Models of Intention for Human-Robot Interaction
AuthorKelley, Richard Charles
StatisticsView Usage Statistics
As demand for robots grows in non-industrial settings, there is acorresponding need to develop systems that engage with humans ona social level. A key component of this social interaction is theprocess of inferring humans' intentions from their observedbehavior. In this dissertation, we define the "intent recognitionproblem" in robotics, describing what it is and why itmatters. We then describe a series of systems that we havedesigned and deployed on several physical robot platforms. Thisincludes a system based on hidden Markov models, a system thatincorporates contextual information from such sources as a parseof the simplified English Wikipedia, and systems based onHewitt's actor model. For each system, we describe its design andevaluate its performance in simulation or on one of severalphysical platforms, including wheeled mobile robots andhumanoids.As a result of our evaluations, we describe several featuresrequired for the successful operation of an intent recognitionsystem. In particular, we demonstrate through multiple systemsthe importance of modeling social contextual information in orderto interpret and predict human actions in unstructuredenvironments. We also offer guidance on important challenges thatremain to be solved as roboticists attempt to build more sociallycapable systems.