KR102571419B1 - Disaster response system through artificial intelligence-based planning and decision-making and method using the same - Google Patents

Abstract

The present invention relates to a disaster response system through artificial intelligence-based planning and decision-making and a method using the same. To create a disaster initial state and disaster response target state, and a disaster plan setting unit that manually or automatically programs domain files and problem files, and automatically creates disaster response plans through AI planners according to the initial conditions of disaster occurrence and disaster response goals Disaster plan action creation unit, disaster plan action execution unit that performs the generated unit actions, disaster situation and response status are visually displayed, and disaster execution application unit and disaster occurrence information for reporting, supervision and instruction on disaster situations are provided. It relates to a disaster response system through artificial intelligence-based planning and decision-making, characterized in that it includes a knowledge base to store, and a method using the same.

Description

Disaster response system through artificial intelligence-based planning and decision-making and method using the same}

The present invention relates to a system for generating an optimal disaster response scenario for a disaster situation using artificial intelligence (AI)-based decision support technology and a method for using the same, and more particularly, based on a disaster situation and crisis response manual. It is about a disaster response system through artificial intelligence-based planning and decision-making that collects information to cope with disasters, and based on this, AI planner automatically creates and executes a plan to cope with disaster situations and how to use it .

Government agencies, public institutions, and general corporations have a crisis response manual that specifically describes the procedures and methods to respond most efficiently in case of a disaster, and through this, regular or irregular disaster response training is being carried out.

However, these manuals are prepared by predicting possible disaster situations in advance, but when an unexpected situation occurs or the disaster intensity is greater than expected, a problem that cannot be effectively dealt with only by responding according to the manual may occur.

In addition, even though highly trained disaster managers tend to lean toward instinctive and emotional responses rather than rational responses when faced with an emergency situation, they perform tasks and actions other than those specified in the manual, such as identifying the situation individually. As a result, they cannot respond quickly to disasters or disasters, so they often miss the golden time. As a result, disaster response cannot be performed smoothly, resulting in additional damage such as increased human and material damage.

For example, when unexpected disaster situations occur, such as the Sewol Ferry incident or the Yongsan Itaewon incident, most of the disaster response managers, regardless of high-ranking or low-level positions, fail to implement the optimal method specified in the manual, and even the manual clearly If it is not presented, problems that cannot effectively cope with disaster situations may occur.

In order to solve these problems, a decision support technology (IDSS, Intelligent Decision Support System) has recently been developed through artificial intelligence, and there is a tendency to replace the role of a disaster response control tower through this.

However, specific implementation means using IDSS have not been clearly presented, and a platform capable of efficiently coping with various disaster situations has not yet been presented.

Therefore, when a disaster occurs, a platform that can automatically create and execute optimal disaster response scenarios and plans faster than humans using decision-making support technology based on artificial intelligence is required, and based on this, a national It is necessary to minimize damage by acting as a control tower to prevent and effectively respond to disasters.

Patent Document 1 provides a system that collects as much information as possible on the type of disaster and the surrounding situation in a disaster situation and intelligently searches for and provides a shelter for users to evacuate, or simply provides an active means of action to automatically respond to a disaster. It does not provide, but merely suggests a place to evacuate.

Patent Document 2 is characterized by providing a user-customized response manual as well as a universal response manual in the event of a disaster to provide customized response measures to the facility or site actually managed by the user, but simply presenting only the disaster manual. It is unable to present concrete and practical unit actions (actions) that can be dealt with.

Korean Patent Publication No. 2021-0064579 Korean Patent Registration No. 10-2490062

In order to solve the above problems, in the present invention, an artificial intelligence planner automatically generates disaster response plans to be performed by members of organizations or institutions to cope with disaster situations and assists in performing disaster tasks. Its purpose is to provide a disaster response system through intelligence-based planning and decision-making and a method using the same.

In addition, in the present invention, for unit actions generated in the AI plan, unit tasks are instructed through the portable terminal of individual members or whether or not the instructed unit tasks are performed, and by synthesizing them and providing necessary next-step task instructions, etc. The purpose of this study is to provide a disaster response system through artificial intelligence-based planning and decision-making that can systematically respond to situations and a method using the same.

Disaster response system through artificial intelligence-based planning and decision-making according to the present invention for achieving the above object is an information disaster information collection unit that collects information about the situation in which a disaster occurs, and a plan to cope with a disaster situation Disaster plan setting unit for generating disaster initial conditions and disaster response goals, disaster plan action generation unit (300 ), a disaster plan action execution unit that performs unit actions created in ), a disaster execution application unit that visually displays disaster situations and response status, and reports, supervises, and instructs disaster situations, and a knowledge base that stores disaster occurrence information. It is characterized in that it includes.

In addition, in the disaster response system through artificial intelligence-based planning and decision-making according to the present invention, the disaster information collection unit collects situation information, and the disaster area, disaster type, and emergency level from the collected situation information. It is characterized in that it includes a disaster situation information analysis unit that analyzes the disaster situation through a disaster situation feature extraction unit and parameters to extract parameters to be determined.

In addition, in the disaster response system through artificial intelligence-based planning and decision-making according to the present invention, the disaster plan setting unit prerequisites for manually or automatically responding to a disaster according to sensor information or optional query words for prioritization It is characterized by including an input unit, an initial state defining situation information and a disaster plan generation unit for setting disaster response goals, and a PDDL file generation unit for generating domain files and problem files.

In addition, in the disaster response system through artificial intelligence-based planning and decision-making according to the present invention, the domain file 231 and the problem file 232 are written in PDDL language.

In addition, in the disaster response system through artificial intelligence-based planning and decision-making according to the present invention, the problem file includes unit action functions describing preconditions and effects for coping with disaster situations.

In addition, in the disaster response system through artificial intelligence-based planning and decision-making according to the present invention, the disaster plan action execution unit executes the unit actions automatically generated by the AI planner and the unit action is performed to respond to disasters. If the execution is successful after instructing the portable terminal or robot of the member who is involved in the process, the next action continues in a way that the system continuously instructs the person, and if it fails, the action re-execution unit that requests re-planning to the AI planner It is characterized by including.

In addition, in the disaster response system through artificial intelligence-based planning and decision-making according to the present invention, the disaster execution application unit visually displays the performance and progress of the unit action, a disaster situation information visualization unit, whether unit actions are performed, and The disaster management department summarizes and reports the report to report on-site information and crisis response situation information to the reporting line of the disaster management manager by synthesizing the results, and briefing or broadcasting the disaster situation based on the report data generated by the disaster management department. It is characterized in that it includes a disaster situation information provision unit provided to the media.

In addition, the disaster response method through artificial intelligence-based planning and decision-making according to the present invention includes the first step of collecting and analyzing disaster situation information in the disaster information collection unit 100, setting the initial state and target state of disaster occurrence, and PDDL 2nd step of generating domain files and problem files, 3rd step of automatically generating unit actions that operate plans in response to disasters by passing the generated PDDL file to the AI planner, and generating unit actions through the AI planner It is characterized in that it includes a fourth step of executing and a fifth step of determining whether to perform a unit action.

In addition, in the disaster response method through artificial intelligence-based planning and decision-making according to the present invention, when the execution of the unit action is successfully completed after the fifth step, the step of calculating the input cost according to the execution of the unit action is further included. and, when the unit action fails, further comprising requesting re-planning through the AI planner.

In addition, in the disaster response method through artificial intelligence-based planning and decision-making according to the present invention, it is characterized in that it further comprises the step of transmitting the performance results of the unit actions to the disaster execution application unit after the fifth step.

As described above, according to the disaster response system and method using the same through artificial intelligence-based planning and decision-making according to the present invention, accident site information of disaster response is delivered to the relevant agency and the final decision-maker in real time, and after the disaster The advantage is that it can serve as a control tower where key decisions on decision-making and resource allocation are automatically generated and executed within the initial golden time.

In addition, according to the disaster response system and method using artificial intelligence-based planning and decision-making according to the present invention, rescuers necessary at the scene of an accident can perform important tasks such as evacuation, finding a group with a possibility of survival, and optimizing response and recovery options. It has the advantage of being able to quickly solve problems that would otherwise consume a significant amount of work.

In addition, according to the disaster response system and method using the artificial intelligence-based planning and decision-making according to the present invention, there is an advantage in that the resilience of society can be increased by reducing the cost of disaster interruption borne by the government, businesses and citizens.

In addition, according to the disaster response system and method using the artificial intelligence-based planning and decision-making according to the present invention, the location of the disaster can be visually displayed accurately in real time, so that the user can more quickly understand the distribution and predicted route of the disaster. There are advantages to being able to

In addition, according to the disaster response system and method using artificial intelligence-based planning and decision-making according to the present invention, installation manuals and maintenance manuals for fault diagnosis and repair of complex systems such as semiconductor factories and aircraft maintenance are provided for crisis response If it is used in the same principle as a manual, it has the advantage of solving complicated steps that may occur during installation or system maintenance and showing them to workers as unit work steps and helping them.

1 is a functional block diagram showing overall functions of a disaster response system through artificial intelligence-based planning and decision-making according to a preferred embodiment of the present invention.
Figure 2 is a functional block diagram showing the function of the disaster information collection unit according to a preferred embodiment of the present invention.
3 is a functional block diagram showing the functions of a disaster plan setting unit according to a preferred embodiment of the present invention.
4 and 5 are diagrams illustrating domain files generated by the disaster plan setting unit according to a preferred embodiment of the present invention.
6 is a diagram showing a disaster plan generated by a disaster plan action generation unit according to a preferred embodiment of the present invention in codes.
7 is a functional block diagram showing the functions of a disaster plan action execution unit according to a preferred embodiment of the present invention.
8 is a functional block diagram showing the functions of a disaster execution application unit according to a preferred embodiment of the present invention.
9 is a flowchart illustrating a disaster response method through artificial intelligence-based planning and decision-making according to a preferred embodiment of the present invention.

Hereinafter, an embodiment in which a person skilled in the art can easily practice the present invention will be described in detail with reference to the accompanying drawings. However, in the detailed description of the operating principle of the preferred embodiment of the present invention, if it is determined that a detailed description of a related known function or configuration may unnecessarily obscure the subject matter of the present invention, the detailed description will be omitted.

In addition, the same reference numerals are used for parts having similar functions and actions throughout the drawings. Throughout the specification, when a part is said to be connected to another part, this includes not only the case where it is directly connected, but also the case where it is indirectly connected with another element interposed therebetween. In addition, including a certain component does not exclude other components unless otherwise stated, but means that other components may be further included.

Hereinafter, a disaster response system through artificial intelligence-based planning and decision-making according to the present invention and a method using the same will be described with reference to the accompanying drawings.

First, the types of disasters are very diverse, such as earthquakes, fires, typhoons, and floods, and there are various ways to cope with disasters according to the degree or grade of emergency and the area or time of occurrence.

Responses are carried out according to a disaster response manual prepared in advance, but a control tower must be organized in advance so that instructions or a command system according to the manual are possible, so a quick control tower configuration is very important.

Here, a series of scenarios for instructions or commands are defined as a plan in the present invention, and actions or operations for constructing and executing these plans are expressed as a plurality of unit actions in a time-series sequential or parallel type. do. In particular, it is assumed that the performing human member or robot is composed of multiple entities, not a single entity, and these multiple members create and perform unit actions required to perform necessary actions simultaneously and in parallel in different time and space.

1 is a functional block diagram showing overall functions of a disaster response system through artificial intelligence-based planning and decision-making according to a preferred embodiment of the present invention.

As shown in FIG. 1, the disaster response system through artificial intelligence-based planning and decision-making includes a disaster information collection unit 100, a disaster plan setting unit 200, a disaster plan action generation unit 300, and a disaster plan action It is composed of an execution unit 400, a disaster execution application unit 500, an AI planner 600, and a knowledge base 700.

First, the disaster information collection unit 100 collects various and comprehensive information from various IoT sensors to recognize a situation in which a disaster has occurred, and can use this as basic information for establishing a plan to respond to a disaster. .

The disaster plan setting unit 200 creates an initial disaster state and a disaster response target in order to establish a plan for coping with a disaster situation. In addition, it performs a function of building a knowledge-based framework that can automatically create a goal-oriented work flow, and a function of establishing a disaster response plan based on a disaster response manual can be programmed by a programmer.

The disaster plan action generation unit 300 may automatically generate disaster response plan actions to be automatically performed by the AI planner 600 according to the initial conditions and set goals set by the disaster plan setting unit 200.

Here, the AI planner 600 is a type of software or algorithm used to create a plan or schedule for a given series of tasks or goals, generally utilizing artificial intelligence techniques such as search algorithms, constraint satisfaction, and optimization methods. to create an efficient and effective plan.

In addition, the AI planner 600 is a "constraint-based planner", "constraint satisfaction problem solver", or a reinforcement learning engine, which creates a plan by satisfying a series of constraints or rules. In the present invention, the AI planner 600 A description of an algorithm describing a specific operation is a tool that can be easily changed and adopted at the level of a person skilled in the art, so a detailed description thereof will be omitted.

Again in FIG. 1 , the disaster plan action execution unit 400 performs a function of performing unit actions generated by the disaster plan action generation unit 300 .

In addition, the disaster execution application unit 500 provides a monitoring means for visually displaying the disaster situation and response status, and provides communication media such as disaster work for reporting, supervision and instruction on the disaster situation, and briefing or broadcasting on the disaster situation. It performs the notification function through

The knowledge base 700 serves to store collected disaster information, disaster response manuals, disaster-specific decision-making rules, disaster-specific required resources, geographical information, and the like.

The AI planner 600 provides an optimal algorithm for solving one problem under given constraints, and presents an optimal solution by using the knowledge base accumulated by regularizing the experience and conditions for the problem for search decision-making. do.

2 is a functional block diagram showing the functions of the disaster information collection unit 100 according to a preferred embodiment of the present invention.

Referring to FIG. 2, the disaster information collection unit 100 collects all information related to disasters from various IoT sensors that measure human verbal reports, telephone reports, visual sensors such as CCTV, sound, vibration, gas, and air quality, and Disaster situation information collection unit 110, which is comprehensively collected through various media, disaster feature extraction that extracts parameters for determining the disaster area, disaster type, emergency situation, etc. from disaster situation information collected to establish a disaster response plan It includes a disaster situation information analysis unit 130 that comprehensively analyzes the disaster situation through the unit 120 and the extracted features.

The disaster information collection unit 110 may collect disaster information by accessing a disaster information providing system operated by a public institution. At this time, the disaster information may be provided using a public API or a private API provided by the disaster information providing system. there is. In addition, real-time weather information and disaster information that change frequently can be collected through disaster text messages or broadcasting. Additionally, the disaster situation information collection unit 110 may collect static information such as a disaster response manual, a government organization table, an educational institution organization table, school, fire station, and hospital related information.

The disaster feature extraction unit 120 maps the disaster manual stored in the knowledge base 700 and the collected disaster information to extract various parameters required for the execution of the disaster manual, and allows the disaster situation information collection unit 110 to add additional parameters as needed. You may be directed to collect information.

The disaster situation information analysis unit 130 may set resources to be dealt with according to the degree of the disaster situation, and present the analyzed results as reference data when the disaster plan setting unit 200 sets up a disaster plan. In addition, it can perform roles such as searching for disaster response manuals, selecting related institution information, disaster information delivery system, and disaster work personnel according to the manual, and the analyzed knowledge data is in the form of an ontology in the knowledge base (700). storage is used.

For example, in the event of an injury due to an earthquake or fire, the information from the Korea Meteorological Administration, such as the strength of the earthquake and predicted aftershock information, and disaster site information such as the state of collapse of the building, the number of injured people and the region, etc. It collects and stores information on dispatchable fire trucks, firefighters, ambulances, and nearby hospitals, and analyzes them according to the degree of disaster occurrence to suggest utilization plans.

3 is a functional block diagram showing the functions of the disaster plan setting unit 200 according to a preferred embodiment of the present invention, and FIGS. 4 and 5 are generated by the disaster plan generation unit according to a preferred embodiment of the present invention. It is a diagram illustrating a domain file.

First, prior to generating disaster plan actions generated through the AI planner 600 using the domain and problem files generated by the disaster plan setting unit 200, a series of general time-series disaster response procedures will be described.

For example, a patient occurs due to a disaster in a specific area, and a detailed procedure according to a series of manuals for transporting the patient to a hospital may be configured as follows.

Here, in the conventional method, the agent is a person in charge or in charge of disaster management of the control tower for convenience, but in the present invention, it means an artificial intelligence planner and decision-making system. Although the overall disaster response means is performed through these agents, the procedure below is simply an example of a basic procedure and is not limited thereto as it can be changed to various procedures at the level of a skilled technician.

1) Check access (School manager's mobile device -> Agent access)

2) Report (find patients at evacuation site, number of patients, patient condition, ambulance request)

3) Wait for response

4) Connection confirmation (agent -> 119 fire department access)

5) Transmission of information (find patients at the evacuation site, number of patients, patient condition, 119 ambulance request)

6) dispatch request

7) Dispatch report (ambulance dispatch, expected arrival time)

8) Transmission of information (firefighter, ambulance, destination information)

9) Dispatching report (firefighters, school administrators).

10) Dispatch arrival report (firefighter, agent)

11) Initiate patient transfer (firefighter, agent),

12) Check access (agent -> emergency hospital manager)

13) Notification during patient transfer (agent -> emergency hospital manager)

14) Receipt of patient acceptance

15) Notification of expected patient arrival time

16) Notification of patient arrival

17) End of connection

However, when various factors such as a large number of patients occur, some bridges and roads in the disaster area are damaged, vehicles cannot be operated, or hospital beds are insufficient, it is not possible to respond quickly with only the basic manual as described above. When a complex situation occurs in which the control tower cannot perform its role, rapid response becomes impossible, and even highly skilled experts cannot comprehensively judge the disaster situation and give instructions, establishing a plan based on artificial intelligence planning and decision-making. this becomes inevitable

Referring back to FIG. 3, the disaster plan setting unit 200 performs a function of building a knowledge-based framework capable of manually or automatically synthesizing a goal-oriented workflow, and plans a disaster response based on a disaster response manual. It may be composed of a query input unit 210, a disaster plan generator 220, and a PDDL file generator 230 that perform a function of establishing a.

Here, Planning Domain Definition Language (PDDL) is one standardized language defined for use in artificial intelligence to describe various aspects of plan problems, designed to represent plan domains in a way that is independent of any particular plan algorithm, so that different plans It allows designers to reason about domains in a consistent way. Also, PDDL consists of two main components, domain file and problem file. The domain file defines the actions that can be performed and the conditions that can be performed in the plan problem, and the problem file specifies the initial state, the target state to be achieved, the preconditions of each action function, and the constraints and target conditions to be satisfied. Through this, the plan designer does programming that presents “what to do” in a more declarative way about the problem, which is different from “what steps to do” in the conventional procedural method, which is a general programming method. When a domain and problem file are given in a declarative programming method, a wide range of search algorithms can be applied inside the planner engine to find the optimal combination of a series of unit actions.

First, the query input unit 210 may be selectively used or omitted to establish a response target for a disaster situation. A person manually giving a query is to make disaster response goal setting clearer. As a means of receiving text and audio/video signals for setting disaster response goals, it is an essential initial procedure for artificial intelligence-based decision-making. In the case of a fire, it can be established at the initial inquiry stage whether the goal is to extinguish or prevent spread, or to prioritize saving lives. If the manual query input is omitted, the AI agent of the control tower automatically judges the situation information and presents the goal setting, and the programmer can refer to it for programming.

The disaster plan generation unit 220 includes an initial state setting unit 221 defining information on a situation in which a disaster has occurred and a final goal state setting unit ( 222).

In particular, the initial state setting unit 221 determines the initial state in which a disaster has occurred, which is used as very important information for establishing a disaster plan. For example, when there are injured people due to an earthquake, it is possible to establish a specific plan based on this only when the number or location of the injured people has to be confirmed and set in advance.

After collecting disaster situation information, extracting and analyzing features, setting the initial state and target state using this basically requires a programmer to manually program, but it can be semi-automated with the help of the knowledge base 700. For example, it can be deduced from the knowledge base 700 that an earthquake requires emergency transportation (target state) to a nearby emergency hospital in case an injured person occurs (initial state) due to an earthquake.

In addition, the PDDL file generation unit 230 is a method in which a human programmer manually creates a domain file and a problem file according to the planning and decision-making conditions set by the disaster plan generation unit 220 through programming. generate

4 and 5 are diagrams illustrating domain files generated by the disaster plan setting unit according to a preferred embodiment of the present invention.

4 and 5, in the domain file described in PDDL, as an example of describing the type of object according to the disaster situation, the state of the predicates of the proposition, and the action, the injured person in the school due to the earthquake It shows the case of disaster response that occurs and transports it to the hospital.

Referring to FIG. 4, when a disaster situation occurs in a school, the types to cope with the disaster include AI agents, resources (school, hostipal, firestation, ambulance, bed), school information (school_info), patient information (patients_info), Numerical information such as ambulance-related information (ambu_arrival_status, ambulance_worker, emergency_patient), hospital information, school-hospital, fire station-school-hospital distance, and transfer time can all be described, and AI agent’s It describes the current state information of ambulances and hospitals as a means of coping with whether or not they are in operation, the occurrence of patients, and occurrences.

These actions may be defined in various forms according to the type of disaster, which may be defined in advance in the knowledge base 700 .

In one embodiment of the present invention, 10 actions can be provided as follows for a disaster situation, especially a coping situation related to patient transfer, and each action is an input action precondition and an effect indicating an output action ) is structured.

- check_connection: communication connection

- send_information: send information

- start_ambulance : Ambulance departure request

- report_ambulance_arrival : Ambulance arrival report

- load_ambulance : Ambulance ride

- move_ambulance ambulance movement

- unload_ambulance Unload the ambulance

- go_to_bed : Transfer to bed

- move_back_ambulance : Ambulance regression

- transfer_end : end

5 is an example of an action for 'start_ambulance' of one of the above actions, and shows definitions of related parameters, preconditions, and effects.

The problem file 231 generated by the PDDL file generation unit 230 according to a preferred embodiment of the present invention will be described in detail below.

For example, the initial situation set in the initial state setting unit 221 is defined as "4 emergency victims, in the affected area, 2 firefighters required, 4 transport vehicles required, reservation at an emergency hospital", and the target state Assuming that the goal set by the setting unit 222 is additionally set to "evacuation of 4 emergency victims to an emergency hospital" or "to arrive at the nearest emergency hospital or in the shortest time", based on this The domain file 231 generated by the PDDL file generation unit 230 may be generated as follows.

(define (problem safety-pb2)

(:domain safety)

(:requirements :strips :typing :negative-preconditions)

(:objects

ai_agent - AI_agent

si-school_info

cs - connection_status

pi - patients_info

aas - ambu_arrival_status

school1 school2 - school

hospital1 hospital2 - hospital

firestation1 firestation2 - firestation

ambu1_1 ambu1_2 ambu1_3 ambu1_4 - ambulance

;ambu_worker1_1 ambu_worker1_2 ambu_worker1_3 ambu_worker1_4

bed1_1 bed1_2 bed1_3 bed1_4 - bed

patient1 patient2 patient3 patient4 - emergency_patient

)

(:init

;patient1~2 - school1 (2 patients)

;patient3~4 - school2 (2 patients)

;-> total patients: 4

(have ai_agent si)

(not (have ai_agent cs))

(not (have ai_agent aas))

(not (have ai_agent pi))

;patient1~2- school1

(in patient1 school1)

(in patient2 school1)

;patient3~4 - school2

(in patient3 school2)

(in patient4 school2)

;hospital1 - bed1_1 to 1_4

(enough bed1_1 hospital1)

(enough bed1_2 hospital1)

;hospital2 - bed2_1 to 2_4

(enough bed2_1 hospital2)

(enough bed2_2 hospital2)

(enough bed2_3 hospital2)

(enough bed2_4 hospital2)

;firestation1 - ambulance 3

;firestation2 - ambulance 4

;firestation3 - ambulance 4

;firestation4 - ambulance 4

; -> total ambulances: 15

(at ambu1_1 firestation1)

(at ambu1_2 firestation1)

(at ambu1_3 firestation1)

(at ambu2_1 firestation2)

(empty ambu4_1)

(empty ambu4_2)

(empty ambu4_3)

(empty ambu4_4)

)

(:goal

;init>

;patient1~2 - school1 (2 patients)

;patient3~4 - school2 (2 patients)

;-> total patients: 4

;hospital1 - bed1_1 ~ 1_2 (2 beds)

;hospital2 - bed2_1 ~ 2_2 (2 beds)

; -> total beds: 4

)

)

The problem file above describes the objects, init state, and goal state according to the disaster situation.

For example, with initial information, there are two patients (patient1, patient2) in school1 (school1), two patients (patient3, patient4) in school2 (school2), and four ambulances are ready. am. Based on this, the disaster response goal is to transfer these patients by ambulance to four beds (bed1 to bed4) of hospitals 1 and 2 (hospital1, hospital1).

6 is a diagram showing a disaster plan automatically generated by the disaster plan action generation unit 300 according to a preferred embodiment of the present invention in codes.

The code shown in FIG. 6 is a code automatically generated by the AI planner 600 using the basic data of disaster response defined in the domain file 231 and the problem file 232 generated by the action plan setting unit 200. , Multiple patients (85 people), disaster schools (5 locations), hospitals (4 locations), ambulances (15 units) are assumed, and the optimal unit action code is automatically generated according to the initial state and goal setting. .

In FIG. 6, the AI agent corresponds to the disaster plan action execution unit 400, and the communication connection state (eg, call connection with the school manager) for checking the disaster situation in the school in the disaster plan action execution unit 400 A series of procedures, such as an action to check availability), transmission of information for transport to the school where the patient occurred and an ambulance that can be dispatched, requesting departure of the ambulance and confirming its arrival, for subsequent transportation according to the occurrence of the patient, are automatically can be created and executed. In the AI planner 600, even though a plurality of patients have occurred, it is possible to proceed with an optimal transfer procedure using the given resources as much as possible, and this can be performed instead of a problem in which the disaster manager cannot act after immediately determining the situation in the field.

7 is a block diagram showing the functions of the disaster plan action execution unit 400 according to a preferred embodiment of the present invention.

Referring to FIG. 7 , the disaster plan action execution unit 400 is a function of performing unit actions generated by the disaster plan action generation unit 300 and is composed of an action execution unit 410 and an action re-execution unit 420 .

The action executor 410 is a component that performs a unit action, and one basic unit action means one action to be performed by a person. The start command is given to the portable terminal of the performer to be performed, and upon completion of the action One unit action ends by pressing the finish button. In addition, if one unit action succeeds, it is possible to comprehensively evaluate later whether the entire action operates in an optimal state in the final stage by calculating the corresponding cost and required time.

For example, in the case of 'start_ambulance' action, a program commanding departure to a specific moving place with the cell phone number of the corresponding ambulance driver can be operated.

In addition, each unit action may be executed in parallel as needed by the action execution unit 410 .

The action re-execution unit 420 may request an additional action for re-execution by determining whether the unit action to be performed is an appropriate action or action or whether the corresponding action has been properly performed. In addition, when it is impossible to normally perform the unit action generated for a disaster situation that changes from time to time or a change is necessary, the AI planner 600 can immediately be instructed to modify the unit action by changing the unit action, and through this, the modified unit action is re-executed. can do.

For example, the execution of the 'start_ambulance' action was commanded, but in the planning stage, there was no problem because the ambulance was in a standby state. Follow-up actions can be taken with the action to change and request departure.

8 is a functional block diagram showing the functions of the emergency execution application unit 500 according to a preferred embodiment of the present invention.

As shown in FIG. 8, the disaster execution application unit 500 includes a disaster situation information visualization unit 510 for visually displaying the disaster situation and response status, and a disaster management unit 520 for reporting, supervising and instructing the disaster situation. ) and a disaster situation information providing unit 530 that notifies briefings and broadcasts about disaster situations through communication media.

The disaster situation information visualization unit 510 can visually display the disaster situation and coping status and location information of the unit actions performed by the disaster plan action execution unit 400 and the progress status in real time to the disaster manager accurately. In addition, location and movement status information can be mapped to a geographic information system and displayed.

The disaster management unit 520 summarizes a report to report on-site information and crisis response situation information to the reporting line of the disaster management manager by synthesizing the unit actions being performed by the disaster plan action execution unit 400 and the results. Performs a writing and reporting role. Here, a detailed description of the specific embodiment of the report creation is omitted because it can be easily configured at the level of a person skilled in the art.

The disaster situation information providing unit 530 performs a function of comprehensively briefing the disaster situation based on the report generated by the disaster management unit 520 or providing basic data for real-time disaster service to newspapers or broadcasting media.

9 is a flowchart illustrating a method of using a disaster response system through artificial intelligence-based planning and decision-making according to a preferred embodiment of the present invention.

Referring to FIG. 9, the disaster response method through artificial intelligence-based planning and decision-making includes the first step of collecting and analyzing disaster situation information in the disaster information collection unit 100, setting the initial state and target state of disaster occurrence, and The second step of programming and generating PDDL domain files and problem files, delivering the generated PDDL files to the AI planner 600, and automatically generating unit actions that operate plans in response to disasters through the AI planner 600 It consists of a third step of executing the generated unit action, a fourth step of executing the generated unit action, and a fifth step of determining whether or not the unit action is performed.

In the first step, the AI planner 600 for disaster response is prepared, inputting disaster information to be processed, and extracting features for collecting and analyzing various disaster information reported in disaster occurrence areas. In addition, for analysis, relatively static information such as disaster response manuals, government organization tables, educational institution organization tables, and student records can be collected and stored in the knowledge base 700 in the form of an ontology.

In the second step, constraint conditions for the overall disaster situation are input through a separate query procedure for disaster response, and based on this, the programmer sets the initial state and target state, and programs the domain file 231 and the problem file 232 to create

Here, the unit actions described in the problem file may be stored in the knowledge base 700 in advance in various forms according to the type and grade of each disaster, and the programmer may select and configure them by referring to them during programming, if necessary. Alternatively, such programming can be prepared in advance and stored in the knowledge base, and necessary parts can be automatically generated and configured in case of emergency.

In the third step, a series of unit actions constituting a disaster response plan are generated through the AI planner 600. At this time, the AI planner 600 specifies constraint conditions for selecting unit actions, and optimally determines the search tree for the search tree. If a heuristic function that can be searched can be presented and an objective function for generating an optimal goal is defined, the AI planner 600 itself performs a procedure for finding an optimal solution at a minimum cost and outputs an optimal solution. . For example, it is possible to find an optimal choice through a situation in which a patient is to be transferred with priority by integrating the patient's emergency situation, the location of an ambulance and a hospital.

In the fourth and fifth steps, unit actions generated by the disaster plan action execution unit 400 are executed, and actions capable of sequential processing or independent parallel processing are executed. In addition, it is possible to determine whether to execute the action of the next step by checking the success of each executed action.

In addition, in the fifth step, if the unit action is successfully completed, the step of calculating the cost according to the action is further included, and if the unit action fails, replanning is performed through the AI planner 600 A step of requesting and re-executing the changed unit action may be further included.

After the fifth step, the performance results of the unit actions may be delivered to the disaster execution application unit 500 . In general, as a disaster situation is reported to a control tower in the event of a disaster, the disaster management unit 520 can report the results of unit actions to the disaster manager or supervisor according to a pre-set reporting line in oral or written form, wired or wireless, Re-planning may be performed through the action re-execution unit 420 when a correction instruction or the like occurs.

The disaster situation and response status through the disaster situation information visualization unit 510 and the execution and progress of the unit actions performed by the disaster plan action execution unit 400 accurately in real time based on the location information of the mobile terminal of the disaster manager and supervisor It may further include a step of visually displaying or displaying location and movement situation information mapped to a geographic information system.

In addition, based on the report generated by the disaster management unit 520 through the disaster situation information provision unit 530, basic data that can comprehensively brief the disaster situation or provide real-time disaster services through newspapers or broadcasting media can be provided. Additional steps may be taken.

So far, the present invention has been described with reference to specific examples. However, those skilled in the art to which the present invention pertains will be able to perform various applications and modifications within the scope of the present invention based on the above information.

100: Disaster Information Collection Department
110: Disaster situation information collection department
120: disaster feature extraction unit
130: Disaster Situation Information Analysis Department
200: disaster plan setting unit
210: query input unit
220: disaster plan generation unit
221: initial state setting unit
222: target state setting unit
230: PDDL file generation unit
231: domain file
232: problem file
300: disaster plan action generator
400: Disaster Plan Action Execution Department
410: action execution unit
420: action re-execution unit
500: disaster execution application department
510: Disaster situation information visualization department
520: Disaster Management Department
530: Disaster situation information provision department
600: AI planner
700: knowledge base

Claims (10)

Disaster information collection unit that collects information on situations in which disasters occur;
Disaster plan setting unit for generating initial disaster conditions and disaster response goals to establish a plan for coping with disaster situations;
a disaster plan action generation unit that automatically creates a disaster response plan through an AI planner according to the disaster initial conditions and the disaster response goal;
a disaster plan action execution unit that performs unit actions generated by the disaster plan action generation unit;
A disaster execution application unit that visually displays the disaster situation and response status, reports, supervises, and instructs the disaster situation, and
Including a knowledge base for storing situation information in which the disaster occurred,
The disaster plan action execution unit action execution unit for executing the unit actions automatically generated by the AI planner, and
Disaster response system through artificial intelligence-based planning and decision-making, characterized in that it comprises an action re-execution unit requesting re-planning to the AI planner when the execution of the unit action fails. According to claim 1,
The disaster information collection unit includes a disaster situation information collection unit for collecting situation information in which the disaster has occurred;
A disaster situation feature extraction unit for extracting parameters for determining the disaster area, disaster type, and degree of emergency from the disaster situation information; and
Disaster response system through artificial intelligence-based planning and decision-making, characterized in that it comprises a disaster situation information analysis unit for analyzing the disaster situation through the parameters. According to claim 1,
The disaster plan setting unit optional query input unit for determining prerequisites or priorities for coping with disasters;
A disaster plan generation unit for setting the disaster initial condition and the state of the disaster response target; and
Disaster response system through artificial intelligence-based planning and decision-making, characterized in that it comprises a PDDL file generation unit for generating a domain file and a problem file. According to claim 3,
Disaster response system through artificial intelligence-based planning and decision-making, characterized in that the domain file and the problem file are written in PDDL language. According to claim 4,
Disaster response system through artificial intelligence-based planning and decision-making, characterized in that the problem file includes the precondition and action function for coping with the disaster situation. delete According to claim 1,
The disaster execution application unit includes a disaster situation information visualization unit for visually displaying the execution and progress of the unit action;
A disaster management department that prepares and reports a report to report on-site information and crisis response situation information to the reporting line of the disaster management manager by synthesizing the results and whether the unit actions have been performed; and
Disaster response system through artificial intelligence-based planning and decision-making, characterized in that it comprises a disaster situation information providing unit for briefing the disaster situation or providing it to broadcast media based on the report data generated by the disaster management unit. A first step of collecting and analyzing disaster situation information in a disaster information collection unit;
A second step of setting an initial state and a target state of a disaster and generating a PDDL domain file and a problem file;
A third step of automatically generating unit actions for transmitting the PDDL domain file and the problem file to an AI planner and operating a plan corresponding to a disaster through the AI planner;
a fourth step of executing the unit action; and
Disaster response method through artificial intelligence-based planning and decision-making, characterized in that it comprises a fifth step of determining whether to perform the unit action. According to claim 8,
After the fifth step,
Further comprising calculating an input cost according to the execution of the unit action when the execution of the unit action is successfully completed, and requesting re-planning through the AI planner when the unit action fails. Disaster response method through artificial intelligence-based planning and decision-making, characterized in that. According to claim 8,
After the fifth step, the disaster response method through artificial intelligence-based planning and decision-making, characterized in that it further comprises the step of transmitting the performance results of the unit actions to the disaster execution application unit.