PUB - Publikationen an der Universität Bielefeld

For the development of multi-purpose robots that can operate in a wide range of different situations we will need sophisticated behavioral building blocks to compose the desired performances. In this paper we argue for a behavior modeling framework which provides specific behavior interfaces for implementing robot skills that stay close to evaluation cycles and to rich fused sensory data. This way we ensure reusability and facilitate what we call Informed Strategies. As a specific application we deploy the framework to an object search task for a domestic service robot. The presented behavior involves an attention mapping mechanism based on 2D and 3D visual cues. We show the advantages of the proposed approach by conducting an evaluation in a real-world apartment scenario as well as by successfully taking part in the RoboCup@HOME competition.