Communication for Enhancing Human-Robot Collaboration
Abstract
?While robotics has made considerable strides toward more robust and adaptive manipulation, perception, and planning, robots in the near future are unlikely to be as dexterous, competent, and versatile as human workers. Rather than try to create fully autonomous systems that accomplish tasks independently, we believe that a more practical approach is to construct robots that work side-by-side with a person, allowing the human worker to concentrate on the tasks to which they are best suited while obtaining assistance on the tasks where the robot is as capable or perhaps more capable. A collaborative robot might hand you a screwdriver just when you need it, stabilize a piece of lumber while you drill at one end, clean up the workspace in anticipation of your next task, alert you to the fact that an assembly is wobbly, or help you update a checklist. We propose to take the next step toward collaborative robot assistants by integrating robotic task representations with the knowledge resources of language-endowed, mindreading agents. We envision robots that are able to respond to requests for role allocation or assistance, to query the human partner to resolve ambiguity, and to respond proactively to clarify its own actions when (and only when) it judges that they would appear to be ambiguous or unclear to the human user. Our approach to this task highlights four common threads that permeate all of our work: 1. Imperfect agents in imperfect worlds: Our work is based on the development of practical, near-future systems. 2. Models of self and other: One shortcoming of existing robotic systems is that they cannot adequately reason about their own capabilities and limitations. We focus on systems that build explicit, modifiable representations of self. 3. Integration of knowledge representations: In order to be successful, we must combine a set of representation systems that have been developed in different research communities including task, dialogue, and self models. 4. Dealing with off-task activities and interruptions: Real-world deployments are marked by a multitude of interruptions and off-task actions, remarks, and distractions. Our systems must operate with these partial, interruptible scenarios. We will develop two lines of research that are critical to maintaining grounded communication for a collaborative robot: (1) using communication to reduce uncertainty, and (2) aligning and correcting mental models between the robot and the human user. These two research thrusts require a shared cognitive representation which combines an adaptive model of the robotÕs own capabilities that can be explicitly manipulated by the reasoning system, a task model that can represent both the activities of the robotÕs own plans and the joint plans of the human partner, a metacognitive model of the beliefs, intentions and biases of self and other agents as well as (meta)knowledge for assessing risk, self-confidence and other parameters essential for efficient decision-making.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jun 10, 2016
- Source ID
- N000141612212
Entities
People
- Brian Scassellati
Organizations
- Office of Naval Research
- United States Navy
- Yale University