An Integrated Planning and Learning Framework for Human-Robot Interaction

International audience Assistive robot systems require a robot to interact closely with people and to perform joint human-robot tasks. Interaction with humans comes with additional challenges to those in other real-world scenarios. Robot plans must be especially flexible and take into account human abilities and preferences. For providing this level of flexibility, we propose a framework for trans-formational reactive planning that includes the capability to learn models of the human during plan execution. We show how this framework can be extended to the special requirements of human-robot interaction.