Understanding human-robot teams in light of all-human teams: Aspects of team interaction and shared cognition

As robots become more autonomous, their roles shift from being operated and controlled by humans to interactively teaming with humans. The current research focuses on how human operators can effectively team with autonomous urban search and rescue agents in a dynamic and complex task environment. To do so, we empirically examined how shared cognition and restricted language capabilities impacted performance of human-robot dyad search teams using a simulated Minecraft task environment. In order to examine the effects of shared mental models and language the following modified conditions were applied: (1) participants were either able to communicate using natural language or the internal participant's communication was limited to three-word utterances; and (2) shared mental models were manipulated by either the internal participant being made fully aware of the external participant's restricted representation of the environment and inaccurate map or the internal was unaware of these challenges. The primary findings from this study are: (1) teams in the natural language and shared mental model conditions performed better than teams in the limited language and restricted model conditions; (2) when the internal participant was unaware of the challenges of the external, the external perceived higher workload than when there was a shared mental model; (3) teams with natural language and shared mental model demonstrated more predictable behavior than the other teams; (4) some amount of systems' predictability was good but too much predictability was not good. Overall, these results indicate that effective team interaction and shared cognition play an important role in human-robot dyadic teaming performance.