RAPTOR Plus : Implementing a Robust Systems Architecture foSocial Hierarchical Learning

Much of contemporary robotics research focuses on methods by which robotic agents can operate in environments that have little, if any, human presence. Humanrobot collaboration, where robots perform tasks alongsidea human confederate, has been almost wholly ignored by curren t research. Social Hierarchical Learning is a method by which robots can learn representations for primitive actions, become experts at performing those actions, decompose a complex ta sk into some superset of those skills, and then enter a mixed human-robot environment to begin actively collaborating.The purpose of this independent research project was to aid in th e design and implementation of such a framework, decomposing the complex pipeline described into a set of discrete compon e ts that can be investigated independently of one another or lin ked together to form a robust and flexible system. The result is prototype software for several components in the framework and a detailed architecture for implementing the remaining modules.