Bradley Hayes attended the 23rd IEEE International Symposium on Robot and Human Interactive Communication (RO-MAN ‘14) in Edinburgh, Scotland, to present the paper “People Help Robots Who Help Others, Not Robots Who Help Themselves”, by Bradley Hayes, Daniel Ullman, Emma Alexander, Caroline Bank, and Brian Scassellati.
The theme of the symposium this year was “Human-Robot Co-Existence: Adaptive interfaces and systems for daily life, therapy, assistance and socially engaging interactions.”
While in Edinburgh, Brad also gave an invited talk at the IPAB seminar at the University of Edinburgh.
Title: Algorithmically Constructing Task Hierarchies for Human-Robot Collaboration
Abstract:
Robots capable of collaborating with people provide tremendous value, and bring with them the potential to revolutionize a wide array of industries ranging from healthcare to education to manufacturing. Particularly in domains where modern robots are ineffective, we wish to leverage human-robot teaming to improve the efficiency, ability, and safety of human workers. Central to building systems capable of human-robot teaming are the problems of teammate goal inference and multi-agent coordination, both of which can be extremely challenging without a priori structural knowledge.
In this talk, I describe a novel approach for constructing task hierarchies with direct applications to facilitating goal inference and multi-agent coordination. By leveraging properties of a task’s underlying Semi-Markov Decision Process representation and using edge contraction techniques on its conjugate graph, we present a recursive hierarchy construction algorithm that is universally applicable to finite-horizon tasks. The talk will include a discussion of applications for these hierarchies within the collaborative robotics domain, including preliminary work on methods of estimating a collaborator’s task-relevant mental state, as well as applications within tele-operated and assistive robotics.