Planning Challenges in Human-Robot Teaming: An Integrated Exploration of Representations, Algorithms and Human Factors
Abstract
Statement of Work:Investigate the challenges of modeling of humans and the resulting planning faced by a robot in scenarios where humans are in the loop. Develop effective frameworks for handling these challenges. Validate the developed approaches in different realistic domains and investigate how general these methods are.Objective:The long term goal is to develop foundations of human-robot teaming. The objectives of this effort are to understand the challenges of modeling of humans and the resulting planning faced by a robot in scenarios where humans are in the loop, and to develop effective frameworks for handling these challenges. Approach:This effort is a collaboration of Rao Kambhampati and Nancy Cooke. The proposal focuses primarily on cognitive aspects of human-robot teaming. Effective teaming requires the robot to model the intentions and capabilities of the human teammates (like human-human teaming) and take these models into account while planning its own behavior. They propose to address several critical challenges that come up in supporting cognitive teaming that include (a)automatically constructing and learning models of human teammates, (b) using these models to guide the interactions between the robots and humans to facilitate effective teaming (while ensuring human preferences and safety), and (c) evaluating the developed models and interaction approaches in realistic human-robot teaming scenarios (e.g. urban search and rescue). They propose to take an integrated approach to understanding and addressing these challenges. In particular, use the lessons from human factor studies in human-human as well as human-robot studies to develop effective desiderata, representations and algorithms for planning in human-robot teaming, and systematically evaluate their effectiveness. An important challenge in this effort is various types of uncertainty that will lead to partial models. The PIs will continue extending their pioneering work in model-lite planning and plan recognition.Overall Merit and ONR Mission/Relevance:This effort in human-robot teaming is of high relevance to ONR Autonomy Focus Area.Human-robot teaming is an important and challenging research area with little systematic work heretofore. This effort addresses fundamental problems in this area and is expected to lead to advances in modeling humans, planning and plan recognition in uncertain, weakly modeled domains.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 30, 2016
- Source ID
- N000141612892
Entities
People
- Subbarao Kambhampati
Organizations
- Arizona State University
- Office of Naval Research
- United States Navy