h1

TY  - THES
AU  - Schiffer, Stefan
TI  - Integrating Qualitative Reasoning and Human-Robot Interaction for Domestic Service Robots
PB  - RWTH Aachen University
VL  - Dissertation
CY  - Aachen
M1  - RWTH-2015-00660
SP  - VIII, 224 S. : Ill., graph. Darst.
PY  - 2015
N1  - Prüfungsjahr: 2014. - Publikationsjahr: 2015
N1  - RWTH Aachen University, Diss., 2014
AB  - The last decade has seen an increasing interest in domestic service robots. Particularchallenges for such robots especially when performing complex tasks are deliberation,robust execution of actions, and flexible human-robot interaction. Despite progress inqualitative reasoning and human-robot interaction their integration is an open issue.In this thesis, we build on an existing cognitive mobile robot platform and make aseries of contributions to integrate qualitative representations, high-level reasoningand human-robot interaction for an intelligent domestic service robot. We start byintroducing the domestic service robotics domain and parts of the ROBOCUP@HOMEmethodology that we contributed to. Before we can actually turn to our main focus, weequip the system with a set of basic capabilities that are required for a service robotin human environments. As a bridge between perception and symbolic reasoning weprovide a semantic mapping scheme that allows to centrally manage information aboutthe environment. With a novel hierarchical object recognition method we are furtherable to classify even yet unseen objects.Then we move on to the main contributions of this thesis. First, we extend the robot withimportant modes for human-robot interaction by adding components for speech, face,and gesture recognition as well as for speech synthesis and a virtual facial display. Forthe speech input we proceed with a simple form of natural language understanding thatallows a limited form of error recovery. Second, we introduce qualitative representationsand control to our high-level control system. After integrating a general account forqualitative information based on fuzzy sets into our high-level language we also addmeans to specify and use fuzzy controllers for behaviour specification. Then we focuson spatial data and provide a formalization that allows for representing and reasoningwith qualitative positional information in our high-level language. Lastly, we increasethe robustness of the robot against internal errors and add to the flexibility in dealingwith possibly faulty external input. We integrate a basic form of self-maintenance thatallows the robot to recover from internal errors by itself.
LB  - PUB:(DE-HGF)11
UR  - https://publications.rwth-aachen.de/record/462448
ER  -