Robot-Assisted Medical Reachback: Using Shared Visual Information * Dawn R. Riddle Robin R. Murphy and Jennifer L. Burke CHI Systems, Inc. Center for Robot-Assisted Search and Rescue 1200 Research Parkway, Suite 120 University of South Florida Orlando, FL 32826, USA Tampa, Fl 33620, USA driddle@chisystems.com murphy@cse.usf.edu, jlburke4@cse.usf.edu * This work is partially supported by a grant from the DARPA Synergistic Cyber-Forces Seedling and Cognitive Systems Exploratory Effort programs, and SAIC, Inc. Abstract - Robot assisted medical reachback (RAMR) involves remote medical personnel conducting operator- and robot- mediated victim assessment in an urban search and rescue environment. A simulated medical reachback exercise was developed to examine RAMR. Key findings suggest it is critical for providers and operators to maintain a shared visual space for developing mental models and facilitating team coordination, Communication analysis across the RAMR task revealed that shared visual information was used approximately 50% of the time to facilitate the development of shared mental models, and again approximately 50% of the time to facilitate team coordination activities. Future efforts in this research domain must further investigate the use of shared visual information to facilitate shared mental models and team coordination. Index Terms – Group and organization interfaces, synchronous interaction, user studies, robots, video analysis, distributed work. I. INTRODUCTION Robot assisted medical reachback (RAMR) involves remote medical personnel conducting operator- and robot- mediated victim assessment and triage decision making in an urban search and rescue environment. This new type of technological mediation between medical providers, robot operators, and victims trapped within a collapsed space offers a challenging application in which to apply the principles, theories and methodologies used in human-robot interaction domains. In conventional medical missions, the point of injury (POI) is removed from the point of care (POC). For example, in a building collapse after an earthquake, a victim may be trapped for over 10 hours after being found, with direct medical assistance in only the last 20 minutes or so before the final extrication. The medical team works to stabilize the victim during final extrication and arranges for transport to POC. There is some communication with the POC personnel, but it is limited and usually only verbal (over radio). The advent of robot-assisted medical reachback (RAMR) is both changing the separation of POI and POC teams and the method of communication. The Center for Robot-Assisted Search and Rescue has small robots that can be inserted 20 to 30 meters into a void and provide 2-way audio, video, simple life sign sensing, and even transport fluids to the victim. The robot can be linked to the internet providing reachback to remote medical specialists who are not on site. RAMR creates a distributed POC situation, where the caregivers include the robot directly interacting with the victim, the robot’s operator, medical personnel on site, and remote medical specialists. RAMR also introduces video as a common resource, complementing or surpassing audio as the dominant information source between team members. The use of robots for medical missions is still new and so RAMR affords the opportunity to study how new technology is integrated and accepted by distributed teams. One of the most interesting aspects of RAMR is its reliance on visual information; it provides a domain to study the effects of 1-way and 2-way video with and without audio. This paper describes an exploratory case study in RAMR utilizing a victim assessment scenario. Section 2 describes the participants, setting, procedure, and methods of data collection and analysis used in the exercise. In Section 3, we identify five phases or subtasks of robot-assisted medical reachback (RAMR): orientation (consisting