Autonomous Recognition of Emergency Site by Wearable Sensors

To support rescue activities of first responders is crucial at disaster sites. Especially, provisioning location and situation information is indispensable for those first responders to efficiently rescue injured people in unknown places with a lot of buildings such as private properties (like factories) and university campus. In such an extreme situation, seamless indoor/outdoor maps will be of substantial aid for the first responders. In this paper, we propose a method of creating an indoor/outdoor map by a first responder team. We assume some of them are equipped with range scanners and all the members have GPS and WiFi devices. Then the presence of obstacles and movable areas is estimated by combining information from different sources (like GPS, WiFi and range scanners) with different confident levels. Since these confident levels depend on scenarios and environments, we design an "adaptive information fusion" algorithm that automatically estimates the confident levels to optimize the precision of the generated map. We have demonstrated our method in several experiments with real sensor data.