A Flexible Task Knowledge Representation for Service Robots

Robots that are designed to act in human-centered environments put up the need for a flexible and adaptive representation of task knowledge. This results on the one hand from the continuously changing and hardly predictable state of an environment that is populated with humans and robots. On the other hand, a task knowledge description of a robot which cooperates with humans has to be adaptable and extendable. This paper presents a task knowledge representation for service robots called Flexible Programs (FP) and the environment for execution of FPs. Flexible Programs can be created manually, or by using the results of machine learning approaches like Programming by Demonstration. It is possible to change, adapt and extend this task knowledge at runtime