Örebro Studies in Technology 87 I ÖREBRO 2019
2019JA
SM
IN
G
RO
SI
N
GE
R
On Ma
king R
ob
ots P
ro
ac
tiv
e
jasmin grosinger received her MSc degree in Medical Informatics from Technische Universität Wien (Vienna University of Technology), Austria, in 2012. She became a doctoral student at the Center for Applied Autonomous Sensor Systems (AASS) at Örebro University, Sweden, in 2013. Her research interests include proactivity, goal reasoning, and knowledge representation and reasoning in robotics. We are increasingly exposed to Artificial Intelligence (AI) systems via the pro-ducts and services we use every day. Researchers and industry agree that future AI systems will have to exhibit proactive behavior if they are to be accepted in human-centered environments. Proactivity is the ability to generate one’s own goals and pursue them. Proactive behavior is opposed to reactive behavior which is merely responding to external events and explicit requests (by the user). In this thesis we focus on AI systems that are intelligent robots, hence, the questions addressed are: Can we make robots proactive? Can we achieve preferable outcomes by proactive behavior? Making robots proactive requires making them capable of reasoning about the world around them (including humans), as well as predicting the evolution of the world over time, knowing about the robot’s own capabilities and knowing what is desirable. This thesis proposes a formal model which comprises all of the above. To achieve proactive behavior the model is made computational in an algorithm called equilibrium
maintenance. We formally show that equilibrium maintenance leads to
proac-tive behavior and preferable outcomes. This finding is confirmed empirically by experiments in simulation in the domain of robot disaster management and in a real physical robot system in the domain of domestic robotics.
issn 1650-8580 isbn 978-91-7529-312-7