel, and ambient temperature and humidity. This information is also serialized into JSON format to ensure that the data can be effectively transmitted and processed.

[0084] 4. Serialize data and transfer it to LLM

[0085] This invention enables agents to automatically adjust their decision-making strategies based on constantly changing environments and different task requirements through Large Language Models (LLM), thereby improving the flexibility and adaptability of decision-making.

[0086] Once the basic and contextual information is serialized, it will be transmitted to a Large Language Model (LLM), such as ChatGPT or Spark Large Language Model. The LLM will process this data, transforming it into natural language descriptions. Specifically, this includes:

[0087] (1) Basic Attribute Natural Language Description: LLM transforms the basic information of the agent into easily understandable natural language descriptions. This transformation ensures that the data can be accurately interpreted through the language model, providing support for subsequent decision-making.

[0088] (2) Environmental Information Natural Language Description: Environmental information is also converted into natural language description according to a preset format. This conversion provides LLM with a clearer environmental context, helping it to fully understand the current situation.

[0089] This invention serializes information and describes it in natural language using a large language model, enabling the unified and effective processing of various heterogeneous information. At the same time, it allows various types of information (such as basic agent information, environmental information, etc.) to be transformed into easily understandable and manipulable formats, thereby improving the decision-making quality of agents in complex environments.

[0090] 5. Make decisions based on behavior databases and historical records, and explain the reasons.

[0091] After receiving natural language descriptions of basic and environmental information, LLM combines standard operating procedures, standard behavioral patterns, and historical decision records stored in its behavior database to perform decision analysis and provide reasonable action plans. LLM not only provides action decisions but also generates natural language explanations for those decisions, detailing "why this decision was made," thereby significantly enhancing the transparency and credibility of the decision-making process and improving the acceptability and controllability of the agent in human-machine collaboration. These action decisions and their explanations are recorded in a message history database for subsequent analysis and feedback.

[0092] LLMs leverage the deep language understanding and generation capabilities cultivated through training on large amounts of text data, possessing abundant common-sense knowledge, domain expertise, and factual data. Therefore, LLM-based agents typically perform exceptionally well on new tasks with only a few samples. Their superior generalization ability enables them to excel in previously unseen situations. This invention optimizes the decision-making process through large language models, enabling agents to generate decisions rapidly with lower computational costs, thus improving the efficiency of agent decision-making module development.

[0093] 6. The decision results are transmitted to the execution module.

[0094] After the decision is made, the decision generated by the LLM will be transmitted to the agent's execution module. The execution module is responsible for implementing specific actions according to the LLM's decision, such as carrying out rescue missions or adjusting behavioral strategies.

[0095] 7. End

[0096] The completion of this process marks the end of a single decision-making process for the intelligent agent. Subsequent decision-making processes may be initiated based on new environmental changes or task requirements.

[0097] In one embodiment of the present invention

[0098] The behavior library includes standard operating procedures and standard behavioral patterns;

[0099] Standard operating procedures include evacuation route selection, distribution of relief supplies, and search for emergency shelters;

[0100] Standard behavioral patterns include behavioral classification, behavioral pattern mapping for different environmental states, and behavioral decision priority rules.

[0101] The method of this invention includes a behavioral database designed for disaster relief scenarios. This database contains standard operating procedures, such as evacuation route selection, relief material distribution, and the search for emergency shelters. The core of the behavioral database lies in integrating and summarizing domain-specific behavioral patterns. Taking storm surge scenarios as an example, the behavioral database can be summarized based on research literature: by reviewing research papers in disaster management, disaster psychology, and sociology, it can summarize human behavioral patterns during storm surges.

[0102] The content of a domain-specific behavior library should cover the following aspects:

[0103] Behavior classification: Breaking down behaviors into different types, for example:

[0104] Individual behaviors: escaping alone, calling for help, etc.

[0105] Group behavior: coordinated actions (such as family members evacuating together), congestion.

[0106] Environmental factor behavior mapping: Setting corresponding behavioral patterns for different environmental states. For example:

[0107] People are more inclined to walk when the water level is below knee level.

[0108] When the water level is above waist level, people are more likely to stay put or use rescue equipment.

[0109] Decision prioritization rules: These provide priority rules for agents to choose their actions, for example:

[0110] Refugees should be prioritized for targets with high accessibility and sufficient capacity.

[0111] Prioritize avoiding dangerous areas (such as bridges that have been submerged).

[0112] By using a domain-specific behavior library, the method of this invention can improve the decision-making quality and action efficiency of intelligent agents in disaster scenarios. For example, when a flood occurs, it can quickly make decisions for humanoid intelligent agents to go to the nearest shelter or escape exit.

[0113] In one embodiment of the present invention, the method further includes:

[0114] The execution module returns the execution results to the decision module, which then performs dynamic learning and optimization of the large language model based on the execution results to ensure that the decision matches the current environment state.

[0115] This invention introduces a closed-loop feedback mechanism, in which the execution module can return the execution result to the decision module with a short update interval, enabling LLM to dynamically learn and optimize based on new data, ensuring that the decision matches the current environmental state.

[0116] In one embodiment of the present invention

[0117] The current environmental information in a disaster relief scenario includes one or more of the following: the occurrence of a natural disaster, the availability of relief resources, the availability of relief personnel, traffic conditions, geographic information, and weather and climate information.

[0118] If the disaster relief scenario is a storm surge scenario, the environmental information includes one or more of the following: water level and depth changes, traffic conditions, topographic map, population distribution, location of rescue resources, location of shelters, occupancy status of shelters, presence of rescue personnel, ambient temperature, and ambient humidity.

[0119] This invention emphasizes how to integrate and process multi-dimensional information from different sources. The current environmental information in a disaster relief scenario may include, but is not limited to, one or more of the following: the occurrence of natural disasters, the status of rescue resources, the status of rescue personnel, traffic conditions, geographic information, and weather and climate information. Examples include the processing flow of environmental monitoring data, real-time sensor data (such as water level sensors), and geographic information (such as topographic maps and building distribution). This invention employs serialization technology (such as JSON format) to standardize multi-dimensional data, enabling the effective integration of data in different formats within the system and facilitating further analysis by an LLM (Local Language Model). After serialization, this invention transforms the data into natural language descriptions, facilitating semantic understanding and comprehensive decision-making by a large language model. For example: "The current flood depth in the area is 1.2 meters, the traffic congestion index is 75%, and the nearest shelter is 500 meters away."

[0120] This invention provides more comprehensive and in-depth decision support through multi-dimensional data processing. Through dynamic data updates and real-time feedback, the intelligent agent can also quickly respond to environmental changes in disaster scenarios. If the disaster relief scenario is a storm surge, environmental information may include, but is not limited to, one or more of the following: water level changes, traffic conditions, topographic maps, population distribution, location of rescue resources, location of shelters, occupancy status of shelters, presence of rescue personnel, ambient temperature, and ambient humidity. In a storm surge scenario, the intelligent agent not only considers the current water level but also combines traffic conditions, population distribution, and the location of rescue resources to make more reasonable evacuation and rescue decisions. For example, a sensor monitors the current water level at 1.2 meters, the intelligent agent's environmental perception module shows a water depth of 1.2 meters, a traffic congestion index of 75%, and shelter A is full. This data is serialized and transmitted to the LLM (Local Management Module), then converted into a natural language description. The LLM combines historical records and a behavior database to decide whether to proceed to shelter B.

[0121] In one embodiment of the present invention, the method further includes:

[0122] The large model makes decisions periodically at preset decision time intervals, and / or determines whether to adjust the decisions based on changes in environmental information.

[0123] This invention focuses on evacuation and rescue in emergency disaster scenarios, requiring intelligent agents to respond quickly to real-time changes in environmental information (such as flood spread, shelter occupancy, etc.). Therefore, real-time performance is the core objective of the technical solution. Furthermore, the LLM agent's decision-making interval is relatively short, for example, it can be set to 5 seconds. Thus, LLM allows for frequent, real-time decision-making by the agent. Moreover, through a real-time acquisition and dynamic update mechanism of multi-dimensional information such as sensor data and geographic information, this invention enables the agent to continuously receive the latest data on environmental changes and adjust its decisions accordingly.

[0124] The present invention has the following advantages:

[0125] 1. Enhance the flexibility and adaptability of agent decision-making.

[0126] Traditional agent decision-making systems typically rely on rules or predefined models, lacking the flexibility and adaptability to cope with complex and dynamic environments. This invention, however, utilizes LLM (Limited Learning Model) to adjust the decision-making model based on real-time environmental changes, execution results, and historical data. This enables the agent to flexibly respond to changing situations and task requirements, ensuring more accurate and efficient decision-making in complex or unknown environments.

[0127] 2. Efficient integration and processing of multi-dimensional data

[0128] This invention serializes the agent's basic and environmental information into standardized data (such as JSON format) and performs natural language processing (NLP) to efficiently integrate multi-dimensional information from different sources. This approach effectively avoids the limitations of insufficient information processing or information silos in traditional methods, providing more comprehensive and accurate decision-making support. Compared with existing methods, this invention demonstrates stronger capabilities in processing complex and heterogeneous data, enabling more efficient fusion and analysis of multi-source information and improving decision-making accuracy.

[0129] 3. Transparency and explainability of the decision-making process

[0130] Traditional decision-making systems (especially deep learning-based systems) often suffer from the "black box" problem, meaning that decision outcomes lack explanation, and users cannot understand why the agent made a particular choice. In contrast, this invention provides transparency and traceability to the decision-making process through natural language explanations generated by LLM. The agent can not only provide the actions to be performed but also explain the rationale behind its decisions, making the decision-making process more credible and easier to understand. This transparency is particularly valuable in critical areas such as public safety and emergency evacuation.

[0131] 4. Improved efficiency in agent decision-making development

[0132] The decision-making process of this invention leverages the powerful computing and parallel processing capabilities of LLM to rapidly process large amounts of data and make immediate decisions. Compared to traditional rule-based systems or deep learning models that require long training times, this demonstrates higher efficiency in practical system development.

[0133] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0134] Reference Figure 3 The diagram illustrates a structural block diagram of an intelligent agent simulation decision-making device based on a large language model, provided in an embodiment of the present invention. Specifically, it may include the following modules:

[0135] The real-time information collection module 301 is used to acquire basic attribute information of the intelligent agent and collect current environmental information of the disaster relief scenario in real time through multiple data sources.

[0136] The serialization processing module 302 is used to serialize the basic attribute information and current environmental information of the agent to obtain serialized basic attribute data and serialized environmental information data, and then transmit the serialized basic attribute data and serialized environmental information data to the decision module.

[0137] The decision result generation module 303 is used to convert the serialized data of basic attributes and the serialized data of environmental information into natural language descriptions of basic attributes and environmental information through the large language model of the decision module, and generate decision results based on the natural language descriptions of basic attributes and environmental information, combined with historical decision records and behavior databases; the decision results include action plans and decision explanations.

[0138] The decision result transmission and recording module 304 is used to transmit the decision result returned by the decision module to the execution module for execution, and record the decision result in the message history database.

[0139] Optionally,

[0140] The behavior library includes standard operating procedures and standard behavioral patterns;

[0141] Standard operating procedures include evacuation route selection, distribution of relief supplies, and search for emergency shelters;

[0142] Standard behavioral patterns include behavioral classification, behavioral pattern mapping for different environmental states, and behavioral decision priority rules.

[0143] Optionally, the device further includes:

[0144] The decision model dynamic update module is used by the execution module to return the execution results to the decision module. The decision module dynamically learns and optimizes the large language model based on the execution results to ensure that the decision matches the current environment state.

[0145] Optionally,

[0146] The current environmental information in a disaster relief scenario includes one or more of the following: the occurrence of a natural disaster, the availability of relief resources, the availability of relief personnel, traffic conditions, geographic information, and weather and climate information.

[0147] If the disaster relief scenario is a storm surge scenario, the environmental information includes one or more of the following: water level and depth changes, traffic conditions, topographic map, population distribution, location of rescue resources, location of shelters, occupancy status of shelters, presence of rescue personnel, ambient temperature, and ambient humidity.

[0148] Optionally,

[0149] Historical decision records are stored in the message history database, including action plans and decision explanations.

[0150] Optionally, the device further includes:

[0151] The timed dynamic decision-making module is used for the large model to make decisions periodically according to a preset decision time interval, and / or to determine whether to adjust the decisions based on changes in environmental information.

[0152] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0153] In addition, embodiments of the present invention also provide an electronic device, such as... Figure 4 As shown, it includes a processor 401, a communication interface 402, a memory 403, and a communication bus 404, wherein the processor 401, the communication interface 402, and the memory 403 communicate with each other through the communication bus 404.

[0154] Memory 403 is used to store computer programs;

[0155] When the processor 401 executes the program stored in the memory 403, it implements the intelligent agent simulation decision-making method based on a large language model as described in the above embodiments.

[0156] The communication bus mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.

[0157] The communication interface is used for communication between the aforementioned terminal and other devices.

[0158] The memory may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0159] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0160] like Figure 5 As shown, in another embodiment of the present invention, a computer-readable storage medium 501 is also provided, which stores instructions that, when executed on a computer, cause the computer to execute the intelligent agent simulation decision-making method based on a large language model as described in the above embodiments.

[0161] In another embodiment of the present invention, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute the intelligent agent simulation decision-making method based on a large language model as described in the above embodiments.

[0162] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0163] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0164] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the apparatus embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0165] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of protection of the present invention.

Claims

1. A simulation decision-making method for intelligent agents based on a large language model, characterized in that, The method includes: Acquire basic attribute information of the intelligent agent and collect real-time environmental information of the disaster relief scenario through multiple data sources; The basic attribute information and current environmental information of the intelligent agent are serialized to obtain serialized data of basic attributes and serialized data of environmental information, and then the serialized data of basic attributes and serialized data of environmental information are transmitted to the decision module. The decision module uses a large language model to transform serialized data of basic attributes and environmental information into natural language descriptions of basic attributes and environmental information, respectively. Based on these descriptions, and combined with historical decision records and a behavior database, decision results are generated, including action plans and decision explanations. The behavior library includes standard operating procedures and standard behavior patterns. Standard operating procedures include evacuation route selection, distribution of relief supplies, and search for emergency shelters; Standard behavioral patterns include behavioral classification, behavioral pattern mapping for different environmental states, and behavioral decision priority rules; Behavioral categories include individual behavior and group behavior. Individual behaviors include: escaping alone and calling for help; group behaviors include: family members evacuating together and congestion. Behavioral pattern mappings for different environmental conditions include: when the water level is below the knee, people tend to walk; when the water level is above the waist, people tend to stay put or use rescue equipment. Historical decision records are stored in the message history database, including action plans and decision explanations; The decision results returned by the decision module are transmitted to the execution module for execution, and the decision results are recorded in the message history database. The execution module returns the execution results to the decision module, which then performs dynamic learning and optimization of the large language model based on the execution results to ensure that the decision matches the current environment state.

2. The method according to claim 1, characterized in that, The current environmental information in a disaster relief scenario includes one or more of the following: the occurrence of a natural disaster, the availability of relief resources, the availability of relief personnel, traffic conditions, geographic information, and weather and climate information. If the disaster relief scenario is a storm surge scenario, the environmental information includes one or more of the following: water level and depth changes, traffic conditions, topographic map, population distribution, location of rescue resources, location of shelters, occupancy status of shelters, presence of rescue personnel, ambient temperature, and ambient humidity.

3. The method according to claim 1, characterized in that, The method further includes: The decision-making module's large language model makes decisions periodically at preset decision time intervals, and / or determines whether to adjust the decisions based on changes in environmental information.

4. A smart agent simulation decision-making device based on a large language model, characterized in that, The device includes: The real-time information collection module is used to acquire basic attribute information of the intelligent agent and collect current environmental information of the disaster relief scenario in real time through multiple data sources; The serialization processing module is used to serialize the basic attribute information and current environmental information of the agent to obtain serialized basic attribute data and serialized environmental information data, and then transmit the serialized basic attribute data and serialized environmental information data to the decision module. The decision result generation module is used to transform the serialized data of basic attributes and environmental information into natural language descriptions of basic attributes and environmental information through the large language model of the decision module. Based on the natural language descriptions of basic attributes and environmental information, and combined with historical decision records and behavior database, the module generates decision results. The decision results include action plans and decision explanations. The behavior library includes standard operating procedures and standard behavior patterns; Standard operating procedures include evacuation route selection, distribution of relief supplies, and search for emergency shelters; Standard behavioral patterns include behavioral classification, behavioral pattern mapping for different environmental states, and behavioral decision priority rules; Behavioral categories include individual behavior and group behavior. Individual behaviors include: escaping alone and calling for help; group behaviors include: family members evacuating together and congestion. Behavioral pattern mappings for different environmental conditions include: when the water level is below the knee, people tend to walk; when the water level is above the waist, people tend to stay put or use rescue equipment. Historical decision records are stored in the message history database, including action plans and decision explanations; The decision result transmission and recording module is used to transmit the decision result returned by the decision module to the execution module for execution, and record the decision result in the message history database; The decision model dynamic update module is used by the execution module to return the execution results to the decision module. The decision module dynamically learns and optimizes the large language model based on the execution results to ensure that the decision matches the current environment state.

5. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; The memory is used to store computer programs; When the processor executes the program stored in the memory, it implements the intelligent agent simulation decision-making method based on a large language model as described in any one of claims 1-3.

6. One or more computer-readable media having instructions stored thereon that, when executed by one or more processors, cause the processors to perform the agent simulation decision-making method based on a large language model as described in any one of claims 1-3.

Images