H u m a n -a g e n t -R o b o t t e a m w o R k In real-life situations, robots often need to collaborate with humans. An experience and communication model supports the necessary shared activities.says something, we need to understand how it relates to the world around us: the when, what, and where. Our interpretation draws on a deep understanding of why someone says something and how we are supposed to act on it. What role does it play in the joint activity? How does it help us better understand each other and work together?The question here is what it takes to make a robot understand human communication and produce communication that humans can easily understand in a given context. This requires an understanding of human communication-yet robots see and experience the world differently from humans. Uncertainty and incompleteness pervade a robot's experience, but it must relate communication to that experience to figure out what the humans are discussing.In this article, we explain how we can model robot experience as a collection of representations that bridge the gap between low-level sensing and high-level conceptual structures and multiagent beliefs. Examples from urban search and rescue (USAR) scenarios illustrate the process. The examples are drawn from our experience in the NIFTi project, an ongoing European project to develop systems for human-robot teams to jointly explore urban disaster sites so as to make a situation assessment in the early phase of a disaster response. (NIFTi is the project acronym for the Natural Human-Robot Cooperation in Dynamic Environments project.) NIFTi teams are geographically distributed: the robots are deployed in the hot zone, while the humans can be either at a remote control post or in the field. NIFTi adopts a user-driven view, involving first responders in the entire development cycle. The project evaluates its systems each year in real-life conditions, at firstresponder training sites. Currently, NIFTi is at a stage in which several humans remotely collaborate with one or more semiautonomous unmanned ground vehicles (UGVs) using a multimodal GUI including spoken dialogue, and with an unmanned aerial vehicle (UAV) acting as a roving sensor. M any real-life applications require robots to carry out their tasks together with humans rather than acting autonomously. That requires communication to coordinate activities, work together, and act as a team. Team communication is about more than just words, however. When a team member 28 www.computer.org/intelligent iEEE iNTElliGENT SYSTEMS H u m a n -a g e n t -R o b o t t e a m w o R k
Modeling ExperienceJoint activity between humans and robots takes place in a mutually experienced situation, in which all parties' experience provides the basis for how plans, actions, and their expected outcomes are individually understood. Communication plays a crucial role in building up a shared form of situational awareness to aid coordination and collaboration. Gary Klein and his colleagues outlined several basic requirements for turning a...