Road disaster response support system and program, road management method

Abstract

In recent years, advanced sensing technologies such as artificial satellites, optical fibers, and connected cars have advanced, but a system has not yet been established to integrate this diverse information and dynamically update and optimize road clearing plans in stages according to the elapsed time after a disaster. As a result, prioritizing and managing the progress of road clearing operations is dependent on individual efforts, which could lead to delays in achieving the plan and inefficient allocation of resources. [Solution] A road disaster response support system and program, and a road management method characterized by acquiring information on road disaster conditions, analyzing the acquired information to grasp the road disaster conditions, registering a pre-formulated road clearing plan (including at least time milestones), evaluating the feasibility of achieving the time milestones based on the road clearing plan and the analyzed results according to the elapsed time since the disaster, and dynamically formulating or updating a road clearing implementation plan based on the evaluation results.

Description

Technical Field

[0001] The present invention relates to a technology related to road maintenance management and road opening during disasters. In particular, based on information obtained by a plurality of sensing means (such as satellite sensing, fiber optic sensing, vehicle running sensing, etc.), during normal times, it evaluates the risk of cavity formation in road infrastructure and judges its soundness, and when a disaster occurs, it utilizes a pre-established road opening plan to quickly formulate an implementation plan for road opening to ensure the passage of emergency vehicles and the like. The present invention relates to a road disaster response support system, a related program, and a road management method. In addition, it aims to achieve both wide-area and continuous infrastructure monitoring, which was difficult in the past, and quick judgment and plan support at the initial stage of disasters, and to provide a flexible and advanced support system that functions effectively in both normal and disaster phases. This system contributes to the construction of a road management system with high disaster resilience by seamlessly supporting the preventive maintenance phase and the disaster response phase.

Background Art

[0002] In conventional road infrastructure management, it is common to have an emergency response system after accidents such as road surface subsidence and settlement occur, and a system for comprehensively and continuously grasping the risk of cavity formation and signs of aging of roads during normal times has not been sufficiently established. Also, during a disaster, rapid road opening is required to ensure the passage of emergency vehicles such as ambulances, fire trucks, and the Self-Defense Forces. However, information acquisition and judgment processing necessary for formulating the implementation plan tend to be personnel-intensive, and there is a lack of efficient and scientific support technology. In recent years, with the development of advanced sensing technologies such as satellites, fiber optics, and connected cars, wide-area road monitoring has become technically possible. However, implementation examples of integrating these technologies in both infrastructure soundness evaluation and road opening plans during disasters have been limited. Also, exploration means such as ground-penetrating radar are point-by-point and intermittent, and cannot provide continuous and comprehensive evaluation. Therefore, there are limitations in detecting signs of disaster risk and making immediate route judgments immediately after the onset of a disaster. On the other hand, in the event of a large-scale disaster, it is considered a crucial administrative task to carry out emergency restoration (road clearing) prior to normal recovery procedures to ensure access to medical facilities, evacuation centers, and disaster relief bases. Road clearing, in this context, refers to the process of removing obstacles and making simple repairs immediately after a disaster to enable the passage of emergency vehicles as quickly as possible. A system is needed to quickly compare pre-formulated road clearing plans with actual information at the time of the disaster and incorporate it into the road clearing implementation plan. However, conventional technologies lack the configuration to dynamically utilize road clearing plans. For example, they do not adequately implement processes such as updating plans to reflect damage conditions and traffic disruption information after a disaster, or optimizing the timing of implementation. The configuration for flexibly processing information interpretation, weighting, and route optimization using AI technology is also insufficient, which has contributed to the burden on on-site decision-making and delays in initial response. Furthermore, with the recent increase in inbound tourism, the risk of foreign tourists facing natural disasters has also increased. However, foreign travelers have limited means to quickly and accurately obtain information on road access and evacuation routes during disasters, and language barriers and differences in information acquisition channels have led to delays and confusion in responses. Therefore, there is a need to develop a system that provides easy-to-understand and timely disaster-related transportation information to a diverse range of users, including inbound tourists. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2024-014948 [Patent Document 2] Japanese Patent Publication No. 2024-071243 [Patent Document 3] Japanese Patent Publication No. 2023-162007 [Patent Document 4] Japanese Patent Publication No. 2024-136347 [Overview of the Initiative] [Problems that the invention aims to solve]

[0004] In road infrastructure, there is a high risk of sinkholes and traffic disruptions during disasters due to voids beneath the road surface and ground deformation. Traditionally, emergency responses have been implemented only after these events have become apparent, and widespread and continuous monitoring and management of void risks during normal times have been insufficient. Furthermore, while individual technologies such as fiber optic sensing, satellite remote sensing, and vehicle driving data exist, a system that integrates and analyzes these technologies to support both preventive maintenance and emergency response during disasters had not yet been established. For example, the following prior art is known: Patent Document 1 discloses a technology for detecting road anomalies using optical fiber sensors, but it does not disclose integrated analysis with satellite data and vehicle driving data, or the use of ground survey results to improve AI models. Patent Document 2 describes a technology for detecting road obstacles using satellite imagery and determining whether or not it is passable, but it does not disclose the integration of heterogeneous sensors, the output of risk scores using AI models, or learning improvements. Patent Document 3 describes using AI to analyze in-vehicle camera images and determine the disaster situation, but it does not explicitly mention continuous improvement through heterogeneous sensor fusion or retraining. Patent Document 4 discloses a configuration for visualizing data from radar-equipped vehicles on a map, but it does not mention multi-level integration, AI-based cavity risk assessment, or the utilization of ground survey results. Thus, most conventional technologies are limited to localized anomaly detection based on a single sensor, and a comprehensive system encompassing area-wide and continuous hollowing-out risk assessment during normal times, as well as rational and immediate response support processing during disasters, has yet to be realized. [Means for solving the problem]

[0005] To solve the above problems, the road disaster response support system according to the present invention may have the following configuration. The Road Clearing Department registers pre-formulated road clearing plans and, in the event of a disaster, formulates or updates road clearing implementation plans based on those road clearing plans. An analysis unit that evaluates and predicts the health of road infrastructure using multiple sensing means (e.g., artificial satellites, optical fibers, vehicle movement, etc.) during normal times, The system includes an improvement unit that learns and improves upon the analysis results in order to use them for infrastructure maintenance management decisions or disaster response support decisions. The road clearing unit may maintain time milestones (e.g., 24 / 48 / 72 hours) corresponding to the elapsed time since the disaster as internal parameters, and sequentially revise the road clearing implementation plan according to the estimated probability of achievement based on indicators such as the extent of damage, traffic obstruction, and progress of restoration. The analysis department may conduct ground surveys in areas with a high risk of cavities, and the results may be re-inputted into the AI ​​model to continuously improve prediction accuracy. After a disaster occurs, the AI ​​model and accumulated geospatial data may be used to support the determination of a rational road clearing route. Furthermore, the system may be configured to identify sections that are difficult to pass immediately after a disaster using high-frequency information such as optical fiber and satellites, to determine priority routes for emergency vehicles, and to include a function to visualize risk scores and the reasons for decisions using explainable AI. Furthermore, the system may include methods for enhancing the formulation or updating of road clearing implementation plans by combining functions such as authority determination, equipment allocation, base selection, multi-hazard assessment, communication, transportation, and robotics coordination, and learning and intellectual property analysis. Furthermore, in the embodiment, an intellectual property research unit may be provided to analyze the claim text of publicly available web data and patent databases, perform normalization to absorb variations in claim expression and differences in the order of constituent elements, evaluate the correspondence between the processing procedures or functions included in the road clearing plan / road clearing implementation plan and the constituent elements described in the claims of third parties, and estimate the likelihood of applicability (potential infringement) based on the correspondence. This may include automatic claim chart generation, terminology normalization, requirement mapping, semantic matching of equivalent concepts, etc. The output of the intellectual property research unit can be used for deciding whether to adopt a function, suggesting alternatives, considering licensing policies, notifying risks when collaborating with external parties, and creating risk review materials. An example of the configuration of the present invention is shown below in accordance with the claims. [1] A road disaster response support system comprising: an information acquisition unit that acquires information on the road disaster situation; an analysis unit that analyzes the information acquired by the information acquisition unit to grasp the road disaster situation; and a road clearing unit that registers a pre-formulated road clearing plan (including at least road clearing bases, road clearing routes, and time milestones); and with the goal of opening wide-area transportation routes within 24 hours, access routes within 48 hours, and routes within the affected area within 72 hours, depending on the elapsed time since the disaster, the analysis unit evaluates the feasibility of achieving the time milestones based on the road clearing plan and the analysis results obtained by the analysis unit, and the road clearing unit dynamically formulates or updates a road clearing implementation plan based on the evaluation results. A road disaster response support system as described in [2][1], wherein the analysis unit evaluates the feasibility of achieving the time milestones using a reinforcement learning model. A road disaster response support system as described in [3][1], wherein the analysis unit takes at least one of the following as input: traffic volume simulation, abandoned vehicle information, short-term weather forecast, rescue request information, damage status information, recovery status information, or operational status of equipment and personnel, and evaluates the feasibility of achieving the time milestone using a reinforcement learning model. A road disaster response support system according to [4][2] or [3], characterized in that the reward in the reinforcement learning model is designed based on at least one of the following: the degree of achievement for each time milestone, the change in the survival rate over time, the amount of equipment consumed, and the degree of expansion of the section where traffic is secured. A road disaster response support system as described in [5][1], wherein the road disaster response support system comprises an improvement unit, the improvement unit refers to at least one of the following generated by the analysis unit or the road clearing unit: information on the results of road clearing work, information on the consumption of materials and equipment, and information on the achievement rate of road function restoration, the improvement unit relearns evaluation indicators based on the degree of achievement of the time milestones, and cooperates with the analysis unit or the road clearing unit based on the results of the relearning to improve or update the road clearing implementation plan. A road disaster response support system as described in [6][1], wherein the information acquisition unit acquires at least one of the following: information regarding patrol status, information regarding optical fiber survey status, information regarding satellite survey status, information regarding vehicle driving status, information regarding road service status, or information regarding weather conditions. A road disaster response support system as described in [7][1], wherein the road clearing unit registers at least one of the following as part of the road clearing base: port facilities, temporary landing bases, temporary maritime transport bases, airport facilities, or river transport bases, and may, if necessary, register the Self-Defense Forces' air cushion boats, transport ships, or other special equipment as part of the road clearing base; the analysis unit evaluates at least one of the following for each transport route: maritime transport route, air transport route, and the aforementioned road clearing route on land: transport time, fuel consumption, weather conditions, tidal conditions, and safety; the road clearing unit optimizes the transport efficiency of relief supplies, medical supplies, or equipment based on the evaluation results and a spatial optimization model, reflects the optimization results in the achievement targets of the time milestones, and enables the road clearing implementation plan to be achieved within the deadline. [8] A program comprising a sequence of instructions for causing a computer to perform at least one of the following (A) or (B): (A) a function of a road disaster response support system as described in any one of [1] to [3], [5] to [7]; (B) a function as described in [4] for designing the reward in a reinforcement learning model based on at least one of the achievement level of each time milestone, the change in the survival rate over time, the amount of equipment consumed, and the degree of expansion of the section of road that can be opened to traffic. [9] A road management method using a computer, wherein the computer obtains information on the road disaster situation via a network, analyzes the obtained information to grasp the road disaster situation, registers a pre-formulated road clearing plan (including at least road clearing bases, road clearing routes, and time milestones), and, with the goal of opening wide-area transportation routes within 24 hours, access routes within 48 hours, and routes within the disaster area within 72 hours, depending on the elapsed time since the disaster, evaluates the feasibility of achieving the time milestones based on the road clearing plan and the analyzed results, and executes a process to dynamically formulate or update a road clearing implementation plan based on the evaluation results. A road management method according to

[10] [9], characterized in that the computer performs a process to evaluate the feasibility of achieving the time milestone using a reinforcement learning model. A road management method as described in

[11] [9], characterized in that the computer takes at least one of the following as input: traffic volume simulation, abandoned vehicle information, short-term weather forecast, rescue request information, damage situation information, recovery situation information, or operational status of equipment and personnel, and performs a process to evaluate the feasibility of achieving the time milestone using a reinforcement learning model. A road management method according to

[12]

[10] or

[11] , wherein the computer is characterized in that the reward in the reinforcement learning model is designed based on at least one of the following: the degree of achievement for each time milestone, the change in the survival rate over time, the amount of equipment consumed, and the degree of expansion of the section where passage is secured. A road management method according to

[13] [9], characterized in that the computer refers to at least one of the generated information regarding the results of road clearing work, information regarding the consumption of materials and equipment, and information regarding the achievement rate of road function restoration, relearns evaluation indicators based on the degree of achievement of the time milestones, and performs a process to improve or update the road clearing implementation plan based on the results of the relearning. A road management method as described in

[14] [9], characterized in that the computer performs a process to acquire at least one of the following via a network: information regarding patrol status, information regarding optical fiber survey status, information regarding satellite survey status, information regarding vehicle driving status, information regarding road service status, or information regarding weather conditions. A road management method as described in

[15] [9], wherein the computer registers at least one of the following as part of the road clearing base: port facilities, temporary landing bases, temporary maritime transport bases, airport facilities, or river transport bases, and may, if necessary, register the Self-Defense Forces' air cushion boats, transport ships, or other special equipment as the road clearing bases; evaluates at least one of the following for each transport route: transport time, fuel consumption, weather conditions, tidal conditions, and safety; optimizes the transport efficiency of relief supplies, medical supplies, or equipment using the evaluation results and a spatial optimization model; reflects the optimization results in the achievement targets of the time milestones; and performs processing to enable the road clearing implementation plan to be achieved within the deadline. [Effects of the Invention]

[0006] According to the present invention, the risk of hollowing out of road infrastructure can be grasped comprehensively and continuously during normal times, and advanced preventive maintenance, which was difficult with conventional inspection-based management, can be achieved. In the event of a disaster, a road opening implementation plan can be quickly formulated based on a pre-established road opening plan, and by selecting a reasonable route that reflects the analysis results of the infrastructure status and disaster impact by AI, it becomes possible to ensure the passage of emergency vehicles and optimize the initial response. In addition, by integrating and analyzing heterogeneous sensor information such as optical fiber, satellite, and vehicle travel data, accurately grasping the cavity risk and obstacle status, and continuously training these information into the AI model, the prediction accuracy and responsiveness can be improved in the long term. Furthermore, by providing a visualization function and feedback mechanism based on explainable AI, it supports the judgment of on-site users and local government staff, enabling practical and highly reliable disaster response. In addition, by providing a function for obtaining correction information by robotics, a multilingual support function, and a notification function according to traveler attributes, it is possible to quickly provide information on passable roads and danger avoidance information during disasters to foreign tourists visiting Japan, etc., contributing to reducing confusion and secondary disasters.

Brief Description of the Drawings

[0007] [Figure 1] It is a system configuration diagram related to the road disaster response support system of the present invention. [Figure 2] It is a flowchart showing an example of the flow of patrol. [Figure 3] It is a flowchart showing an example of the flow of a fixed-point camera. [Figure 4] It is a diagram showing an example of the patrol situation. [Figure 5] It is a diagram showing an example of the optical fiber inspection situation. [Figure 6] It is a diagram showing an example of the satellite inspection situation. [Figure 7] It is a diagram showing an example of the weather situation. [Figure 8] It is a diagram showing an example of the vehicle travel situation. [Figure 9] It is a flowchart showing an example of the flow of the road disaster response support system of the present invention. [Figure 10] It is a diagram showing an example of the judgment criteria in the decision-making part of the present invention. [Figure 11] This figure shows an example of the hardware configuration related to the road disaster response support system of the present invention. Note that the configuration diagrams and flowcharts shown in Figures 1, 9, 10, and 11 are examples and do not comprehensively illustrate all elements related to the present invention. Configurations and processes not explicitly shown in the drawings (e.g., road clearing unit, robotics unit, intellectual property investigation unit, authority determination, equipment allocation, base selection, multi-hazard assessment, communication / transportation / robotics coordination, intellectual property investigation, etc.) may also be included in embodiments of the present invention. [Modes for carrying out the invention]

[0008] An embodiment of the road disaster response support system of the present invention will be described below with reference to the drawings. The following description is not intended to unduly limit the technical concept of the present invention as described in the claims. Not all of the configurations described in this embodiment are essential components of the present invention, and each element constituting the feature group may constitute an independent invention. The system of this invention can be applied not only to emergency response support during disasters, but also to evaluating the soundness of road infrastructure during normal times and detecting early signs of the risk of underground voids. This contributes to preventing accidents and road collapses, and to optimizing long-term maintenance costs through infrastructure maintenance support. Furthermore, this system may be configured to provide information in multiple languages ​​to foreign users, including inbound tourists, via smartphones or other devices. This would allow foreign travelers to immediately grasp the passability of their current location and the vicinity of their destination, avoidance routes, and road clearing status in the event of a disaster, thereby supporting them in making safe decisions. Furthermore, "information regarding road disaster conditions" is a comprehensive concept that encompasses multiple types of information used to understand the condition of road infrastructure during a disaster, including disaster information, road information, traffic information, and weather information.

[0009] Figure 1 shows an example configuration of the road disaster response support system 600 of the present invention. The road disaster response support system 600 includes an information acquisition unit 610 that acquires information on patrol status, optical fiber survey status, satellite survey status, weather conditions, and vehicle driving status; an analysis unit 620 that analyzes the various conditions acquired by the information acquisition unit 610; and a decision unit 630 that determines the necessity of road disaster response based on the analysis results. The road disaster response support system 600 of this embodiment is connected to a patrol status provision server 100, an optical fiber survey status provision server 200, a satellite survey status provision server 300, a weather status provision server 400, and a vehicle driving status provision server 500 via a network NW. To understand the patrol situation, Figure 1 shows only one vehicle Vh and terminal device TM, but multiple vehicles Vh and terminal devices TM may be connected to the network NW. To understand disaster situations, etc., Figure 1 shows only one fixed-point camera CAM, but multiple fixed-point cameras CAM may be connected to the network NW. Based on the analysis results, the decision unit 630 may determine whether road infrastructure maintenance is necessary and whether disaster response is necessary, and may instruct the unit to take appropriate action, such as formulating and updating repair plans and road clearing implementation plans, as needed. Road clearing is handled by the road clearing unit, which may distribute and update plans based on estimated time milestones and achievement probabilities. Furthermore, the road disaster response support system 600 may include functional blocks such as an infrastructure maintenance unit, an improvement unit, a robotics unit, and an intellectual property research unit. If necessary, it may also be configured to incorporate information on road service status, output from a prediction unit, and multilingual distribution by an information provision unit. The output from the intellectual property research unit can be used to determine whether to adopt a function, suggest alternative solutions, consider contracts and licenses, and provide risk notifications when collaborating with external parties.

[0010] The terminal device TM, fixed-point camera CAM, patrol status provision server 100, fiber optic survey status provision server 200, satellite survey status provision server 300, weather status provision server 400, vehicle driving status provision server 500, and road disaster response support system 600 communicate via a network NW. The network NW includes some or all of the following: WAN (Wide Area Network), LAN (Local Area Network), the Internet, provider equipment, wireless base stations, dedicated lines, satellite lines, etc. Furthermore, data can be exchanged not only via the network (NW) but also via a memory card. Downloading and uploading data via the network (NW) is also acceptable. Furthermore, in the present invention, the road disaster response support system 600 may be equipped with a function to provide information on the status of road services, and the system may be configured to acquire such information from a dedicated server or terminal device via a network NW.

[0011] Terminal devices TM are used by passengers riding in the vehicle Vh. Terminal devices TM include mobile phones such as smartphones and tablet devices. The terminal device TM may be a communication-type drive recorder mounted on the vehicle Vh or a stationary in-vehicle device, or it may be equipped with an AI (artificial intelligence) image analysis function. The vehicle Vh may also be equipped with a subsurface cavity detection function (a technology that irradiates electromagnetic waves from the road surface toward the road surface and estimates the location of cavities and buried pipes from the reflected waves), and the vehicle Vh may be a subsurface cavity detection vehicle. The terminal device TM has a built-in road patrol application that works in conjunction with the patrol status provision server 100. The terminal device™ includes a positioning device such as a GPS (Global Positioning System) receiver, a communication device for connecting to a network NW, input / output devices such as a G-sensor (accelerometer), camera, and touch panel, and a processor such as a CPU (Central Processing Unit).

[0012] Figure 2 is a flowchart illustrating an example of a patrol. The terminal device TM starts collecting location information, acceleration information, video, etc. when the patrol start button of the road patrol app is pressed (S1) (S2). After the patrol is completed, pressing the "end patrol" button on the road patrol app (S3) transmits the terminal device TM's location information, acceleration information, video, etc. to the patrol status provision server 100 (S4). The patrol status server 100 determines the presence or absence of unevenness on the road surface based on measurement information transmitted from the terminal device TM, and identifies the location of the road surface determined to be uneven. It also determines the extent of damage to the road surface based on transmitted video and images, and identifies the location of the road surface determined to be a risk area.

[0013] Fixed-point cameras (CAMs) are installed on buildings, roadside supports, poles, etc., around areas prone to flooding, such as roads (highways, major arterial roads, roads with heavy traffic, major bus routes, roads connecting to schools, public facilities, and emergency hospitals, roads along mountains and mountainous areas, etc.), underpasses (roads below elevated intersections), roads along rivers, and roads along the coast. Fixed-point cameras (CAMs) include communication-enabled live cameras, web cameras, and network cameras. The fixed-point camera CAM may be a connected dashcam or a small unmanned aerial vehicle camera such as a drone, or it may be equipped with AI (artificial intelligence) image analysis capabilities. The fixed-point camera CAM has a built-in camera application that communicates with the patrol status provision server 100. A fixed-point camera (CAM) includes a lens, an image sensor, a positioning device such as a GPS (Global Positioning System) receiver, a communication device for connecting to a network (NW), and a processor such as a CPU (Central Processing Unit).

[0014] Figure 3 is a flowchart showing an example of the workflow for a fixed-point camera CAM. The fixed-point camera CAM collects road conditions periodically or intermittently (S5), and periodically or intermittently transmits video, location information, date and time information, etc. to the patrol status provision server 100 (S6). The patrol status server 100 determines the extent of road surface damage, etc., based on video and images transmitted from the fixed-point camera CAM, and identifies the locations determined to be at risk.

[0015] The patrol status provision server 100 provides patrol status to the road disaster response support system 600 via the network NW. The provided patrol status is road-specific information and includes some or all of the following: road damage, road collapse, roadbed washout, roadway collapse, pavement damage, gutter damage, uneven road surface, voids under the road surface, liquefaction, snow cover / snow quality / freezing on the road surface, collapsed buildings, vehicle traffic history, power outages, earthquake damage, tsunami damage, typhoon damage, volcanic eruption damage, volcanic activity damage, tornado damage, landslides, mudslides, mudslides, slope failures, tunnel collapses, shoulder collapses, fallen trees, rockfalls, road erosion, total bridge loss, river breaches, river flooding, inundation, dense fog, avalanches, blizzards, presence of accident vehicles / stuck vehicles, etc., and other disaster conditions or things that hinder vehicle traffic. Figure 4 shows an example of a patrol situation, where road damage location 110 is shown in black on the map.

[0016] The Optical Fiber Survey Status Provision Server 200 utilizes optical fiber sensing technology, which uses optical fibers as sensors. It receives backscattered light from communication optical fibers contained in cables laid on roads, etc., detects vibration patterns corresponding to the vehicle driving conditions on the roads, etc., based on the backscattered light, and obtains information on the vehicle driving conditions on the roads, etc., and the surrounding road conditions from the detected vibration patterns and a learned model (information from cameras connected to optical fibers may also be used). Furthermore, by analyzing the intensity, frequency changes, and continuous abnormal patterns of waveforms of minute vibrations propagating underground, it is also possible to grasp the risk of underground cavities and signs of ground deformation. The fiber optic survey status provision server 200 provides fiber optic survey status to the road disaster response support system 600 via the network NW. The provided fiber optic survey status is road-specific information and includes some or all of the following: vehicle traffic history, traffic volume, traffic congestion, sudden vehicle stops, traffic accidents, snow accumulation on the road surface, cavities beneath the road surface, tunnel collapses, accidents, power outages, water leaks / outages, earthquake prediction information, earthquake damage, tsunami damage, typhoon damage, volcanic eruption damage, tornado damage, fiber optic cable breaks / disconnections, and other disaster conditions (including images of the road) and obstacles to vehicle traffic. Figure 5 shows an example of the fiber optic survey situation, with the collapsed section 210 inside the tunnel indicated in black on the map.

[0017] The satellite survey status provision server 300 utilizes satellite remote sensing technology, which involves observations by artificial satellites equipped with SAR (Synthetic Aperture Radar), optical sensors, microwave sensors, etc. It uses data (scattering intensity values, phase information, polarization information, etc.) and images (optical images, SAR images, etc.) observed by the satellites to acquire road conditions, vehicle conditions, etc., from the difference before and after a disaster. In particular, by using time-series interferometric analysis such as InSAR (Interferometric SAR), it is possible to detect minute displacements such as ground settlement, uplift, and tilt with high precision, and estimate signs of underground cavities and the risk of their formation from the deformations that appear on the ground surface. Furthermore, AI (artificial intelligence) analysis may be used to compare and detect road conditions, vehicle conditions, etc., before and after a disaster, or to detect road conditions, vehicle conditions, etc., during a disaster. The satellite survey status provision server 300 provides satellite survey status to the road disaster response support system 600 via the network NW. The provided satellite survey status is road-specific information and includes some or all of the following: collapsed buildings, landslides, mudslides, mudslides, slope failures, tunnel collapses, road damage, roadway collapses, shoulder collapses, cavities under the road surface, fallen trees, falling rocks, complete bridge loss, river breaches, river flooding, inundation, avalanches, vehicle traffic records, ground subsidence, power outages, water leaks / outages, earthquake prediction information, earthquake damage, tsunami damage, typhoon damage, volcanic eruption damage, tornado damage, forest damage, crop damage, presence of accident vehicles / stuck vehicles, etc., as well as any other information that may obstruct vehicle traffic or cause other disasters. Figure 6 shows an example of the satellite survey situation, where the landslide area 310 is shown in black on the map.

[0018] The weather information server 400 provides weather information to the road disaster response support system 600 via the network NW. The weather information provided is regional information and includes some or all of the following: time, weather (sunny, rainy, snowy, etc.), temperature, rainfall, snowfall, snow depth, wind speed, special warnings (heavy rain, strong winds, storm surge, high waves, heavy snow, blizzard), warnings (heavy rain, strong winds, flood, heavy snow, blizzard, etc.), record-breaking short-term heavy rainfall information, landslide warning information, earthquake prediction information, earthquake information, tsunami information, eruption information, information on volcanic activity, typhoon information, tornado information, and disaster situation (including images of roads). Figure 7 shows an example of weather conditions, where warning 410 and seismic intensity 420 are indicated by letters and numbers.

[0019] The vehicle driving status server 500 utilizes automotive sensing technology to acquire various data from vehicles such as connected cars, including vehicle driving conditions and surrounding road conditions. In particular, based on information such as sudden braking, ABS activation, acceleration abnormalities, tire slippage, and bump reactions, it is possible to estimate abnormal behavior when a vehicle approaches a cavity, making it an effective source of information for detecting risk areas caused by underground cavities. The vehicle driving status server 500 provides vehicle driving status to the road disaster response support system 600 via the network NW. The vehicle driving conditions provided include road-specific information obtained from vehicles (including electric vehicles) such as private cars, taxis, buses, and trucks, and include some or all of the following: temperature, sudden braking locations, skidding locations, tire spin locations, tire lock locations, ABS activation locations, flooded locations, underwater cavities, etc. from vehicle sensors, etc.; power outages, earthquake damage, tsunami damage, typhoon damage, volcanic eruption damage, tornado damage, river flooding, liquefaction, road obstacles, etc. from camera footage and images, etc.; collapsed buildings, landslides, slope failures, tunnel collapses, total bridge losses, road damage and impassable areas from 3D data, etc.; vehicle traffic history (including separate data for passenger cars and large vehicles), traffic volume, traffic congestion, passing speed, average speed, acceleration, etc. from probe information (including ETC2.0), etc., and the presence or absence of rainfall and snowfall from wiper operation status, etc., including disaster conditions and things that hinder vehicle traffic. Furthermore, data may be acquired using autonomous driving technology (sensing technology for autonomous vehicles), or it may be detected and identified through AI (artificial intelligence) analysis, such as disaster conditions or vehicle obstruction information. Figure 8 shows an example of vehicle traffic conditions, and areas 510 where there is no record of vehicle traffic are shown in black on the map.

[0020] The information acquisition unit 610, which operates in the road disaster response support system 600, acquires patrol status from the patrol status provision server 100, fiber optic survey status from the fiber optic survey status provision server 200, satellite survey status from the satellite survey status provision server 300, weather conditions from the weather conditions provision server 400, and vehicle driving status from the vehicle driving status provision server 500 via the network NW. Patrol status provided by patrol status server 100 is stored as patrol information 640. Fiber optic survey status provided by fiber optic survey status server 200 is stored as fiber optic survey information 650. Satellite survey status provided by satellite survey status server 300 is stored as satellite survey information 660. Weather conditions provided by weather status server 400 are stored as weather information 670. Vehicle driving status provided by vehicle driving status server 500 is stored as vehicle driving information 680. Furthermore, the information acquisition unit 610 may also be equipped with information regarding road service status. This information regarding road service status is road-specific and mainly consists of information managed by road service providers (such as the Japan Automobile Federation) (such as information from a road service management system), and includes some or all of the following: rescue requests (date, time, location, rescue details, etc.), rescue requests due to abnormal weather (date, time, location, rescue details, etc.), rescue requests due to natural disasters (date, time, location, rescue details, etc.), dead batteries, locked-out keys, running out of gas, flat tires, wheels coming off / falling off, flooding / submersion, recovery from snow / mud, accidents, slips, disaster / damage situations (including images of the road), vehicle towing / transportation, removal / towing / transportation of abandoned vehicles, removal / towing / transportation of damaged vehicles, removal / towing / transportation of accident vehicles, road conditions (including images of the road), traffic conditions, EV charging availability, vehicle inspection results, etc. This road service status is stored as road service information. In addition to the above-mentioned information on receiving rescue requests, the system may also include information on rescue activities obtained through rescue vehicles, communication-type dashcams, small unmanned aerial vehicles (drones), portable information devices, etc., as well as member-reported information provided by members. Hereinafter, in the present invention, "rescue vehicle" means a vehicle owned by the road service provider and used to rescue members or general vehicles (which may include emergency vehicles), and includes tow trucks, service cars, transport vehicles, or vehicles equipped with vehicle rescue equipment. "Rescue request reception information" means information related to rescue requests transmitted to the road service provider by telephone or app from members or third parties, and includes the requester's contact information, location information, reason and circumstances for needing rescue, and information on the target vehicle. "Rescue activity acquisition information" means information acquired at the rescue activity site using a rescue vehicle or a communication-type drive recorder, a small unmanned aerial vehicle camera (drone), portable information device, etc., and includes video data, still image data, audio data, location information, on-site environment information, etc. Furthermore, the information acquired by the information acquisition unit 610 may be limited to only some of the following: patrol status, fiber optic survey status, satellite survey status, weather conditions, vehicle driving status, and road service status. Based on the results of an integrated analysis of this information, the decision unit 630 can be configured to determine the need for preventative maintenance of road infrastructure and the necessity of responding to disasters (e.g., securing emergency routes, determining repair priorities, etc.), and to execute the necessary response processes.

[0021] The analysis unit 620, which operates within the road disaster response support system 600, analyzes patrol information 640, fiber optic survey information 650, satellite survey information 660, weather information 670, vehicle travel information 680, and road service information obtained from the information acquisition unit 610, and stores the analysis results 690. The stored analysis results 690 may be configured to be viewable in formats such as map display, list display, or time-series graph display. Furthermore, the analysis unit 620 may be configured not only to process individual pieces of information individually, but also to perform matching and comparison between these different types of data and to make a comprehensive evaluation based on integrated correlations. For example, by comprehensively matching and comparing at least two or more data from satellite data, fiber optic data, and vehicle driving data, it is possible to detect the overlap and consistency of anomalies at the same location from a multifaceted perspective such as ground displacement, vibration intensity, and driving anomalies, and to evaluate the level of the risk of subsurface cavities. Furthermore, the analysis unit 620 may be equipped with an AI (artificial intelligence) model. The AI ​​model may be configured to use a learning algorithm such as a neural network to integrate multiple sensing data as input and output a risk score (for example, a continuous value from 0.0 to 1.0 or a risk classification category) for each location. The output score can be used as auxiliary information for decision-making in road infrastructure maintenance and disaster response. Furthermore, the system may also include configurations in which information obtained from on-site ground surveys, such as the presence or absence of cavities, cavity location information, cavity depth information, and cavity shape information, is input into an AI (artificial intelligence) model as training data or update data, and processes (retraining, model updating) are carried out to continuously improve prediction accuracy and judgment results. For example, measurement data (date and time, location, seismic intensity, weather data, vibration, vehicle behavior, probe information, three-dimensional terrain data, etc.) included in various information (patrol information, fiber optic survey information, satellite survey information, weather information, vehicle driving information, road service information) acquired by the information acquisition unit 610 can be used as training data to analyze and visualize road disaster conditions and the degree of infrastructure damage risk. Furthermore, the analysis results 690 may be utilized in conjunction with the prediction and improvement units as needed, contributing to future improvements in disaster prediction and decision-making accuracy. Furthermore, the information analyzed by the analysis unit 620 may be configured to analyze only a portion of the information from among the patrol information 640, fiber optic survey information 650, satellite survey information 660, weather information 670, vehicle driving information 680, and road service information.

[0022] Furthermore, the analysis unit 620 may use a deep learning model that automatically detects abnormal signs from multiple disaster-related information sources. For example, by using LSTM (Long Short-Term Memory), abnormal patterns can be detected from time-series changes in weather information, seismic waveforms, vehicle behavior, etc. This makes it possible to predict precursors to disasters and the risk of secondary damage with high accuracy. Alternatively, a GNN (Graph Neural Network) may be applied to analyze the road network structure and the spatial relationships of damaged nodes, thereby extracting preferred road clearing route candidates and routes with a high risk of damage expansion.

[0023] Furthermore, the analysis unit 620 may include a process for assigning a reliability score to each data source. The reliability score is dynamically calculated based on the source of acquisition (public institution / general user), observation conditions, past accuracy history, etc. The analysis unit 620 excludes or weights data below a predetermined threshold to prevent judgment errors due to inaccurate information. In addition, the analysis results 690 may be used to evaluate the priority of response, and may be configured to score the results by weighting factors such as the probability of human casualties, the status of non-functioning shelters and hospitals, and the impact of damage to lifelines, and visualize them as a heat map on a map.

[0024] Furthermore, the analysis unit 620 may be equipped with a reliability score generation function that evaluates the reliability of various types of information used in the analysis. The reliability score generation function may be configured to score each information source acquired by the information acquisition unit 610 based on factors such as the frequency of information acquisition, past accuracy, and acquisition method (automatic measurement / manual reporting), and to prioritize the supply of highly reliable information to the analysis process. For example, for disaster reports and resident reports collected from social media, a reliability score is calculated by taking into account the consistency of location information, the degree of agreement with the damage situation based on image analysis, and past reporting history. Information with a score below a threshold is then excluded from the analysis or corrected. Furthermore, the reliability score may be referenced in each process of the analysis unit 620, prediction unit, and improvement unit, and may be configured to contribute to improving information weighting and anomaly detection accuracy. In addition, the system may be configured to enhance the transparency of decision support by providing an interface (such as a reliability label display) that allows the user to explicitly check the information rating.

[0025] Furthermore, the analysis unit 620 may be equipped with a resident-participatory learning function that utilizes reports, posts, and feedback information from residents and users to improve the accuracy of the AI ​​(artificial intelligence) model and optimize the decision criteria. For example, one configuration involves collecting reports (text, photos, videos, etc.) regarding road damage, sinkholes, and traffic obstructions, along with location and time information. This information is then evaluated to determine if it aligns with analysis result 690 or judgment result 695, and accurate reports are used as training data to retrain an AI (artificial intelligence) model. Furthermore, the system may be configured to encourage improvement in the quality of user submissions through feedback functions (report evaluation and report correction) for false reports and inappropriate reports, thereby contributing to an overall improvement in the model's performance. This type of community-participatory learning structure is expected to link information flow from society as a whole with an AI (artificial intelligence) learning platform, leading to increased awareness of participation in the maintenance and management of public infrastructure and strengthening regional disaster prevention capabilities.

[0026] Furthermore, the analysis unit 620 may also be equipped with a reliability visualization function that outputs the data that forms the basis of the judgment and the degree of confidence (reliability) of the judgment as supplementary information for various judgment results made by AI (artificial intelligence). In this configuration, the reason for the decision and the score are displayed together, for example, "Road closure recommended for this section: Confidence level 87% (basis: satellite imagery + vibration anomaly + SNS report)," allowing users or administrators to verify the transparency of the AI's (artificial intelligence) decisions. The confidence score can be calculated using the probability output (such as the softmax output) from within the AI ​​(artificial intelligence) model or the consistency of the evidence (the degree of agreement between multiple sources). Furthermore, if the judgment score falls below a threshold, notes such as "caution required" or "further investigation recommended" may be added to help prevent overconfidence in the judgment. This type of decision-making visualization configuration enhances social acceptance of AI (artificial intelligence) implementation and contributes to improving a sense of security and satisfaction in actual operation at disaster response sites.

[0027] Furthermore, the analysis unit 620 may be equipped with a reliability evaluation mechanism based on multiple information sources. During a disaster, in addition to official sensing data (satellite images, vibration sensors, vehicle behavior, etc.), near real-time information such as resident reports, social media posts, and patrol reports may be collected, but there is variability in the reliability of this information. Therefore, in this system, a predefined confidence score (e.g., Japan Meteorological Agency = high, SNS = medium, unregistered reports = low) may be assigned to each information source, and a processing configuration may be used to weight these scores when inputting them into the AI ​​(artificial intelligence) model. Furthermore, the system may include a process to evaluate the degree of consistency (consistency) and spatiotemporal consistency of content across multiple information sources, automatically determining the reliability rank (high, medium, or low) of the information, and adding labels such as "needs verification" or "pending" to the judgment results for low-reliability information. Such an information reliability assessment framework is effective in preventing misjudgments at disaster sites and improving the social transparency of AI (artificial intelligence) decision-making.

[0028] Furthermore, the road disaster response support system 600 may be configured to cooperate with external mobility services and in-vehicle equipment. For example, it may be configured to mutually link information with mobility systems such as car navigation systems, smartphones, MaaS (Mobility as a Service) apps, autonomous vehicles, and electric vehicles (EVs) to provide real-time information on road clearing, passable routes, and hazard avoidance instructions during disasters. In this configuration, disaster conditions and road clearing route determination results are automatically notified to in-vehicle terminals and smartphones, and adaptive navigation instructions are possible based on the user's current location, direction of travel, and driving intentions. Furthermore, for autonomous vehicles, the system may be configured to directly reflect "avoidance orders" and "stop orders" in response to road disaster conditions into the control system. Furthermore, API integration with MaaS service providers enables processes such as optimizing emergency transportation methods, restructuring routes, and adjusting traffic demand during disasters, thus contributing to improving the overall disaster response capabilities of society.

[0029] Furthermore, the road disaster response support system 600 may also include a function for setting conditions regarding the timing and triggers for updating the AI ​​(artificial intelligence) model. The model update process may be configured to be executed automatically or semi-automatically when any of the following conditions are met: (1) When a new disaster occurs and actual damage information is obtained from the site (e.g., results of cavity detection, structural damage data, image diagnostic results, etc.) (2) If the system's judgment accuracy falls below a predetermined threshold (e.g., a continuous decline in the confidence score, an increase in false positive feedback, etc.) (3) When a model update order is issued by the government or a specialized agency (e.g., revision of disaster response specifications, change of regulations, etc.) By pre-setting and managing these update conditions, it is possible to prevent the decision-making model from becoming obsolete, thereby ensuring continuous system maintenance and optimizing the accuracy of disaster response.

[0030] Furthermore, the road disaster response support system 600 may be configured to dynamically change judgment rules or evaluation criteria depending on the type, scale, and extent of damage from the disaster. For example, the system can be configured to change the types of data to be targeted, the evaluation items to be prioritized (such as vibration intensity, flood depth, and traffic disruption rate), and the judgment thresholds depending on the type of disaster, such as earthquakes, heavy rains, and landslides. Furthermore, even within the same disaster, the judgment criteria may be optimized for each region by considering the characteristics of the affected area (such as urban / mountainous areas, aging rate, and damage to lifelines). Furthermore, by dynamically switching priority items and implementation content from a "life-saving-focused phase" to a "material support phase" and then to a "life recovery phase" depending on the stage of disaster progression, it becomes possible to enhance the flexibility of on-site response and real-time adaptability.

[0031] Furthermore, the analysis unit 620 may be configured to dynamically adjust the priority of disaster response according to regional characteristics. For example, in areas where facilities requiring special consideration, such as elderly care facilities, welfare facilities, hospitals, evacuation centers, and elementary and junior high schools, are concentrated, the system may be configured to prioritize road clearing for roads surrounding these facilities. It is also possible to perform a multidimensional weighting evaluation based on population density, the proportion of vulnerable people, the degree of concentration of lifelines, and the distribution of medical resources, and to automatically derive the optimal response order for each region. This enables flexible and rational decision-making that is not only based on the physical damage situation but also on the social needs of the region.

[0032] Furthermore, the Road Disaster Response Support System 600 may be configured to handle damage to communication infrastructure and network disruptions during disasters. For example, various information acquisition units and terminals may be configured to ensure communication without infrastructure dependency by using communication methods such as local networks, mesh networks, or LPWA (Low Power Wide Area Network). In addition, a failover configuration can be used to enable field terminals to autonomously continue decision-making, road clearing plan formulation, etc., even if communication with the central server is impossible, thereby enhancing the system's continued operational capability (robustness) during disasters.

[0033] Furthermore, the road disaster response support system 600 may be equipped with a switching control mechanism between automatic decision-making and manual intervention. For example, while the system normally proceeds with responses based on automatic decisions made by AI (artificial intelligence), it is possible to configure the system so that in cases of high urgency or high uncertainty, a specialist or manager can manually intervene and make decisions. This creates a safety net against the risk of misjudgments by AI (artificial intelligence), improving reliability and flexibility in disaster response. In addition, records of manual interventions can be accumulated in the system, contributing to future model improvements.

[0034] Furthermore, the road disaster response support system 600 may be configured to perform specialized processing according to the type of disaster. For example, it may be configured to change the evaluation criteria and weighting in analysis processing, decision processing, and road clearing route selection, taking into account the different damage characteristics, progression speed, and affected area for each type of disaster, such as earthquakes, heavy rains, volcanic eruptions, and tsunamis. This enables optimal decision-making for each disaster and improves the accuracy of the response. Specifically, it may be implemented to select and switch AI (artificial intelligence) models that reflect the characteristics of the disaster, such as emphasizing shaking and ground deformation during earthquakes and emphasizing inundation depth and drainage capacity during heavy rains.

[0035] Furthermore, the Road Disaster Response Support System 600 may be configured to switch decision criteria and processing methods in stages according to the time-series phases of a disaster (pre-disaster, immediately after disaster, emergency response phase, and full-scale recovery phase). For example, it may prioritize rule-based processing that emphasizes speed immediately after a disaster, and then switch to AI (artificial intelligence) analysis that emphasizes accuracy once things have calmed down, thereby achieving optimal decision-making according to the time series. The configuration may also include RNN (Recurrent Neural Network) models using time-series data and event trigger determination based on the disaster progression flow.

[0036] Furthermore, the Road Disaster Response Support System 600 may also incorporate a hybrid AI (artificial intelligence) configuration. For example, by combining disaster response decisions made by an AI model with explicit rule-based (IF-THEN) decisions, the explainability of the AI ​​model's output and the basis for its decisions can be clarified. In particular, during large-scale disasters, it is important for residents and commanders to understand "why that route was chosen," and the presentation of decision results with explanations may be implemented. A configuration that utilizes a Large-Scale Language Model (LLM) to automatically generate explanations of route selection reasons in natural language may also be included.

[0037] Furthermore, the Road Disaster Response Support System 600 may be configured to utilize live video footage from the site acquired by patrol vehicles, drones, fixed-point cameras, surveillance cameras, robots, etc., and to perform AI (artificial intelligence) image recognition and anomaly detection processing. For example, the AI ​​can automatically detect visual anomalies such as rubble accumulation, flooding, and ground fissures in the video footage and execute control processing to increase the priority of response at those locations. This enables real-time assessment of disaster conditions and decision-making support without relying on visual inspection, improving the accuracy of road clearing and guidance of support vehicles.

[0038] Furthermore, the road disaster response support system 600 may also be configured to support disaster response training and simulations during peacetime. For example, a processing configuration could be used in which past disaster data and hypothetical disaster scenarios are input, a prediction unit and an analysis unit 620 perform virtual judgments, and based on the results, simulated road clearing plans, support route generation, and confirmation of material transport plans are performed. This enables practical training before a disaster occurs, thereby improving the ability to respond quickly and accurately during an actual disaster.

[0039] Furthermore, the Road Disaster Response Support System 600 may also be configured to perform real-time processing of disaster information not only through cloud processing but also through edge AI (artificial intelligence) implemented on local terminals. For example, a configuration could be used in which patrol vehicles, drones, robots, etc., deployed at the disaster site perform on-site processing even in environments where cloud communication is difficult, and execute localized decision-making, notification, and control. This would enable a certain level of decision-making support, road clearing decisions, and danger avoidance even when communication is interrupted, resulting in a highly resilient disaster response.

[0040] Furthermore, the analysis unit 620 may be equipped with a triage processing function that prioritizes a large amount of disaster-related information according to its urgency and importance, anticipating situations where a large amount of disaster-related information is concentrated at once. For example, when a large number of disaster reports and sensor data are input simultaneously, the system can be configured to prioritize disaster locations directly related to life and essential infrastructure, and to make decisions regarding the appropriate allocation of resources and road clearing, thereby avoiding delays in decision-making and insufficient or excessive responses. The AI ​​(artificial intelligence) model can calculate a priority score considering the extent of damage, surrounding conditions, and the reliability of the communication source, and determine the processing order based on that score.

[0041] Furthermore, the Road Disaster Response Support System 600 may have a hybrid configuration that flexibly utilizes both cloud and local processing. For example, in situations where network bandwidth is limited, such as immediately after a disaster, important decision-making processes can be performed on local terminals or base servers, and then synchronized and integrated with analysis on the cloud once the network is restored. This ensures both continuity of processing during a disaster and overall optimization. Additionally, an architectural configuration may be adopted that records and shares the basis for decisions and processing results, contributing to subsequent analysis and improvement processes.

[0042] The decision unit 630, which operates in the road disaster response support system 600, determines the necessity of road disaster response and the necessity of road infrastructure maintenance in an integrated and comprehensive manner based on some or all of the analysis results 690 of various information (including patrol information, fiber optic survey information, satellite survey information, weather information, vehicle driving information, and road service information) analyzed by the analysis unit 620, and stores the determination result 695. This judgment result 695 may include criteria that reflect risk score thresholds, correlation patterns between analysis data, consistency with ground surveys, etc. The judgment result 695 may be visualized on a map, list, time-series graph, or dashboard. Furthermore, the decision unit 630 may be equipped with an AI (artificial intelligence) model for decision support, and may be configured to determine maintenance priorities, disaster response urgency, etc., based on statistical threshold judgment, rule-based judgment, or a machine learning prediction model. In this case, the decision execution method can be configured as fully automated processing, semi-automatic processing with confirmation by an operator, or a suggestion-presenting support mode. Furthermore, the decision unit 630 may be configured to function as an execution trigger for disaster response support processing, including proposing repair plans, determining road clearing routes, recommending emergency vehicle routes, and deciding on traffic restrictions, as needed. This enables automatic or semi-automatic disaster response support linked to the analysis results 690. In addition, the decision-making unit 630 may be configured to cooperate with the road clearing unit or robotics unit to formulate a road clearing implementation plan or to instruct robots to carry out road clearing work based on the decision result 695. This links decision-making and execution, enabling a rapid and efficient disaster response. Furthermore, the judgment result 695 and analysis result 690 may be used in the improvement unit to retrain the AI ​​(artificial intelligence) model and update the judgment criteria, and may also be combined with past historical data and future scenario data to be reflected in the prediction unit's prediction of future disaster risks and the formulation of advance plans. Furthermore, the road disaster response support system 600 may include an information provision unit that provides some or all of the judgment results 695 and analysis results 690 to external parties. This information provision unit has the effect of quickly disseminating and sharing road disaster response policies and countermeasures information with administrative agencies, related businesses, and the general public. Furthermore, the decision unit 630 may be configured to automatically generate an explanatory text in natural language using an LLM (Large-Scale Language Model) for the judgment made by the AI ​​(Artificial Intelligence), thereby increasing the transparency and explainability of the basis for the judgment. Furthermore, the judgment result 695 may also be configured to be output in a foreign language (e.g., English, Chinese, etc.) via a multilingual translation configuration, and may be used for notifications and explanations to foreign users.

[0043] Figure 9 is a flowchart showing an example of the flow of the road disaster response support system 600. The information acquisition unit 610 periodically acquires information on patrol status from the patrol status provision server 100 (for example, every few minutes) (S10). The information acquisition unit 610 periodically acquires information on optical fiber survey status from the optical fiber survey status provision server 200 (for example, every few minutes) (S11). The information acquisition unit 610 periodically acquires information on satellite survey status from the satellite survey status provision server 300 (for example, every few minutes) (S12). The information acquisition unit 610 periodically acquires information on weather conditions from the weather conditions provision server 400 (for example, every few minutes) (S13). The information acquisition unit 610 periodically acquires information on vehicle driving conditions from the vehicle driving status provision server 500 (for example, every few minutes) (S14). Then, the analysis unit 620 extracts road damage locations, etc. from patrol information 640 (S15). The analysis unit 620 extracts traffic records, etc. from optical fiber survey information 650 (S16). The analysis unit 620 extracts landslide locations, etc. from satellite survey information 660 (S17). The analysis unit 620 extracts earthquake and tsunami information, etc. from weather information 670 (S18). The analysis unit 620 extracts areas where travel is impossible, etc. from vehicle travel information 680 (S19). Next, based on the analysis results described above, the decision unit 630 makes a comprehensive judgment regarding the necessity of responding to the road disaster and, if necessary, decides to implement response measures (for example, determining a repair plan or a road clearing route) (S20). Furthermore, the Road Disaster Response Support System 600 may also be equipped with information from road users and residents using SNS (Social Networking Service) (disaster information, damage information, rescue information, recovery information, etc.), as well as information from government agencies (police, fire department, Self-Defense Forces, etc.), infrastructure operators (telecommunications, electricity, gas, water, sewage, etc.), transportation operators (railways, buses, ferries, airplanes, etc.), construction and civil engineering companies (including construction industry associations), delivery companies, tourism businesses (inns, hotels, tourist facilities, roadside stations, etc.), and designated public institutions (Disaster Countermeasures Basic Act) (disaster information, damage information, rescue information, recovery information, etc.), and information from the government's Emergency Disaster Countermeasures Headquarters. Note that Figure 9 shows one example of a typical processing flow, and is not limited to this example.

[0044] Figure 10 shows an example of the criteria used by the decision unit 630. The decision unit 630 can determine the necessity of road disaster response based on triggers such as: when a tsunami occurs (1 element) 631; when a slope collapses and optical fibers are severed (2 elements) 632; when there is road damage and collapsed buildings and no history of vehicle traffic (3 elements) 633; ​​when a heavy rain warning is issued and mudslides cause tires to spin, preventing vehicles from moving forward and resulting in severe traffic congestion (4 elements) 634; or when a heavy snow warning is issued, there is snow on the road surface, ABS is activated, and there is severe traffic congestion with many stranded vehicles (5 elements) 635. Furthermore, the system can provide some or all of the judgment results 695 and analysis results 690 determined by the decision unit 630, and various information (patrol information, fiber optic survey information, satellite survey information, weather information, vehicle driving information, road service information) acquired by the information acquisition unit 610, through the information provision unit, etc., with functions such as email notification to road administrators, data provision to road information board systems, data provision to road restoration visualization maps (which unify and display road restoration status, main damaged areas and damage status, road traffic restrictions, intercity travel time, vehicle speed data, vehicle traffic history, population mesh data, etc., on a web map, etc.), data provision to car navigation systems, data provision to autonomous driving systems, data provision to MaaS (Mobility as a Service), information provision to government agencies (police, fire department, Self-Defense Forces, etc.) and the mass media, and information disclosure to road users and residents (homepage, smartphone app, etc.), and it is not necessarily required to go through the information provision unit. An example of the function of disclosing information to road users and residents is as follows, but is not limited to this: A road disaster response support system characterized by transmitting one or more pieces of information acquired by the information acquisition unit 610, or the results of analysis by the analysis unit 620, or the necessity of road disaster response determined by the decision unit 630, or the road clearing implementation plan formulated by the road clearing unit, or the recovery status by the robotics unit, or the prediction of the necessity of road disaster response by the prediction unit, to a mobile terminal device (mobile phone, smartphone, tablet device, laptop computer, game console, etc.) or a fixed terminal device (desktop computer, smart TV, set-top box, digital signage, kiosk terminal, car navigation system, car display audio, etc.). Figure 10 shows an example of typical judgment criteria, and is not limited to this example.

[0045] The road disaster response support system 600 may include an improvement unit that continuously improves some or all of the analysis results 690 (including the analysis method) obtained by the analysis unit 620, the judgment results 695 output by the decision unit 630, and the judgment criteria. The improvement unit aims to improve the accuracy of analysis processing and judgment processing and suppress misjudgments by utilizing disaster-related data (images, numerical data, simulation results, etc.) acquired from external sources. The improvement unit can also cooperate with the infrastructure maintenance unit and the road clearing unit, and can optimize the improvement targets in each process (cavity risk assessment, repair judgment, road clearing route determination, etc.). One example of an improvement process is a processing configuration in which hypothesis data, verification data, and future prediction data related to disasters are input, and after referring to the history of past judgment results 695 and analysis results 690, the improvement unit performs re-evaluation and retraining using methods such as statistical analysis and machine learning, and the results are fed back into the existing model to improve the accuracy of the judgment. The improvement unit may be equipped with AI (artificial intelligence) analysis functions and may perform model retraining processing for image recognition and risk score calculation. For example, it may take image data and various associated sensing data (location information, vibration, temperature, speed, terrain information, etc.) generated from patrol information, fiber optic survey information, satellite survey information, weather information, vehicle driving information, and road service information acquired by the information acquisition unit 610 as input, predict and output disaster situations, and tune the AI ​​(artificial intelligence) model and judgment criteria using the corrected and supplemented results. Furthermore, the improvement department can use information obtained from on-site ground surveys regarding the presence or absence of cavities, cavities, cavities, depths, or cavities as "ground truth data (training data)" and continuously improve the prediction accuracy of cavity risk assessments and the results of infrastructure maintenance decisions through retraining of the AI ​​(artificial intelligence) model. The above improvement process may be carried out by the improvement unit alone, or it may be configured to operate in conjunction with the infrastructure maintenance unit or the road clearing unit. Furthermore, the improvement unit may be configured to work with the prediction unit as needed to improve the model and reset the judgment criteria based on the disaster prediction results.

[0046] Furthermore, the improvement unit may be configured to flexibly select and apply different decision logics depending on the type and circumstances of the disaster, in cooperation with the analysis unit 620 and the road clearing unit. For example, in the case of an earthquake, an analysis logic that emphasizes the vibration frequency of structures may be adopted, and in the case of a flood, a priority decision logic based on flood prediction and road flood height may be adopted, dynamically adapting according to the type of disaster. This enables flexible and accurate decision-making that is in line with the reality of the disaster.

[0047] Furthermore, the improvement unit may be configured to continuously collect and learn from the results of various disaster responses (road clearing results, passability, recovery speed, etc.) and improve the accuracy of future decisions and implementation plans through feedback learning. In this case, linking the information reliability evaluation results (score) will prevent bias in learning due to misinformation. In addition, the improvement unit may explicitly maintain model update conditions (accuracy deterioration, change in disaster type, occurrence of new data, etc.) and have a function to automatically update the model when the predetermined conditions are met.

[0048] Furthermore, the improvement unit may be configured to sequentially acquire information such as obstacle removal work performed by robotic equipment in the road clearing unit, information on the passability of the road clearing route, and on-site restoration progress logs, and to utilize this information for retraining the AI ​​(artificial intelligence) model used in the analysis unit 620. This will enable optimization of robotic control and improvement of environmental adaptability, allowing for continuous improvement of road clearing processing accuracy and initial response capabilities in the event of a future disaster. In addition, the improvement unit may be configured to accumulate on-site environmental sensor information (vibration, collision, temperature, terrain changes, etc.) acquired during robotic work as training data, and to use it for model performance evaluation and tuning.

[0049] The road disaster response support system 600 may include a prediction unit that utilizes historical data such as previously acquired analysis results 690 and judgment results 695 to predict future disaster occurrences and road damage risks in advance. The prediction unit can analyze various time-series data related to disasters (actual data, future scenario data, weather forecast data, etc.) as input and output the future need for road disaster response quantitatively or probabilistically. This will enable road administrators to plan and implement preparatory measures, response plans, and training plans in advance to prepare for the risk of disasters, thereby accelerating and streamlining initial responses in the event of an actual disaster. One example of a prediction process involves inputting hypothesis data, verification data, training data, and future prediction data (for example, assumed patterns of earthquakes or heavy rain disasters that may occur once every few decades) into the prediction unit, comparing them with existing analysis history and judgment results, and simulating the scope of impact and necessary responses in advance when a disaster occurs. The prediction unit may be equipped with AI (artificial intelligence) prediction capabilities. For example, it can be configured to use a prediction model that has learned multiple input factors such as weather conditions, topography, past disaster records, and road structure to pre-extract areas where securing road clearing routes is difficult or risk areas that may hinder the passage of emergency vehicles. Furthermore, the prediction results may be used in conjunction with the Analysis Department 620 and the Improvement Department, and may be fed back into the Infrastructure Maintenance Department and the Road Clearing Department's pre-planning as needed.

[0050] Furthermore, the prediction unit may be capable of evaluating not only future disaster risks but also the risk of secondary damage resulting from aging infrastructure and insufficient maintenance. For example, one configuration could take into account past repair history and infrastructure deterioration indicators to predict potential areas of damage expansion and the risk of chain reactions, and reflect these predictions in pre-repair and road clearing plans.

[0051] Furthermore, the prediction unit may be equipped with a scenario simulation function that performs multi-condition analysis using multiple disaster occurrence conditions (rainfall intensity, epicenter location, time of occurrence, etc.) as variables, allowing for comparison and examination of the optimal initial response, traffic routes, and emergency supply transport routes for each case. In this case, it may be linked with a human-in-the-loop type decision support function to create a proposal mechanism that assumes the intervention of administrators' judgments.

[0052] The Road Disaster Response Support System 600 functions effectively by using one or more pieces of information from among patrol status, fiber optic survey status, satellite survey status, weather conditions, vehicle movement status, and road service status. In particular, by combining two or more pieces of information, complementary analysis becomes possible, enabling more accurate situation assessment and decision support. For example, by understanding wide-area ground deformation through satellite surveys, detecting localized underground vibrations through fiber optic surveys, and acquiring driving records and abnormal behavior through vehicle traffic data, it becomes possible to achieve both wide-area monitoring and localized detection. This creates a multi-layered decision-making platform, from predictive monitoring during normal times to emergency response decisions during disasters. Furthermore, by conducting integrated analysis that comprehensively compares and compares these multiple pieces of information, it becomes possible to perform advanced decision-making processes beyond mere enumeration of information, such as the following: (1) Improved accuracy through matching image data with non-image data: For example, if road damage is detected in image analysis, matching it with the corresponding optical fiber displacement and vehicle vibration history can reduce false detections and improve the reliability of disaster assessment. (2) Advanced situation assessment based on multi-viewpoint information: Even in situations where detection is difficult with a single sensor, such as at night or in severe weather, it becomes possible to understand the damage situation in a temporally and spatially complementary manner by using other information sources (satellite, driving, weather, etc.). (3) Improved rationality and responsiveness of decisions: Integrated matching processing based on diverse sensing information and structural analysis using AI (artificial intelligence) will enable rapid and rational assessment of road infrastructure risk, prioritization of disaster response, and determination of road clearing routes. In this way, this system aims to minimize damage and contribute to the early recovery of disaster-stricken areas by achieving both rapid initial response during disasters and advanced preventive maintenance during peacetime.

[0053] In another embodiment of the present invention, a simplified configuration that acquires and analyzes information in stages can be adopted. In this embodiment, first, a preliminary determination of the occurrence and severity of a road disaster is made based on information regarding weather conditions and patrol status. Subsequently, only if it is determined that the road disaster is severe, at least one of the following pieces of information is acquired: vehicle driving status, fiber optic survey status, satellite survey status, and road service status. This configuration allows for detailed analysis and disaster response decisions, thereby enabling practical disaster response support while keeping system costs and load low. In this configuration, the information acquisition unit 610 first acquires information on weather conditions and patrol status, and the analysis unit 620 evaluates the occurrence and severity of road disasters (for example, earthquakes of magnitude 6 or higher, widespread wind and flood damage, snow damage, landslides, etc.) based on this information. If the evaluation result exceeds a predetermined threshold, the information acquisition unit 610 acquires additional sensing information, and the analysis unit 620 re-analyzes this information, enabling more sophisticated disaster response decisions. This allows for efficient system operation through the phased utilization of information. Furthermore, the processing in this embodiment may be similar to that of a normal configuration, with the decision unit 630 determining whether disaster response is necessary and what the response policy should be, and performing emergency response or notification processing as needed. Moreover, this stepwise configuration can be applied to specific operating modes and simplified implementation forms as an auxiliary configuration to the main configuration, which is a multi-sensing integrated analysis processing.

[0054] The road disaster response support system 600 may include a road clearing unit that registers a pre-formulated road clearing plan and determines a road clearing route based on the road clearing plan and acquired information when a disaster occurs. By equipping the road disaster response support system 600 with a road clearing unit, the effect of being able to quickly clear roads in the event of a disaster can be obtained. In typical disasters, the process involves emergency restoration followed by full-scale restoration. However, in large-scale disasters, emergency restoration (road clearing) is necessary before emergency restoration. Road clearing involves quickly removing minimal debris and repairing uneven surfaces to secure rescue routes, enabling emergency vehicles to pass for life-saving and rescue operations, emergency supply support, and restoration work. Road administrators formulate road clearing plans in advance, which include road clearing bases (disaster prevention bases such as bases for support units and collection points for supplies and equipment), road clearing routes (wide-area movement routes, access routes, and routes within the affected area), and specific action plans (timelines). A timeline is an action plan that organizes and shares in advance, in chronological order, when, who, and what actions will be taken by relevant organizations in cooperation during a disaster. One example of a road clearing unit in the road disaster response support system 600 is a configuration in which a road clearing plan formulated in advance before or after a disaster is registered (registration can be either before or after the disaster), and based on multiple disaster, damage, and traffic-related data acquired by the information acquisition unit 610, or the results of integrated analysis by the analysis unit 620, a road clearing route during the disaster is determined, and an optimized road clearing implementation plan is formulated based on that determination. The registration of this road clearing plan may be configured to register plan information that has been entered in advance by the administrator, or it may be configured to allow a computer to automatically register road clearing plans obtained from an external system. This makes it possible to achieve both the flexibility of human input and the speed of automated processing. Furthermore, the road clearing unit may be equipped with AI (artificial intelligence) analysis, interpretation, and route determination functions. For example, two or more pieces of information acquired by the information acquisition unit 610, such as patrol information, fiber optic survey information, satellite survey information, weather information, vehicle driving information, and road service information, or images, sensor data, and three-dimensional terrain data based on these, may be input into a machine learning model, and after outputting, supplementing, and correcting the road disaster situation, the road clearing route may be dynamically determined. Furthermore, the determination of the road clearing route may be configured to select the optimal route from multiple candidate routes that avoid high-risk points that should be avoided, based on the disaster situation. The road clearing route and road clearing implementation plan determined by the road clearing department may be notified to the road administrator via a management terminal or external server and automatically reflected in conjunction with restoration work and traffic restriction instructions. Furthermore, the decision-making process may include a configuration that references on-site information acquired by the robotics unit to supplement or modify the route determination process performed by the road clearing unit. In addition, the road clearing unit may, if necessary, cooperate with the analysis unit 620, improvement unit, prediction unit, or infrastructure maintenance unit to improve the accuracy of road clearing routes during disasters and to facilitate dynamic replanning. In this application, "timeline including time milestones" refers to an action plan that defines status indicators (such as wide-area travel route opening rate, access route opening rate, in-area route opening rate, road clearing base operating rate, equipment arrival rate, and remaining fuel reserve time) to be reached at predetermined points in time after a disaster (e.g., 24 / 48 / 72 hours). The analysis unit 620 estimates the degree of achievement or probability of achievement of the status indicators based on patrol information, fiber optic survey information, satellite survey information, weather information, vehicle travel information, and road service information, and the road clearing unit may dynamically re-optimize the road clearing implementation plan, including route switching, base replacement, equipment relocation, or rearrangement of work order, according to the estimation results. Furthermore, the road clearing unit may register and maintain multiple versions of the road clearing plan, such as an administrative version, a council version, and a training version, and select, combine, or apply differential versions of the plan based on the estimation results of the degree of achievement or probability of achievement, and maintain the application history (version number, application differential, application time, etc.) as an audit log. The operation of the timeline may be linked to the switching of decision-making criteria or processing methods according to the chronological phases of a disaster (pre-disaster, immediate post-disaster, emergency response phase, and full-scale recovery phase). These time milestones are set with the aim of maximizing the survival rate within 72 hours, taking into account the 72-hour critical period in life-saving operations.

[0055] Furthermore, the road clearing unit may be equipped with a function to automatically generate a road clearing order based on prioritizing relief activities, comprehensively evaluating past disaster history, current damage status, and the functional status of evacuation, medical, and logistics infrastructure. For example, it may prioritize areas with significant damage and a high probability of saving lives, while also considering accessibility to important logistics hubs and medical facilities, and determine the priority for each road clearing target segment. This score may be configured to be updated in real time based on dynamic conditions (weather, traffic disruptions, aftershock risk, etc.).

[0056] Furthermore, the road clearing unit may be equipped with a multi-criteria optimization function that automatically generates multiple route candidates and evaluates each candidate from the perspectives of cost, time, safety and probability of arrival, medical accessibility, and infrastructure constraints such as weight / height restrictions to determine the optimal road clearing route. Weighting based on reliability scores may be applied during the evaluation. Additionally, routes may be selected using network flow analysis with a spatial optimization model or preference determination from a Pareto optimal solution set (e.g., weighted synthesis). In this case, a human-in-the-loop configuration can be adopted in which the AI-proposed route is reviewed or feedback is received from humans to update the model. The determined road clearing implementation plan is notified via a management terminal or external server and used for information sharing and work instructions with relevant organizations. Furthermore, the system may be configured to perform dynamic replanning in cooperation with the robotics unit, analysis unit 620, and improvement unit, reflecting on-site information.

[0057] Furthermore, the road disaster response support system 600 may be equipped with security measures to prepare for communication failures and cyberattacks that occur during a disaster. This configuration employs a "zero trust architecture" that implements user authentication, communication encryption, and access control in multiple layers, thereby enhancing the overall security resilience of the system. Furthermore, encrypted communication and mutual authentication are introduced between each subsystem to minimize the risk of unauthorized access and data tampering during disasters. Furthermore, to prepare for main server failures or network outages, a failover configuration (automatic switching to a redundant system) or a configuration with alternative processing capabilities on local terminals may be implemented. For example, even if instructions from the cloud server become unreceivable, the system can be configured so that the local terminal autonomously presents and executes a response plan using pre-downloaded road clearing plans and AI (artificial intelligence) models. This enhanced security and resilience configuration ensures high availability and safety even under large-scale disasters, contributing to improved reliability for full-scale implementation by local governments and public institutions.

[0058] Furthermore, the road disaster response support system 600 may be configured to perform clustering processing according to the characteristics of the affected municipality or region, and to perform individually optimized decision processing on a regional basis. For example, by using regional characteristic data such as population density, topography, traffic infrastructure density, and disaster history to cluster multiple similar municipalities and applying different models and priority evaluation criteria to each cluster, a flexible, non-uniform disaster response becomes possible.

[0059] Furthermore, the road clearing unit may be equipped with a collaborative configuration for performing road clearing work at disaster sites using robotic equipment (such as autonomous heavy machinery and remotely operated removal devices). The type, size, and removal methods of obstacles to the road clearing route are determined in cooperation with the analysis unit 620 or the improvement unit, and work instruction data corresponding to the determination results is transmitted to the robotic equipment to automate or semi-automate on-site work. Furthermore, the system may be configured to transmit work performance information (processing time, obstacle handling history, on-site images, sensor information, etc.) fed back from robotic equipment to the analysis unit 620 or the improvement unit, and to reflect this information in route determination, work estimation, and equipment selection for the next disaster response. This enables the coordination of AI (artificial intelligence) decision-making with robotics as the execution force, significantly improving the responsiveness and safety of disaster response.

[0060] The Road Disaster Response Support System 600 may include a robotics unit that uses AI (artificial intelligence) and robotics technology to have robots (mainly disaster response robots) carry out road clearing work. The system uses AI (artificial intelligence) to determine which roads should be prioritized for clearing, clearly indicating the road clearing route, and then robots use robotics technology to carry out the road clearing work. In carrying out road clearing work, the system uses AI (artificial intelligence) to perform optimal route analysis based on pre-formulated road clearing plans registered in the road clearing unit or road clearing implementation plans formulated by the road clearing unit, and information acquired by the robotics unit (disaster / damage situation, weather conditions, road conditions, traffic conditions, impassable road status, rescue status, recovery status, obstacle information, topographic information, road clearing progress information, latest on-site information, etc.) (the robots do not necessarily have to be autonomous robots). The robotics department will achieve the following benefits: It will be able to perform tasks quickly without relying on human labor by utilizing autonomous robots. Furthermore, autonomous work by robots will minimize the deployment of workers to hazardous areas. In addition, the coordination of heavy machinery, small robots, and drones will enable efficient obstacle removal and other tasks. An example embodiment of the robotics unit in the Road Disaster Response Support System 600 is as follows, but is not limited to this (each robot is connected to a network NW). The collection and learning of these motion data may be organized according to disaster categories such as earthquakes, tsunamis, floods, landslides, snow damage, volcanic eruptions (including ashfall disasters), and nuclear disasters. A road disaster response support system (including road management methods) characterized in that, based on a pre-formulated road clearing plan (a road clearing plan registered with the road clearing department) or a road clearing implementation plan, the robotics department uses artificial intelligence analysis and robotics technology to have disaster response robots (large heavy machinery, autonomous heavy machinery, small robots, humanoid robots, quadruped robots, snake-like robots, multi-legged robots, worm-like robots, shapeshifting robots, articulated robots, crawler robots, autonomous excavation robots, small unmanned aerial robots, underwater exploration robots, rescue robots, etc.) perform road clearing work. Examples of robotics technology (including disaster response robots) are as follows, but are not limited to these: Remotely or autonomously controlling autonomous heavy machinery (bulldozers, excavators, etc.) to remove obstacles and repair uneven surfaces. Removing small-scale debris in cooperation with radio-controlled debris removal robots, etc. Utilizing quadruped robots or drones to support reconnaissance of disaster areas and removal of small obstacles. Furthermore, AI (artificial intelligence) analyzes the progress of road clearing in real time and automatically adjusts the optimal work instructions for robots. In addition, integrated control of multiple different robots (autonomous heavy machinery, small robots, drones, etc.) to ensure optimal work allocation. Examples of AI (Artificial Intelligence) are as follows, but are not limited to these: Route optimization AI for calculating road clearing routes and determining priorities (Dijkstra's algorithm, reinforcement learning, multi-agent, etc.). Image recognition AI for obstacle identification using drone and robot sensors and generation of 3D maps (convolutional neural networks, PointNet, etc.). Robotics control AI for controlling autonomous heavy machinery, collaborative work of small robots, automatic obstacle avoidance and path planning (imitation learning, reinforcement learning, deep reinforcement learning, multi-agent, Simultaneous Localization and Mapping, etc.). Dynamic route update AI (machine learning, Long Short-Term Memory, etc.), and AI-based work monitoring (convolutional neural networks, Long Short-Term Memory, Transformer, etc.). Examples of input and output data in robotics technology and AI (artificial intelligence) analysis are as follows, but are not limited to these: Route Optimization AI: Input data (road network data, obstacle data, real-time traffic data, priority route information, weather / land number data, etc.) → Output data (optimal road clearing route, emergency route, work instruction list, etc.). Image Recognition AI: Input data (drone footage, LiDAR point cloud data, past disaster data, etc.) → Output data (obstacle map, obstacle type determination, work priority map, etc.). Robotics Control AI: Input data (work area map, obstacle information, robot status data, terrain data, etc.) → Output data (robot work plan, movement path instruction, obstacle removal operation, etc.). Work Monitoring AI: Input data (work video data, robot work log, weather information, etc.) → Output data (progress report, anomaly detection alert, work optimization instruction, etc.). By utilizing large-scale language models, road clearing plans, implementation plans, and road clearing operations can be continuously improved regardless of the language used. This is particularly effective for understanding pre-formulated road clearing plans, learning about disaster countermeasures using vast amounts of data on the internet, and making real-time decisions regarding disaster response. Examples of large-scale language model applications are, but are not limited to, the following: supplementing and optimizing road clearing plans and implementation plans, learning from global disaster countermeasure data on the internet, real-time support for robots, and real-time utilization of disaster data. Furthermore, the robotics unit may operate in conjunction with the analysis unit 620, the improvement unit, the prediction unit, and the road clearing unit, and may include a configuration that dynamically updates the target area and priority of road clearing operations based on disaster risk information and judgment results provided by each unit. In addition, it may cooperate with the infrastructure maintenance unit as needed to coordinate with the main recovery work carried out immediately after a disaster response, and to provide support for ongoing infrastructure maintenance. The robot types listed above are merely examples and are not the only ones that may be used. Other types of robots may be used as appropriate depending on the purpose of disaster response and the operating environment. Furthermore, in the event of a nuclear disaster, the aforementioned road clearing operations may be planned and executed based on remote and contactless operation, with the input including dose measurement data, radiation controlled area information, downwind forecast information, etc.

[0061] The road disaster response support system 600 may also include an infrastructure maintenance section. In this specification, "information relating to infrastructure integrity" means information related to maintaining the functional or physical integrity of social infrastructure, such as structural abnormalities, cavity risks, settlement trends, cracks, vibration abnormalities, signs of deterioration, and other such information. The Infrastructure Maintenance Department is responsible for assessing the health of road infrastructure and making maintenance decisions during normal times. In the event of a disaster, it is configured to link the results of these assessments with disaster response procedures, thereby contributing to both maintenance and initial response. This infrastructure maintenance unit may include a configuration that calculates a subsurface cavity risk score using an AI (artificial intelligence) model based on at least two pieces of information acquired by the information acquisition unit 610, such as satellite data, fiber optic data, and vehicle driving data. This allows for the quantitative identification of locations where the formation of subsurface cavities is a concern, and supports the prioritization of infrastructure inspections and repairs. Furthermore, for areas deemed to have a high risk of cavities, on-site ground surveys (e.g., ground-penetrating radar surveys, vibration measurements, camera photography, etc.) are conducted, and the results (presence, location, depth, shape, etc. of cavities) are re-inputted as training data or update data for the AI ​​(artificial intelligence) model. This allows for continuous improvement of the AI ​​model's prediction accuracy or infrastructure maintenance decisions. Furthermore, the system may be equipped with XAI (Explainable AI) technology to visualize the basis for the outputted risk score and anomaly assessment. It may also be configured to preferentially acquire correct data through an active learning strategy, or to use data augmentation processing for training data using GANs (Generative Inverse Networks), etc. This makes it possible to improve model accuracy and learning efficiency. This configuration enables road administrators to make accurate and rational repair decisions based on AI-based infrastructure evaluation results and on-site survey information, contributing to the optimization of maintenance costs and the prevention of accidents. Furthermore, the Infrastructure Maintenance Unit may operate in conjunction with the Analysis Unit 620 or the Improvement Unit, and may be configured to update and optimize cavity risk assessments based on analysis results and judgment results. In addition, it may be configured to cooperate with the Prediction Unit, Road Clearing Unit, and Robotics Unit as needed to support full-scale recovery work after disaster response and ongoing infrastructure maintenance.

[0062] Furthermore, the road disaster response support system 600 may also be equipped with a function for coordinating and controlling multiple robotic devices (unmanned vehicles, unmanned heavy machinery, drones, etc.) deployed at the disaster site. In this configuration, each robotic device can be controlled to switch between remote control mode and autonomous operation mode. For example, in the initial stages of a disaster, the devices can be deployed remotely within a safe range, and once stable operation is confirmed, they can be switched to autonomous operation mode according to the situation on site. Alternatively, the system may be configured to automatically assign missions based on the disaster situation (e.g., obstacle removal, image capture, securing access routes) to robotic devices in cooperation with the analysis unit 620 or the road clearing unit, allowing multiple units to perform tasks in parallel and collaboratively. Furthermore, by incorporating a cooperative control mechanism that aggregates and analyzes sensing information obtained from each device (images, 3D terrain, vibration, obstacle detection, etc.) in real time and feeds it back into the behavior of other units, efficient and safe disaster response operations can be achieved. Such robotics-based integrated control configurations contribute to reducing human risk, improving work efficiency, and expanding the area that can be responded to in disaster situations.

[0063] Furthermore, the road disaster response support system 600 may also be equipped with an emergency supplies transport support function. For example, it may be configured to coordinate roads to be cleared with the logistics network (medical supplies, food, water, fuel, etc.) and prioritize the restoration of roads necessary for emergency vehicle passage. It is also possible to link with information on relief supply collection and distribution centers to perform road selection processing that maximizes logistics efficiency. Processing to optimize the logistics network during a disaster may also be implemented using AI (artificial intelligence) judgment that takes into account transportation schedules, traffic history, and road damage levels.

[0064] Furthermore, the road disaster response support system 600 may be equipped with a user interface configuration that visually presents output information such as analysis results 690 and judgment results 695 in various output formats. Specifically, this configuration may include outputting information such as response priority, road clearing routes, and disaster impact areas using a geographic information system (GIS), augmented reality (AR) navigation, or a list format. This enables the presentation of optimal information according to the situation to various users such as field workers, local government officials, and command centers, thereby improving the effectiveness of decision support.

[0065] The road disaster response support system 600 according to the present invention may include, in addition to the integrated analysis configuration using the AI ​​(artificial intelligence) model described above, analysis processing using non-AI methods such as statistical analysis, threshold comparison, and rule-based inference, not limited to the AI ​​(artificial intelligence) model. This enables flexible configuration selection according to the operating environment and optimization from the standpoint of real-time performance and processing load. Furthermore, the system may include a configuration that allows switching between integrated analysis using non-AI methods and analysis processing using AI (artificial intelligence) models, depending on the application, system configuration, and operating conditions. The following are examples, but are not limited to, a configuration that includes: an information acquisition unit 610 that acquires at least two of the following: information on satellite survey status, information on fiber optic survey status, and information on vehicle driving status; an analysis unit 620 that comprehensively analyzes the information by statistical analysis, threshold comparison, or rule-based inference; and a decision unit 630 that determines whether road infrastructure maintenance or disaster response is necessary based on the analysis results.

[0066] Furthermore, the road disaster response support system 600 according to the present invention may include a configuration that switches its operating policy between a normal operation mode and a disaster operation mode. In normal times, the system's primary objectives are to identify predictive abnormalities in road infrastructure, assess the risk of cavities, and make decisions regarding regular maintenance. However, in the event of a disaster, it shifts to an operational mode that prioritizes securing emergency routes, clearing roads, and supporting rescue operations, based on sensing information and integrated analysis results. Such switching is controlled by software based on the overall system operation policy and does not necessarily require a dedicated "operational switching unit." For example, in normal mode, the analysis unit 620 can apply processing parameters that focus on cavity risk assessment and anomaly detection, while in disaster mode, the decision unit 630 can perform processing that focuses on extracting high-priority road obstructions and selecting corresponding routes.

[0067] Furthermore, the normal operation mode and the disaster operation mode may include a configuration that switches the risk assessment criteria (threshold setting) in the integrated analysis and the prioritization logic for the sites to be surveyed on-site. For example, during normal times, the system prioritizes wide-area and comprehensive early detection of potential problems. To achieve this, the anomaly detection threshold is set relatively loosely, making it easier to identify potential risks. However, in the event of a disaster, the anomaly score determination threshold is set strictly, enabling the system to prioritize the extraction and notification of high-risk locations requiring rapid response. The following are examples, but are not limited to, of the road disaster response support system. The road disaster response support system may be configured to dynamically change the judgment threshold for calculating risk scores and detecting anomalies, or the prioritization logic for on-site survey target locations, depending on whether it is a normal operation mode or a disaster operation mode.

[0068] The road disaster response support system 600 of the present invention may include a configuration that continuously acquires and analyzes sensing information such as disaster, damage, and recovery status, and dynamically re-evaluates and reconstructs the analysis results and response plan based on the latest information as needed. This enables flexible responses that can immediately adapt to changes in the disaster situation.

[0069] The road disaster response support system 600 may, in order to accommodate diverse users such as foreign tourists, perform notification control processing in the analysis unit 620, decision unit 630, or information provision unit according to the user's attribute information and language used. For example, by utilizing the language setting information of the terminal device and GPS information, the system may be configured to automatically notify foreign tourists of disaster information and travel route information in their language if there are dangerous areas within their range of activity. Furthermore, disaster response information may be provided through traveler applications in cooperation with local governments and tourist facilities. Furthermore, the road disaster response support system 600 may also be equipped with a foreign language support configuration. For example, it may be configured to output and notify foreigners of analysis results or road clearing implementation plans translated into multiple languages ​​such as English, Chinese, and Korean via guidance terminals, smartphones, or web portals. This makes it possible to deliver accurate and immediate disaster response information to users whose native language is a foreign language. In this case, for translation processing and multilingual support, an automatic translation configuration utilizing LLM (Large-Scale Language Model) may be adopted. For example, a pre-trained multilingual translation model can be used to accurately and naturally convert specialized disaster terminology and road management terminology into natural expressions. Alternatively, the system may dynamically switch the optimal translation model or output format based on the user's terminal language settings, location information, past usage history, etc., enabling real-time and individually optimized multilingual support. This minimizes delays and misunderstandings in information transmission due to language barriers, enabling safe and effective disaster response support for all users, including foreigners.

[0070] In the following paragraphs, we will sequentially describe the implementation of the road clearing plan in this embodiment, from the framework for designing and optimizing evaluation indicators, to collaboration and authority decisions among related organizations, to the allocation of materials and equipment and selection of bases, to multi-hazard assessment, communication, transportation, and robotics collaboration, and to learning and intellectual property analysis. The road clearing unit according to this embodiment may maintain as internal parameters three time milestones of a timeline targeting the opening of wide-area travel routes (t≦24), access routes (t≦48), and routes within the affected area (t≦72), depending on the elapsed time t (unit: hours) since immediately after the disaster. The analysis unit 620 updates indicators of damage overview, traffic obstruction, and recovery progress in a time series, and the road clearing unit may adjust the road clearing implementation plan in real time by evaluating the deviation between these indicators and the time milestones and updating the prior probability of whether or not the road can be opened using a statistical method (e.g., sequential Bayesian estimation). Furthermore, the analysis unit 620 may associate these evaluation results with time milestones on the timeline (opening of wide-area transportation routes, access routes, and routes within the disaster area) and use them to estimate the degree of achievement or probability of achieving those milestones.

[0071] The above time targets may be determined using a weighting function W(t) based on the survival rate curve R(t). R(t) assumes a decline in the survival rate within 72 hours based on general knowledge of disaster medicine, and the road clearing section may be selected to maximize an evaluation value obtained by multiplying the amount of medical resources reachable by each candidate route by the required time by W(t). This makes it possible to plan with the highest priority on life-saving activities.

[0072] The analysis unit 620 may aggregate the following as a state vector s: traffic volume simulation (demand forecast, oncoming traffic congestion estimation), abandoned vehicle information (reports, roadside devices, image recognition results), short-term weather forecast (rain, wind, snowfall 1-6 hours ahead), rescue request information (fire, medical, and local government request queues), damage and recovery status (road register, work section report, daily work report), and equipment and personnel operational status (deployment, fuel, labor time). The aggregated s may be input into a reinforcement learning model (e.g., a distributed actor-critic) to estimate the achievement probability P(achievement|s) at each time milestone (24 / 48 / 72 hours). The estimation results may be reflected in the dynamic updating of the road clearing implementation plan by the road clearing unit (route switching, priority changes, equipment redistribution, etc.).

[0073] The road clearing unit may simultaneously optimize alternative route generation (avoiding blocked sections, considering elevation and gradient constraints, and weighting bridge structural integrity) and equipment reallocation (fueling sequence, work crew rotation) if P(achievement|s) falls below a predetermined threshold θ. Optimization is performed by updating strategies based on a search that considers multiple objectives such as required time, safety level, and fuel consumption on the road network graph, and then distributing the results with time tags attached to each task in the road clearing implementation plan.

[0074] When working across the management divisions of multiple road administrators or related organizations, the road clearing unit may refer to the results of the analysis unit 620 (priority sections, dangerous sections, alternative routes) and determine whether delegation of authority or substitution of authority is necessary in the relevant section. The determination is made by combining an authority table associated with management boundaries (mapping of clauses of the Road Act, the Disaster Countermeasures Basic Act, and individual agreements) with a score based on the availability, arrival time, and suitability of the equipment held by the implementing body.

[0075] The authorization determination results are reflected as task attributes (approval required, proxy required, notification only) in the road clearing implementation plan, and notifications may be automatically sent according to the chain of command. The notification may include a reference ID of the relevant article or agreement clause and be linked to the approval workflow.

[0076] The Road Clearing Department may analyze relevant laws and prior agreement documents using natural language processing (article segmentation, legal terminology normalization, and modality extraction of obligations, permits, and prohibitions) to extract whether or not authority delegation is necessary. Extraction accuracy can be improved by weakly supervised learning using past agreement implementation records and summaries of precedents as training data. Furthermore, the availability of equipment and materials (number of heavy machines owned, operational rate, and means of transportation) of multiple managers can be assigned to each agent, and the implementing body can be selected by consensus using a multi-agent system, with the results linked to the approval procedure. [Example] In a case where a portion of a national highway and a municipal road were simultaneously blocked due to heavy rain, each agent, such as the national government, prefectures, municipalities, and construction companies, submitted whether or not delegation was legally permissible and the availability of equipment and personnel. The selection algorithm determined a delegation / supplementary system between local governments that met criteria such as shortest reach, avoidance of bridge constraints, and ease of fuel supply, and the approval was automatically forwarded. Note that this example is just one example, and the setting of conditions and combinations of evaluation indicators are not limited to this.

[0077] The Road Clearing Department evaluates the work capabilities (amount of soil processed per hour, cutting and lifting capacity), equipment characteristics (bucket capacity, turning radius, traversability), and operational time (unit regulations, labor, and nighttime restrictions) of the Self-Defense Forces, fire department, police, private construction companies, and transportation companies, and integrates and optimizes the division of roles among multiple units (advance reconnaissance, heavy equipment unit, follow-up transport, and repair team) and shifts (relief, rest, and resupply).

[0078] The above optimization may be formulated as a time-extended graph that combines a three-tiered network of equipment, personnel, and routes, with the aim of minimizing delay penalties to points of demand (medical facilities, bases, isolated settlements). The resulting allocation results are visualized as an operational timeline and distributed to each unit terminal.

[0079] The Road Clearing Department calculates the quantity of equipment and materials necessary for the road clearing implementation plan based on the processing coefficient for each type of obstacle (debris, fallen trees, flooding, road damage, etc.) and the estimated length, width, and depth of the target section, and then compares this with the stockpile ledger. If a shortage is anticipated, potential temporary storage sites and supply bases are dynamically selected by scoring them based on transportation time, disaster resilience, and accessibility, and these selections are reflected in the plan.

[0080] For each disaster scenario (earthquake, tsunami, flood, landslide, snow damage, nuclear disaster, volcanic ashfall), the demand for heavy machinery, personnel, fuel, and equipment is predicted using a Bayesian demand model. If a shortage of stockpiles is foreseen, temporary storage sites and fuel supply bases may be selected by optimizing facility placement using geographic information system data (road hierarchy, bridges, elevation, land attributes) and land use data (vacant lots, warehouses, rest areas / roadside stations, ports) as inputs. [Example] Tsunami inundation areas were set as exclusion constraints, and a logistics warehouse site on high ground was provisionally designated as a temporary storage site. Redundancy was ensured by duplicating the group of gas stations in the port area and the inland oil depots. Note that this example is just one example, and the combination of conditions and evaluation indicators can be changed as appropriate depending on the type of disaster, geographical conditions, and equipment availability.

[0081] For monitoring the operation of heavy machinery and vehicles, fuel level (CAN / OBD), operating time (engine time meter), fault prediction (vibration, temperature, anomaly detection), and available operating time (operator labor, safety regulations) may be acquired in real time. Based on the acquired information, the possibility of continued operation is probabilistically evaluated, and the need for refueling and maintenance is determined. The determination results are directly linked to the selection of temporary storage sites and fuel bases, and the updating of plans.

[0082] In the evaluation process, the distribution of remaining operational time for each vehicle is estimated and superimposed on the distribution of task duration to determine the dispatch and resupply timing that minimizes the risk of mission failure.

[0083] The analysis unit 620 uses a multi-hazard analysis model to quantify the obstruction risk of road clearing routes. This model takes the probability and intensity of each hazard (earthquake, tsunami, flood, landslide, snow damage, nuclear disaster, volcanic ashfall), the vulnerability of road links to the hazard, and exposure indicators such as traffic volume and alternativeity as inputs to calculate obstruction risk indicators for each section, and then weights and aggregates these indicators to output an integrated score for the route. The model may be, but is not limited to, a probabilistic simulation (Monte Carlo), a Bayesian network, a scenario tree, or a machine learning-based surrogate model. The quantification may be re-evaluated with the latest data at time milestones (e.g., 24 / 48 / 72 hours). The road clearing unit dynamically redistributes the priority of road clearing at time milestones using the integrated score as a weight and updates the road clearing implementation plan.

[0084] Prioritization can be achieved by using a method that maximizes the combined utility of the importance of demand points (hospitals, shelters, bases) and their hazard interference levels. The results are visualized on a dashboard.

[0085] The Road Clearing Department will evaluate whether local bases such as roadside stations, logistics centers, school gymnasiums, and industrial parks can be repurposed as equipment and material collection bases or relief activity bases, in addition to road clearing bases, and will incorporate bases deemed usable into the road clearing implementation plan.

[0086] For regional bases, parking capacity, fuel facilities, warehouse functions, communication facilities, disaster preparedness stockpiles, power supply facilities, etc., are scored and dynamically updated in response to changes in the disaster situation. From the perspective of maximizing redundancy, the network reliability of multiple bases is included in the objective function, and a configuration that does not depend on a single point of failure is selected. [Example] A roadside station along the coast and an inland industrial park are paired together, and a dual-base operation is planned to prepare for the isolation of either. Note that this example is just one example, and the evaluation indicators, objective function, and base combination can be appropriately changed according to the disaster situation and regional characteristics.

[0087] As an alternative communication method in the event of a communication failure, satellite communication (satellite phone / satellite IP communication) or mobile base stations may be evaluated as the primary means, with dedicated wireless networks and optical fiber redundant lines as supplementary means, and a redundant communication network may be formed between cleared sites. The placement and operational priority of backup links will be predefined from the viewpoint of link reserve rate and terrain shielding.

[0088] In addition to port facilities, temporary landing sites, temporary maritime transport, airports, and river transport, Self-Defense Forces air cushion boats and transport ships may also be registered as part of the evacuation base. Analysis unit 620 comprehensively evaluates the required time, fuel consumption, weather, tides, and safety for each transport route by sea, air, and land to optimize the efficiency of transporting relief supplies, medical supplies, and equipment. [Example] Under road closure conditions, a segmented transport system combining air and river transport was adopted from the airport to river transport bases around the disaster area, and the plan for setting up temporary piers was simultaneously optimized. Note that this example is just one example, and the combination of transport modes, condition settings, and evaluation indicators can be changed as appropriate depending on the situation.

[0089] The Road Clearing Department may incorporate data obtained from road clearing training and consultations (scenarios, deployment, required time, challenges, etc.) as training data to improve the accuracy of disaster predictions and reflect the results in the road clearing implementation plan. Indicators (arrival time, work efficiency, supply delays) before and after training will be compared, and the degree of improvement will be fed back into the model.

[0090] Training performance data, reports, scenarios, participant feedback, meeting minutes, and video recordings are normalized using natural language processing to identify issues (delay factors, bottlenecks), generate improvement suggestions, and use them as training data. Based on the improved model, the road clearing implementation plan is dynamically updated, and a continuous improvement cycle is applied to the next training and actual disaster.

[0091] Training data, actual disaster data, simulations, and post-event verification materials can be used as input to systematize the differences between human decision-making and AI proposals by extracting, clustering, and justifying the results (explainable AI). [Example] The difference between human judgment prioritizing avoiding unknown bridge risks and AI prioritizing the shortest route was visualized using uncertainty in bridge soundness as an explanatory variable, and the uncertainty penalty was increased in the next model. Note that this example is just one example, and the selection of features, visualization methods, learning settings, etc. can be appropriately changed according to data characteristics and operational requirements.

[0092] This system may collect, summarize, and normalize disaster response-related information from publicly available web data using a large-scale language model, and continuously learn insights (new technologies, equipment, and operational procedures) that contribute to the improvement of road clearing plans, road clearing implementation plans, and work procedures. Multilingual data may be integrated using language-independent embedding.

[0093] The Intellectual Property Research Department analyzes the claim text of patent databases using a large-scale language model to identify processing procedures or functions included in road clearing implementation plans. This analysis includes term normalization, expression variations, and normalization to absorb differences in the order of constituent elements. The department then evaluates the correspondence between the processing procedures or functions and the constituent elements described in third-party patent claims. Based on this correspondence, the department may estimate (score) the likelihood of applicability, and may also include automatic claim chart generation and requirement mapping.

[0094] The Intellectual Property Investigation Department may evaluate the correspondence relationship and estimate the likelihood of infringement by using semantic matching (combination of syntactic trees and semantic role expressions) to absorb variations in claim expression and order differences. [Example] The differences in expression between "formation of alternative communication means" and "establishment of redundant communication routes" were grouped together as equivalent concepts, and an infringement score was calculated. Note that this example is just one example, and the definition method of equivalent concepts, matching method, score calculation logic, etc. can be appropriately changed depending on the situation.

[0095] The Intellectual Property Investigation Department will collect publicly available information relating to third-party systems, programs, and methods, normalize the claim elements of this system, perform a correspondence evaluation, and assess the likelihood that a third-party implementation will satisfy those claim elements. The results may be exported as materials for the risk review meeting.

[0096] The road clearing department will reassess the road clearing implementation plan by considering the combined occurrence of earthquakes, tsunamis, floods, landslides, snow damage, nuclear disasters, and volcanic ashfall, and evaluating the impact of each disaster. In the case of a combined disaster scenario, the road clearing implementation plan will be dynamically updated, including the probability of simultaneous blockages and secondary disasters.

[0097] The information acquisition unit 610 acquires drone aerial footage, in-vehicle dashcam footage, and vehicle driving sensor data in real time. The analysis unit 620 uses image analysis and machine learning to identify debris, flooding, fires, infrastructure damage, recovery progress, and the presence or absence of road obstructions, and performs obstruction classification and estimates the time required for recovery as needed. The road clearing unit immediately reflects the above analysis results in route changes, priority changes, and equipment redeployment.

[0098] The Robotics Department may learn and improve the behavioral models of disaster response robots and robotic equipment based on operational results from peacetime training, normal construction site work, and simulations in each disaster category. The results of activities in actual disasters may also be reflected through online learning.

[0099] The robotics unit may control robots to perform debris removal, excavation, material and equipment transport, and temporary construction based on the analysis results and road clearing implementation plan, and may feed back the results (work progress, fuel consumption, equipment load) to the analysis unit 620 to use for improving the next plan.

[0100] The robotics unit may coordinate the control of multiple robots from different manufacturers and models using a common control signal, and automatically organize the continuous construction of excavation, transportation, soil discharge, and compaction as a workflow. [Example] A four-stage process consisting of excavation robot → automatic transport by dump truck → soil discharge by bulldozer → compaction by vibratory roller was synchronized in terms of cycle time, and control was applied to automatically increase the number of dump trucks in case of work delays. Note that this example is just one example, and the process configuration, synchronization method, number adjustment rule, control algorithm, etc. can be changed as appropriate depending on the situation.

[0101] The Road Clearing Department quantifies multiple evaluation indicators, including the speed of rescue operations, the urgency of saving lives, the importance of restoring lifelines, the need to secure logistics, and the scale of isolation (including the isolation of medical, welfare, and evacuation centers), and calculates a priority score by weighting and integrating these indicators. Based on this score, it dynamically formulates and updates road clearing implementation plans.

[0102] To ensure the explainability of the scores, the contribution of each indicator is calculated (for example, using a method such as the Shapley score) and visualized as the basis for decision-making.

[0103] Secondary disasters such as communication outages, fuel supply disruptions, delays in equipment procurement, personnel shortages, bridge collapses, other infrastructure damage, re-snowfall, and aftershocks may be scenariod, and the feasibility of the road clearing plan may be evaluated using a probabilistic model or multi-agent simulation. [Example] The probability of fuel supply disruption was set, and a redundant route including a reserve tanker truck and inland bases was selected to improve the overall success rate. Note that this example is just one example, and the target events, probability setting methods, redundancy means, evaluation indicators, etc. can be changed as appropriate according to operational requirements. Here, multi-agent simulation refers to a calculation method that represents multiple entities involved in road clearing as agents and simulates their interactions on the road network to calculate indicators such as the feasibility of the plan and the time required.

[0104] The analysis unit 620 categorizes the response into initial response, emergency recovery, and full-scale recovery, and determines whether a phase transition is necessary (example thresholds: degree of damage stability, degree of medical demand urgency, degree of material supply stability). At this time, the determination may also be made considering the achievement status of time milestones (e.g., 24 / 48 / 72 hours). Based on the determination results, the road clearing unit dynamically switches the allocation of materials and equipment, personnel deployment, and work priorities, and updates the road clearing implementation plan to be optimized to suit each phase.

[0105] The Road Clearing Department determines whether abandoned vehicles need to be moved based on the Disaster Countermeasures Basic Act and other relevant laws, and enables the execution of forced removal procedures through API collaboration with towing company associations and contracted companies. The removal results are reflected in the road network graph by removing obstacles, and the passability determination and route optimization are updated.

[0106] The operational status of private construction and transportation companies is determined based on agreement information, disaster risk, and traffic conditions, and the necessity of automatic assembly immediately after a disaster is assessed. Based on the assessment results, a road clearing implementation plan, including automatic assembly notifications or standby orders, is dynamically formulated and distributed to the company terminals.

[0107] By referencing and analyzing the basic disaster prevention plan, regional disaster prevention plan, and disaster prevention operations plan, priority areas and countermeasures will be extracted and linked with the road clearing implementation plan, and adjusted into an integrated disaster response plan that is consistent with higher-level plans. Extraction may be performed using a hybrid of keyword dictionaries and rule-based / large-scale language models.

[0108] Based on the results of the road clearing implementation plan (arrival time, amount of soil processed, degree of repair provisionally completed), at least one of the necessary materials, equipment, process, and personnel allocation for the emergency restoration plan may be automatically selected and output as an emergency restoration plan. [Example] Based on the temporary restoration thickness and traffic demand, the amount of asphalt mixture required for the next day, the shift of the paving team, and the placement of compactors were automatically calculated and sent to the relevant office. Note that this example is just one example, and the input indicators, selection targets, output format, distribution destination, etc. can be changed as appropriate according to operational requirements.

[0109] (Example of evaluation indicator for time target) The road clearing section may calculate an evaluation value W(t)·M / T for a set of candidate routes at time t, based on the survival rate weight W(t), the amount of medical resources reached M (for each route), and the required time T (for each route), and select the route that maximizes this evaluation value while considering the probability of achievement (24 / 48 / 72 hours) as a constraint. W(t) can be parametrically set as a continuous function that decreases monotonically within 72 hours.

[0110] (Reward design in reinforcement learning: example) The reward R may be given as a linear sum of (i) achievement rewards for reaching each 24 / 48 / 72-hour milestone, (ii) waiting time penalties at demand points such as hospitals and shelters, (iii) negative rewards for the risk of passing through dangerous zones, and (iv) penalties for overconsumption of fuel and personnel. The policy is updated using distributed actors, critics, etc., and state transition probabilities are re-estimated online.

[0111] (Knowledge representation for authority determination: example) The knowledge base may store normalized information such as clause ID, modality (obligation / permission / prohibition), subject (national government, prefecture, municipality, road administrator, agreement partner), and action (delegation / instruction / approval). Natural language processing extracts the subject, modality, action, and condition from the clause and maps them to the management boundary.

[0112] (Consensus building for the selection of the implementing body: Example) Multi-agent agreement may be determined by methods such as weighted maximization based on the utility of each body (time to reach, available equipment and materials, degree of legal compliance, etc.), weighted voting, or maximum weight agreement.

[0113] (Unit / Shift Optimization: Example) The objective function can be a combination of factors such as reducing total processing time, minimizing delay penalties at demand points, and risk avoidance, and the solution can be solved as a mixed integer optimization with constraints such as labor regulations, equipment operation, and traffic regulations.

[0114] (Material and equipment demand and facility placement: example) The placement of temporary storage sites for materials and equipment, fuel bases, etc., can be formulated as an optimization problem that simultaneously satisfies the constraint of minimizing the total transportation burden based on the distance between demand points and candidate bases, the constraint of the number of bases to be established p, and exclusion constraints such as flood-prone areas. As an example, the framework of the p-median problem may be used. p is dynamically optimized according to the situation, and the exclusion constraints are updated based on hazard maps.

[0115] (Assessment of operational feasibility: Example) The probability distribution of remaining operational time for each heavy machine / vehicle can be estimated based on sensor information and compared with the task duration distribution. If the probability of failure exceeds a threshold, refueling and maintenance may be prioritized. Sensor characteristics may include fuel level, temperature, vibration, error codes, and continuous operator time.

[0116] (Multi-hazard integration: Example) The probability of each disaster occurring and the degree of disruption to each route are combined using policy weights to calculate the overall risk, and correlated meteorological factors may be corrected using copulas, etc. For road clearing sections, the route score is recalculated using the overall risk as a weight, and the priority is updated.

[0117] In this specification, "spatial optimization model" refers to a general term for methods that use geographic information system (GIS) data, land use data, transportation network data, etc., as input and perform mathematical optimization for purposes such as base placement, route selection, equipment distribution, and transportation efficiency. Specifically, it includes location-allocation models, cost-distance analysis, network flow analysis, and multi-agent simulation, and these can be used individually or in combination to optimize base selection, route planning, equipment distribution, and inter-mode coordination during disaster response. In multi-agent simulation, multiple agents on the network make decisions based on given constraints (e.g., road closures, supply base capacity, working hours) and objectives (e.g., achieving time milestones, reaching high-priority bases), and the resulting spatiotemporal behavior can be evaluated. Furthermore, it is possible to perform an iterative process (optimization simulation coupling) in which a proposed plan (base locations, routes, etc.) obtained by a spatial optimization model is verified as an initial solution in multi-agent simulation, and the results are fed back into the optimization process. This strengthens the plan by considering the effects of probabilistic events and interactions.

[0118] In this invention, "special equipment" refers to equipment that can be used to assist in the transport of goods or personnel in disaster environments where ordinary port facilities, airport facilities, or road transport bases are insufficient. Specifically, this includes air cushion aircraft (LCACs), transport ships, floating piers, temporary bridge erection devices, large hovercraft, heavy equipment barges, landing craft, temporary runway facilities, and large helicopters for air transport (such as the CH-47) owned by the Japan Self-Defense Forces, etc. This special equipment enables temporary landing, loading, unloading, and transport even when ports and airports are unusable, allowing for the rapid delivery of relief supplies, medical supplies, or recovery equipment.

[0119] Specialized equipment can be used not only as a means of transportation but also for emergency recovery work at disaster sites. For example, floating docks can be used as temporary berthing points, and hovercraft can be used to transport equipment and materials in flooded areas. This allows for the development of integrated land, sea, and air material transport plans in conjunction with road clearing implementation plans, and enables the application of spatial optimization models based on multiple evaluation indicators such as transport time, fuel consumption, and safety.

[0120] In this invention, the "Intellectual Property Investigation Department" refers to a functional block that collects, summarizes, and normalizes relevant information from publicly available web data and patent databases regarding processing procedures, functions, data processing flows, etc., included in road clearing plans and road clearing implementation plans; analyzes claim text using a large-scale language model to normalize constituent elements; absorbs variations in claim expression and differences in the order of constituent elements; evaluates the correspondence between the processing procedures or functions and the constituent elements; and estimates the possibility that a third-party implementation will satisfy the constituent elements based on this correspondence. The Intellectual Property Investigation Department may score potential infringement based on automatic claim chart generation, terminology normalization, requirement mapping, and semantic matching that absorbs variations in claim expression and differences in order (e.g., a combination of syntactic trees and semantic role representations). The evaluation results can be used for suggesting alternative functions, proposing modifications to procedures, considering licenses, exporting materials for risk review meetings, etc. Note that this configuration is an example, and the data source, matching method, and score calculation logic can be changed as appropriate according to operational requirements.

[0121] <Hardware Configuration> Figure 11 shows an example of the hardware configuration of a terminal device TM, a fixed-point camera CAM, a patrol status provision server 100, an optical fiber survey status provision server 200, a satellite survey status provision server 300, a weather status provision server 400, a vehicle driving status provision server 500, and a road disaster response support system 600. This figure shows an example where the terminal device TM is a mobile phone such as a smartphone. The terminal device TM has a configuration in which, for example, a CPU 701, RAM 702, ROM 703, a secondary storage device 704 such as flash memory, a touch panel 705, and a wireless communication module 706 are interconnected by an internal bus or a dedicated communication line. Application programs such as road patrol apps are downloaded via the network NW and stored in the secondary storage device 704. The wireless communication module 706 may support cellular, Wi-Fi, LPWA, and satellite communication (satellite phone / satellite IP). The terminal device TM may be equipped with an edge inference accelerator such as an NPU and perform simple object detection and road obstruction recognition on the terminal side. A fixed-point camera CAM has a configuration in which, for example, a CPU 901, RAM 902, ROM 903, secondary storage device 904 such as flash memory, lens / image sensor 905, and communication device 906 are interconnected by an internal bus or dedicated communication line. Application programs such as camera apps are downloaded via a network NW and stored in the secondary storage device 904. The fixed-point camera CAM may also be equipped with an edge inference accelerator such as an NPU / GPU and transmit primary judgment results from images, such as debris, flooding, or fire, as metadata. Each server has a configuration in which components such as a NIC 801, CPU 802, RAM 803, ROM 804, secondary storage devices 805 such as flash memory or HDDs, and a drive device 806 are interconnected by an internal bus or dedicated communication line. A portable storage medium such as an optical disc is mounted on the drive device 806. Programs stored in the secondary storage device 805 or the portable storage medium mounted on the drive device 806 are loaded into the RAM 803 by a DMA controller (not shown) and executed by the CPU 802, thereby realizing the functional parts of each server. Patrol information 640, fiber optic survey information 650, satellite survey information 660, weather information 670, vehicle driving information 680, analysis results 690, and judgment results 695 are stored in the secondary storage device 805. Each server may also be cloud computing. In addition to these, input and output data of the spatial optimization model (candidate locations, network graphs, optimization results, etc.), data to be analyzed by the intellectual property research department (claim text, publicly available information), intellectual property evaluation results (claim charts, requirements mapping, infringement scores), and audit logs and operation history may also be stored in the secondary storage device 805. Furthermore, the Road Disaster Response Support System 600 may be configured to communicate with each information provision server in order to acquire and process information regarding the status of road services. From the perspective of redundant communication, a dedicated wireless network and satellite communication links may be configured as backup lines, and a configuration that automatically failovers in the event of a failure may be implemented. Furthermore, the Road Disaster Response Support System 600 is comprised of a computing environment (cloud or on-premise) equipped with the memory, processor, and storage space necessary for processing each component, such as the Improvement Unit, Prediction Unit, Information Provision Unit, Road Clearing Unit, Infrastructure Maintenance Unit, Robotics Unit, and Intellectual Property Research Unit. The Intellectual Property Research Unit may include a connector for secure (read-only) linkage with external patent databases or public information sources. Furthermore, the system may be configured to include computing resources such as a GPU (Graphics Processing Unit), TPU (Tensor Processing Unit), or AI accelerator for executing the AI ​​(Artificial Intelligence) models used in each component. A hybrid inference configuration may also be used, where lightweight inference is handled on the edge side (terminal devices TM, fixed-point cameras CAM) and batch learning and high-load inference are handled on the center side (each server, cloud). Note that this figure is an example of the hardware configuration shown in Figure 11, and other configurations (edge ​​device configuration, IoT node configuration, distributed processing environment, etc.) may be used depending on the embodiment. To ensure continuity during offline operation, the terminal device TM and fixed-point camera CAM may be equipped with queuing and local caching mechanisms and configured to transmit data with a delay after the line is restored.

[0122] Although embodiments of the present invention have been described above with reference to the drawings, the present invention is not limited to these embodiments or illustrated configurations. For example, configurations not shown but described herein, namely processing functions involved in generating analysis results 690 and judgment results 695, processing for formulating road clearing implementation plans based on registered road clearing plans, integrated analysis processing of various sensing information (optical fiber survey information, satellite survey information, vehicle driving information, etc.), AI (artificial intelligence) model learning improvement processing, implementation configurations for visualization and feedback functions, application of spatial optimization models, utilization of special equipment, intellectual property investigation department's rights infringement assessment and risk notification, failover in case of interruption using redundant communication links (satellite / dedicated wireless), and multilingual information provision are also included in the technical scope of the present invention. Therefore, the present invention can be modified, altered, and substituted in various ways without departing from its gist. Functional blocks not explicitly shown in the drawings (such as road clearing, authority determination, equipment allocation, base selection, multi-hazard assessment, communication / transportation / robotics coordination, intellectual property investigation department, information provision department, etc.) can be implemented by a person skilled in the art using appropriate hardware / software configurations. [Explanation of Symbols]

[0123] 100: Patrol status server 200: Fiber Optic Survey Status Provision Server 300: Satellite Survey Status Provision Server 400: Weather information server 500: Vehicle driving status provision server 600: Road disaster response support system 610: Information acquisition department 620: Analysis Department 630: Decision Section 640: Patrol Information 650: Fiber Optic Survey Information 660: Satellite Survey Information 670: Weather information 680: Vehicle operation information 690:Analysis results 695: Judgment result

Claims

1. Information acquisition unit that acquires information on road disaster conditions, An analysis unit analyzes the information acquired by the information acquisition unit to understand the road disaster situation, It has a road clearing section that registers a pre-formulated road clearing plan (including at least road clearing points, road clearing routes, and time milestones), The goal is to open wide-area transportation routes within 24 hours, access routes within 48 hours, and routes within the affected area within 72 hours, depending on the time elapsed since the disaster. Based on the road clearing plan and the analysis results obtained by the analysis unit, the analysis unit evaluates the feasibility of achieving the time milestones. The aforementioned road clearing unit is a road disaster response support system characterized by dynamically formulating or updating a road clearing implementation plan based on the evaluation results.

2. A road disaster response support system according to claim 1, The road disaster response support system is characterized in that the analysis unit evaluates the feasibility of achieving the time milestones using a reinforcement learning model.

3. A road disaster response support system according to claim 1, The aforementioned analysis unit takes at least one of the following as input: traffic volume simulation, abandoned vehicle information, short-term weather forecast, rescue request information, damage status information, recovery status information, or operational status of equipment and personnel. A road disaster response support system characterized by evaluating the feasibility of achieving the aforementioned time milestones using a reinforcement learning model.

4. A road disaster response support system according to claim 2 or claim 3, A road disaster response support system characterized in that the reward in the reinforcement learning model includes at least one of the following: an achievement reward for reaching a time milestone, a penalty based on waiting time at a demand point for a hospital or shelter, a negative reward based on the risk of passing through a dangerous section, or a penalty based on excessive consumption of fuel or personnel.

5. A road disaster response support system according to claim 1, The aforementioned road disaster response support system includes an improvement unit, The road disaster response support system is characterized in that the improvement unit refers to at least one of the following information generated by the analysis unit or the road clearing unit: information on the results of road clearing work, information on the consumption of materials and equipment, and information on the achievement rate of road function restoration; relearns evaluation indicators based on the degree of achievement of the time milestones; and, based on the results of the relearning, cooperates with the analysis unit or the road clearing unit to improve or update the road clearing implementation plan.

6. A road disaster response support system according to claim 1, The road disaster response support system is characterized in that the information acquisition unit acquires at least one of the following: information on patrol status, information on fiber optic survey status, information on satellite survey status, information on vehicle driving status, information on road service status, or information on weather conditions.

7. A road disaster response support system according to claim 1, The aforementioned road clearing section registers at least one of the following as part of the road clearing section: port facilities, temporary landing bases, temporary maritime transport bases, airport facilities, or river transport bases. If necessary, the Self-Defense Forces' air cushion boats, transport ships, or other special equipment may be registered as road clearing bases. The analysis unit considers at least one of the following transport routes: sea transport routes, air transport routes, and land-based road clearing routes. Evaluate at least one of the following: transport time, fuel consumption, weather conditions, tidal conditions, and safety. The aforementioned road clearing section optimizes the transportation efficiency of relief supplies, medical supplies, or equipment based on the evaluation results and the spatial optimization model. A road disaster response support system characterized by reflecting the optimization results in the achievement targets of the aforementioned time milestones, thereby enabling the road clearing implementation plan to be achieved within the deadline.

8. A program that includes a sequence of instructions to cause a computer to execute at least one of the following (A) to (C): (A) Functions of the road disaster response support system described in any one of claims 1 to 3 or 5 to 7, (B) The road disaster response support system according to claim 2, wherein the reward in the reinforcement learning model is set to a reward that includes at least one of the following: an achievement reward for achieving a time milestone, a penalty based on waiting time at a demand point of a hospital or shelter, a negative reward based on the risk of passing through a dangerous section, or a penalty based on excessive consumption of fuel or personnel. (C) A road disaster response support system according to claim 3, wherein the reward in the reinforcement learning model is set to a reward that includes at least one of the following: an achievement reward for achieving a time milestone, a penalty based on waiting time at a demand point of a hospital or shelter, a negative reward based on the risk of passing through a dangerous section, or a penalty based on excessive consumption of fuel or personnel. A program characterized by causing a computer to execute something.

9. A road management method using computers, The aforementioned computer, via the network, Obtain information regarding road disaster conditions, The acquired information is analyzed to understand the road disaster situation, Register the pre-formulated road clearing plan (including at least road clearing points, road clearing routes, and time milestones), The goal is to open wide-area transportation routes within 24 hours, access routes within 48 hours, and routes within the affected area within 72 hours, depending on the time elapsed since the disaster. Based on the aforementioned road clearing plan and the analyzed results, the feasibility of achieving the aforementioned time milestones is evaluated. A road management method characterized by performing a process to dynamically formulate or update a road clearing implementation plan based on the evaluation results.

10. A road management method according to claim 9, The road management method is characterized in that the computer performs a process to evaluate the feasibility of achieving the time milestones using a reinforcement learning model.

11. A road management method according to claim 9, The computer takes at least one of the following as input: traffic volume simulation, abandoned vehicle information, short-term weather forecast, rescue request information, damage status information, recovery status information, or operational status of equipment and personnel. A road management method characterized by performing a process to evaluate the feasibility of achieving the time milestones using a reinforcement learning model.

12. A road management method according to claim 10 or claim 11, A road management method characterized in that the computer performs a process to set rewards in the reinforcement learning model that include at least one of the following: an achievement reward for achieving the time milestone, a penalty based on waiting time at a demand point for a hospital or shelter, a negative reward based on the risk of passing through a dangerous section, or a penalty based on excessive consumption of fuel or personnel.

13. A road management method according to claim 9, The road management method is characterized in that the computer refers to at least one of the generated information regarding the results of road clearing operations, information regarding the consumption of materials and equipment, and information regarding the achievement rate of road function restoration, relearns evaluation indicators based on the degree of achievement of the time milestones, and performs a process to improve or update the road clearing implementation plan based on the results of the relearning.

14. A road management method according to claim 9, The aforementioned computer, via the network, A road management method characterized by performing a process to acquire at least one of the following: information on patrol status, information on fiber optic survey status, information on satellite survey status, information on vehicle driving status, information on road service status, or information on weather conditions.

15. A road management method according to claim 9, The computer registers at least one of the following as part of the road clearing base: port facilities, temporary landing bases, temporary maritime transport bases, airport facilities, or river transport bases. If necessary, the Self-Defense Forces' air cushion boats, transport ships, or other special equipment may be registered as road clearing bases. For each transportation route, at least one of the following: sea transport routes, air transport routes, and the aforementioned land road clearing routes, Evaluate at least one of the following: transport time, fuel consumption, weather conditions, tidal conditions, and safety. Based on these evaluation results and the spatial optimization model, the transportation efficiency of relief supplies, medical supplies, or equipment will be optimized. A road management method characterized by reflecting the optimization results in the achievement targets of the aforementioned time milestones and executing a process to enable the road clearing implementation plan to be achieved within the deadline.

Images