Human-Robot Emergency Response - Experimental Platform and Preliminary Dataset

Abstract

Abstract : This paper presents progress towards a research infrastructure for studying human-robot performance in laboratory emergency response scenarios and a preliminary dataset. It incorporates an emergency response team that is composed of a human participant, n or = 4 vision sensors in a sensor network, and the uBot mobile manipulator. The apparatus is being assembled to emulate multiple-task emergency response tasks and to study technologies that allows peer-to-peer training and improves the performance of human-robot teams in the tasks. Assets from the team are deployed to perform: (1) situation assessment; (2) decontamination; (3) evacuation of ambulatory victims; and (4) treatment of non-ambulatory victims. A means of scoring emergency response exercises is used to benchmark the performance of individual agents as well as team performance. The effectiveness of a human-robot team was confirmed in our high level emergency response scenarios. The results suggest that the human-robot team of asymmetric roles can significantly outperform individual agents.