Flexible Demonstration Learning System for Variable Number of Robots

In this paper, we present flexMLfD, a robot independent and task independent demonstration learning system that supports a variable number of robot learners. Our approach is based on the Confidence-Based Autonomy (CBA) demonstration learning algorithm, which provides the means for a single robot to learn a task policy through interaction with a human teacher. The generalized representation and adjustable robot autonomy provided by the CBA algorithm enable the flexible system design and multi-robot learning capabilities of flexMLfD. Building upon the CBA single-robot algorithm, we contribute a robot-independent modular software architecture for multi-robot learning, interaction and control. To highlight the generality of the presented learning system, we present three example domains, each utilizing from two to seven real robots.