NSF Award
0415168
Robot developers focus mostly on programming behaviors, particularly in regards to the robot's navigation and sensor systems. All too often, the human-robot interaction (HRI) design is an afterthought, if any thought is given to it at all. Lessons learned from decades of research in human-computer interaction (HCI) are rarely applied to robotic systems. In this project, the PIs will attempt to bridge this gap by studying existing human-robot interfaces and developing novel interfaces using HCI techniques modified for robot applications. To these ends, the PIs will first evaluate the current state of HRI for urban search and rescue, by studying the AAAI Rescue Competition and through formal testing with domain experts. Particular emphasis will be placed in this study on: discovering best practices for providing situation awareness in HRI where the operator is physically separated from the robot; discovering the best methods for HRI when switching autonomy levels in a robot system, either when the switch is robot-driven or human-driven; and investigating how HRI is affected when a robot can change its shape. The PIs will abstract from their study guidelines for designing effective HRI; these in turn will inform the PIs' design of a system for urban search and rescue, where the design of the robot's behaviors occurs simultaneously with the design of the robot's interface.<br/><br/>Broader Impacts: Evaluation methods for human-robot interaction in general, and improved HRI for urban search and rescue applications in particular, will directly benefit society as a whole. Insights acquired in this project into data presentation and fusion will inform the design of future human-machine systems that involve decision-making in safety-critical situations. Similarly, better understanding of the types of awareness required of the human in a human-robot system will impact the multi-user computing domain in general. Since robots often carry more than one on-board sensor (such as video and sonar), information visualization techniques developed during this project that combine disparate sensor information will be applicable to systems such as air defense systems or mass transit control rooms.
