In recent years, various kinds of disasters such as landslide disasters, heavy rain-fall disasters, and earthquake disasters have occurred in any area of Japan. In the case of a large-scale disaster, local governments tend to verify the disaster response, identify issues, and consider solutions. However, in the case of other disasters, the response is managed on a department-by-department basis, and it is rarely summarized comprehensively and the overall picture of the disaster response is not reviewed. On the other hand, assuming training, a disaster response scenario is essential. However, although experienced people and researchers create scenarios reviewing past disasters as examples, they are based on their own experience and they are not necessarily reflected all the actual conditions of past disasters.Against this issue, we designed and developed a web application that comprehensively organizes the action logs in chronological order reflecting the structure of the organization even though those action logs are managed individually by each department. With this tool, based on the accumulated dataset, past response logs are given to a user as situations for each time step, and the user can easily and quickly experience disaster response drill.

At disaster response, it is essential to grab whole picture of damage situation quickly and early after disaster occurrence in order to make disaster response effective and efficient. However, it takes much time to understand damage situation because there is not enough information about it. Against this issue, we proposed implementation of CyborgCrowd for situation awareness in disaster response. In order to validate its possibility, we planned the first international disaster drill in October, 2019. In this drill, we simulated to detect flooded area by West Japan Flood occurred in 2018 from aerial photos by collaboration between crowdsourcing and AIs following Human-in-the-Loop process. Especially, in this drill, AIs were also crowdsourced. In this research, we validated the transition of the efforts from crowdsourcing and AIs to detecting flooded area, and verified the accuracy of result by comparing with the actual flooded area published by Geospatial Information Authority of Japan. Furthermore, we found some suggestion about features of detection results by humans and AIs. For example, some humans detected flooded area roughly, however AIs detected it much closely. Based on those features, we proposed the way to decrease the difference between results by humans and AIs. This was essential for local responders to understand the whole picture of damage situation after disaster occurrence urgently. In this paper, we introduced the framework of international disaster drill, clarified the result of validation, and mentioned the possibility of effective collaboration between crowdsourcing and AIs for quick situation awareness in disaster response.

Once disaster occurs, we have to understand the whole picture of damage situation. However, there is not enough human resources and time to do it. In 2019, we were affected by Yamagata earthquake, and many buildings were damaged in Murakami city. Most of buildings damage were concentrated on their roofs. Local responders tried to inspect those roof damage, however they cannot do it from ground. Against this issue, we decided to take images of roof damage utilizing drones. We designed the flight plan covering over affected area, operated a drone, and got images. Aftermath, we created orthophoto mosaic from those images. We published it for local responders to inspect roof damage of each building in the web-based GIS platform. Furthermore, we detect roof damage by human and let AI learn the result of our detection of roof damage. This is followed the framework of Human-in-the-Loop. Just now, accuracy of the roof-damage detection by AI was not so high. In this paper, we introduce the work-flow of this challenge from designing flight plan for drone to roof-damage detection by AI which is educated with images of actual damage situation after disaster occurrence.

This paper aims to introduce the Cyber-Physical disaster evacuation drill designed by the CyborgCrowd team to implement the collaboration between Human, Information Communication Technology (ICT) and Artificial Intelligence (AI) and hence improve disaster relief planning and effort. We will present some of the preliminary results and the social challenges that the project needs to address in the future.

Based on the experience at past disasters, we have developed the effective method of inspection for each building damage. In this method, we developed only paper-based forms for damage inspection. However, with paper-based forms, we cannot control the quality of collected information, we cannot grab the progress of building damage inspection in real-time, and we cannot draw the whole picture of building damage in the affected area for rational and immediate decision making. Against these issues, we decided to develop an integrated system for effective and efficient building damage inspection utilizing mobile-GIS and cloud-based GIS. We designed the flow of Information Production dealing with information as a product, and developed an integrated system. In this process, we consider about how to realize harmonious collaboration between human and ICT. Finally, we implemented it at Abira town in Hokkaido, which was affected by 2018 Hokkaido Eastern Iburi Earthquake on 6th September. Through the on-site implementation for one month, we and local responders finished the damage inspection for all buildings placed in Abira town, which was over 7,200. Furthermore, we found that system supported their activity effectively, and we detected some issues we should solve for implementation at future disaster.

The earthquake occurred on Sep 6th2018 caused all over the Abrira Town the severe physical damage. The town office had to do the building inspection operation in order to proceed support program for reconstructing livelihoods of the disaster survivors; however, the town had only 145 officials in total, so the town could assign 5 personnel on this specific duty. The research group supported the town effort to realize this duty in the quick and accurate manner utilizing the application with the mobile-GIS and cloud-based GIS. The research group focused on the developing the standardized operational procedure to manage human resources collaborating the system. In order to satisfy two conflicting requests of dealing with a large number quickly but accurately the research team designed the pattern for managing teams working with where we made the total balance among the human activity part and computer processing part. As a result, Abira town welcomed the dispatched the officials of the local administrative organ, a total of 1,128 officials, and 7,213 buildings were inspected in 22 days. Only a few numbers of people raised the objections to the results of the inspection.

To realize an effective disaster response, the entire picture of the damage situation must be grasped. which is time consuming, especially after a huge disaster. After the 2011 East Japan Earthquake, it took about six months for local responders to grasp the general damage situation because they inspected the degree of damage for each building and certified the degree of damage for each victim. This study aims to clarify how much work occurs at an actual disaster site. We selected the 2016 Kumamoto Earthquake and the 2007 Chuetsu-oki Earthquake as case studies, and collected work logs from affected municipalities to determine the transition in the work volume. These logs were analyzed chronologically. It took about three months to certify the degree of building damage for each victim regardless of the severity of damage and area characteristics.