ABSTRACT
Urban search and rescue (USAR) involves the rescue of victims from the collapse of a man-made structure. The environment can be characterized as a pile of steel, concrete, dust, and debris. The areas are dark and perceptually disorienting; they no longer look like recognizable structures due to the collapse. Robot-assisted search and rescue in this field domain requires that physically situated robots operate under these unstructured, outdoor environmental conditions in areas that are either too narrow or unsafe for human or canine entry.The relation between humans and robots in USAR is different than the relation in manufacturing, office, or even security applications of robots. Possibly, the most interesting human-robot interaction aspect of USAR is that robots, much like search dogs, must team with people to perform any activity. The teaming is physical because of the mobility challenges imposed by the USAR environment; robots must be carried in backpacks to the voids targeted to be searched and, once the robots are on the ground, they must be tethered. The teaming also is perceptual and cognitive because people must make decisions for the robots and interpret the video, audio, and thermal imaging data provided from the robots. The robots are short, providing a viewpoint from less than 1 ft off the ground. This viewpoint exacerbates a keyhole effect (i.e., the limited angular view associated with many remote vision platforms that gives remote observers a sense of trying to understand the environment through a peephole; Woods, Tittle, Feil, & Roesler, 2004). People must fuse these data with other data sources (e.g., building plans) and knowledge (e.g., time of day) to identify victims and structural anomalies as well as to conduct and coordinate the rescue efforts. Consequently, human-robot interaction in USAR requires distributed information transfer and cooperation. This task must be accomplished even while operators and decision makers (consumers of information provided by the robots) are under extreme cognitive and physical fatigue, introducing new issues not commonly seen in industrial settings. Furthermore, USAR is a domain where the robots perform tasks that cannot be accomplished by a living creature; thus, the operator has no higher metaphor or example of how to use the robot. The high degree of human involvement is not expected to change in the near future. The robots are not autonomously mobile for the demanding conditions of a rubble pile, and the most optimistic road map posits only navigational autonomy within 10 years (Murphy, 2002).This article investigates human-robot interaction during robot-assisted search and rescue activities observed as part of a high-fidelity USAR field training drill in Miami, Florida, managed by Rescue Training Associates. It should be emphasized that data collection was opportunistic and observa
