When disaster like earthquake occurs, it is possible to derive road closure and simultaneous mass-movement. It provides cause that fleeing civilians circumnavigate road hazard and get seriously jammed. To solve these situations, we present a new route planning algorithm that can select safe routes based on the Hyperstar algorithm with congestion like traffic jams costs. Besides, we evaluate our algorithm on disaster situations, through some experiments.

Recently, consumer robots are expected to support people's lives. The robots have to be able to interact intuitively with human when they help people, like carrying items. The interface that uses gestures is attracting attention in recent years. Because it can be an intuitive instruction to humans and we do not need to mount specialized equipment such as a laser pointer when we use it. In this study, we construct a system of robotic mobilization to the indicated location based on finger gesture recognition using the skeletal tracking capabilities of Kinect (Microsoft Corporation). We use Mindstorms EV3 (LEGO Company) as a mobile robot and mount a Kinect on the robot. In this paper, we report the system and the experimental results on the recognition accuracy of the position pointing.

In this paper, we proposed two Information delivery methods for control of large-scale pedestrian flow. One method is the web site providing congestion information of the routes to the railway station. The other is the guide projection projecting text information onto buildings. We explained our proposed methods and evaluate them based on the questionnaire survey.

In RoboCup Rescue Simulation, a multi-agent simulator intended for disaster relief, it has been found that there is a close dependency between the result of the agent's activity and the complexity of the city. In the previous research we tried to define the complexity of the city and predict the results of the behavior of the agent by multiple regression analysis. However, could not predict the behavior sufficiently. In this paper, using the generalized linear mixed model to clarify the dependency, we have tried to predict the behavior of the results of the agent.

近年，一般家庭やオフィス等の日常生活環境に対するアンビエントシステムの構築に関する研究が注目されている．一般的に日常生活拠点では複数人が生活しており，その各個人の行動を円滑にサポートするインタラクションを自律的に実行することがアンビエントシステムの目的である．そのためには，生活拠点に設置された有限のインタラクティブデバイスを適切に操作できる能力が要求される．我々は，このアンビエントシステムを，各人ごとにインタラクティブデバイス制御のためのプランニングを行うエージェントが，お互いに協調競合することでシステム全体を制御する「マルチエージェントプランニング系」に基づいて構築することを目指す．本研究では，中央制御型と間接協調型の2 つの協調形態が同時に存在しうる共存するモデルを提案した．artisoc にてシミュレーション環境を構築し，提案モデルの分析を行った．その結果，共存型を導入することでプランニング精度が向上する可能性を示した．