Our goal is to develop collaborative agents (software or robots) that can efficiently communicate with their human teammates. Key threads involve designing algorithms for inferring human behavior and for decision-making under uncertainty.
Effective communication between teammates is critical to the success of collaboration, including human-machine collaboration. However, communication typically requires both the human and machine to expend resources; and too much or too little communication has the potential to adversely affect task performance. This project focuses on developing inference and planning algorithms for collaborative agents (software or robots), in order to enable them to efficiently share information with humans. Our approach involves use of tools from inference and machine learning and insights from cognitive science --- to create models and algorithms for the collaborative agent’s decision-making. Our communication algorithm allows the collaborative agent to trade cost and benefits of sharing information with its teammates, to achieve efficient communication. Immediate application of this research include collaborative manufacturing, disaster response and personal digital assistants.