Designing for robust and effective teamwork in human-agent teams

Book Section

We investigated the impact of team structure, task uncertainty, and information-sharing tools on team coordination and team performance in human-agent teams. In applications such as search and rescue, command and control, and air traffic control, operators in the future will likely need to work in teams together with robots. It is critical to understand how these teams could be robust against uncertainty and what influences team performance. We conducted two experiments in which teams of three operators controlled simulated heterogeneous robots on the same testbed. Experiment 1 investigated the impact of team structure and uncertainty of task arrival processes on team coordination and performance. Experiment 2 explored the usage of information-sharing tools under different uncertainty levels. In Experiment 1, it was found that divisional teams were more robust against the uncertainty on task arrival processes. However, this robustness was achieved with an overall worse performance compared to functional teams. Three reasons for the degraded performance were identified, namely duplication on task assignment, under-utilization of vehicles, and infrequent communication. In Experiment 2, it was found that information-sharing tools reduced the duplication on task assignments, improved overall task performance, and reduced workload. These results provide insights for achieving robust and effective teamwork. This goal can be achieved by using a team structure that could adapt to uncertainties together with effective information-sharing tools. These findings could inform the design of robust teams and the development of information-sharing tools to improve teamwork.

Full Text

Duke Authors

Cited Authors

  • Gao, F; Cummings, ML; Solovey, E

Published Date

  • January 1, 2016

Book Title

  • Robust Intelligence and Trust in Autonomous Systems

Start / End Page

  • 167 - 190

International Standard Book Number 13 (ISBN-13)

  • 9781489976666

Digital Object Identifier (DOI)

  • 10.1007/978-1-4899-7668-0_9

Citation Source

  • Scopus