Content area
Full text
We present an innovative approach for largescale data collection in human-robot interaction research through the use of online multiplayer games. By casting a robotic task as a collaborative game, we gather thousands of examples of human-human interactions online, and then leverage this corpus of action and dialogue data to create contextually relevant social and task-oriented behaviors for human-robot interaction in the real world. We demonstrate our work in a collaborative search and retrieval task requiring dialogue, action synchronization, and action sequencing between the human and robot partners. A user study performed at the Boston Museum of Science shows that the autonomous robot exhibits many of the same patterns of behavior that were observed in the online data set and survey results rate the robot similarly to human partners in several critical measures.
We envision the need for robots to be not only functional, but adaptable, robust to the diversity of human behaviors and speech patterns, and capable of acting in a both task and socially appropriate manner. Natural and diverse human-robot interaction (HRI) of this kind has been a long-standing goal for robotics research, and a broad range of approaches have been proposed for the development of robots that support diverse interactions. Among proposed techniques, variants that are dependent on hand-coded rule sets and probabilistic single-task policy learning methods have proven to be too brittle for interactive applications, failing to generalize over the diversity of possible inputs. Such systems typically force the user to adapt their method of interaction to fit the coded requirements of the robot.
A different approach to creating more humanlike robotic systems has focused on imitating human cognitive processes by developing large scale cognitive architectures that support many modalities and interaction styles. While such systems have been shown to successfully support a broad range of interactions, they rely heavily on precoded data. For example, dialogue responses are typically limited to only one or two dozen phrases, which pales in comparison to the diversity of human speech.
We believe that in order for robotic systems to become a truly ubiquitous technology, robots must make sense of natural human behavior and engage with humans in a more humanlike way. Robots must become more like humans instead of forcing humans to be more...