Human-Robot Interactions in Robot-Assisted Urban Search and Rescue

Abstract

Robot-assisted urban search and rescue is an exemplar of a domain where humans must interact with robots and the information they produce, providing capabilities that do exist in traditional methods. It is also a multi-agent domain, where a human might interact with multiple heterogeneous agents. The paper presents data collected in field exercises with subject matter experts (Florida Task Force 3, Hillsborough County Fire Rescue Department, and Fire Department New York/Federal Emergency Management Agency at the World Trade Center). Analysis suggests that execution errors (e.g., controlling the robots) are more prevalent than errors of intention (e.g., where to use robots), the lack of multi-modal interfaces interferes with task completion, many activities are prototypical and could be encapsulated as autonomous behaviors or schemas, and topological mental models are acceptable and perhaps preferable over metric.