A deliberation layer for instantiating robot execution plans from abstract task descriptions

We present an application of Hierarchical Task Network (HTN) planning to create robot execution plans, that are adapted to the environment and the robot hardware from abstract task descriptions. Our main intention is to show that different robotic platforms can make use of the same high level symbolic task description. As an off-the-shelf planning component, the SHOP2 HTN planner is adopted. All the domain knowledge is encoded in the Web Ontology Language (OWL) and stored in a world wide accessible database, which allows multiple systems to reuse and improve upon this knowledge. For task execution, the execution plan is generated using the CRAM plan language (CPL). We demonstrate the functionality of the system in executing a pick-and-place task in a simulated environment with two different service robots, the TU/e Amigo robot prototype and the Fraunhofer IPA Care-O-Bot 3. The experiment shows that although the robots differ in hardware capabilities, the use of HTN planning adds information that is crucial for successful task execution and enables both systems to successfully execute the instructed task.