In this paper, we focus on exhibiting proper homecoming routes to persons difficult to come their homes or refuges after disasters. They cannot nd out appropriate escape routes in the situation because of lack of information. We propose a new route planning algorithm that can select safe routes based on the A-star algorithm with opportunistic communication to solve this problem. In addition, we evaluate our algorithm on disaster situations through some experiments.
It is extremely hard for visually impaired people to walk unknown places. Using a white cane or taking a guide dog is required by law when they go outside. However, even if they are used to using the white cane or familiar with a guide dog, they usually hesitate to go to unknown places because these canes or dogs don't have any ability to navigate the user. In this research, we propose a pedestrian navigation system named "PULLDOG" for visually impaired people with UHF band RFID technology and quasi-zenith satellite system (QZSS) Michibiki. This system calculates as safe routes as possible from current position to destination, such as giving priority to braille block way. We investigated accuracy of positional estimation with RFID and QZSS and implemented the route-search algorithm based on Dijkstra's method. In this paper, the system outline and the results of investigations are described. Additionally, the results and feedbacks from demonstration experiments at Keio Kitano station are also reported.
In this paper, we describe a real estate rent estimation system for restaurant. The traditional real estate market in Japan, the decision of rents for restaurant has been made by veteran salespeople based on their tacit knowledge that made from cultivated intuition and experience. However, the conversion to explicit knowledge (expression) has become an issue when carrying on into other salespeople. Therefore, the evidence-based rent estimation system is needed to assist salesepeople. We proposed the system that considered static information on locational conditions and dynamic information such as traffic around the store. We have developed a sensing system for obtaioning the sense of traffic volume.
A purpose of this paper is considering usage of Word2Vec and Doc2Vec, which are text vectorization tools by machine learning, to classify event notice. Firstly, we calculate feature vector of several characteristic words relate to event notice by Word2Vec, and enumerate five highest similarity words. Secondly, in a similar way, we calculate feature vector of event notice's text by Doc2Vec, and consider five highest similarity event notices.