Integration and evaluation of a gaming situation for long-term human-robot interaction: playing a game of pairs with Flobi using contextual knowledge

This doctoral thesis proposes a system for long-term Interaction between a robot and a human using a gaming context. A robot plays a game of pairs autonomously with a human. The system was developed to evaluate how to implement social assistive robots during space missions that occur under isolation conditions. The first part of the thesis presents the system as designed and implemented. Described are the different components used to realize autonomous interaction. The study itself was conducted in cooperation with the German Aerospace Center. Throughout the study, the proposed system performed robustly, and without major system failures. The participants interacted with the system continuously, and gave it average ratings for acceptability. No significant extraneous effects, such as those related to novelty were found. Nevertheless, problems with perception and classification led to negative ratings for the system’s competence. The second part of the thesis was motivated by findings from the isolation study. Integrated is a context knowledge system to increase interest in the interactions during long-term use. This made it possible to collect data on past interactions for use with further interactions. The results of a study showed that greater commitment to gaming interaction could be promoted by using context knowledge. In addition, implemented is a remotely controlled version of the system to evaluate, whether a robot without visual or speech recognition problems who played perfectly could influence the way the system was perceived. Results showed that ratings on participant enjoyment while playing with the system decreased when the playing system was too perfect. Additionally, a robot that played too perfectly appeared to promote unfair game-play behaviors in the human, likely as an attempt to cope with disadvantages such as lower memory capacity.