FACT: A Full-body Ad-hoc Collaboration Testbed for Modeling Complex Teamwork
This addresses the need for better modeling of natural, ad-hoc collaborative behaviors in robotics research, though it is incremental as it provides a testbed rather than a novel method.
The paper tackles the lack of shared resources for modeling complex, unstructured human-robot collaborations by introducing FACT, an openly accessible testbed that provides an expansive view of individual and team-based behaviors in co-located teamwork, based on preliminary explorations with teams of various sizes.
Robots are envisioned to work alongside humans in applications ranging from in-home assistance to collaborative manufacturing. Research on human-robot collaboration (HRC) has helped develop various aspects of social intelligence necessary for robots to participate in effective, fluid collaborations with humans. However, HRC research has focused on dyadic, structured, and minimal collaborations between humans and robots that may not fully represent the large scale and emergent nature of more complex, unstructured collaborative activities. Thus, there remains a need for shared testbeds, datasets, and evaluation metrics that researchers can use to better model natural, ad-hoc human collaborative behaviors and develop robot capabilities intended for large scale emergent collaborations. We present one such shared resource - FACT (Full-body Ad-hoc Collaboration Testbed), an openly accessible testbed for researchers to obtain an expansive view of the individual and team-based behaviors involved in complex, co-located teamwork. We detail observations from a preliminary exploration with teams of various sizes and discuss potential research questions that may be investigated using the testbed. Our goal is for FACT to be an initial resource that supports a more holistic investigation of human-robot collaboration.