AIBO as a needs-based companion dog José M. Blanco Calvo, Drago Datcu and Léon J. M. Rothkrantz Man-Machine Interaction Group Delft University of Technology Mekelweg 4, 2628CD Delft, The Netherlands E-mail: L.J.M.Rothkrantz@tudelft.nl KEYWORDS AIBO-robot, human-computer interaction, companion robot, needs-based, personality modeling, nPME model. ABSTRACT In this paper we describe the architecture that allows the modeling of an emotionally intelligent robotic companion dog. We chose to implement this architecture for AIBO, which is a dog-like autonomous robot, developed by Sony. AIBO was developed as an entertainment robot and its ‘mind’ is programmable. The focus of this paper is on creating a complex personality model that provides a realistic needs-oriented behavior for robotic companion dogs. The personality model is an extension of the nPME model and implies that needs play the most important role and influence the behavior of the AIBO robot in every situation. Personality, Mood and Emotions are three layers that, in combination with the needs, make AIBO show a real dog- like behavior. The existing personality models need to be modified, adaptated and improved so as to handle the unknown and dynamic environment in which AIBO performs. So, as part of the system design, the importance of incoming events has been considered. A modification in this architecture modeling the function of amygdale has been introduced to speed up the resolution of some situations. The implemented architecture is also designed to be suitable for future extensions in the model. The prototype has been tested in different scenarios using an AIBO robot. An extensive user study has been carried on involving several students and workers to test the realism, emotional responses and event coherency in robot’s behavior. INTRODUCTION The societal relevance of intelligent robots is increasing nowadays covering a wide range of applications: from entertainment robots such as conversational partners, soccer players, or companion dogs, to robots that provide independent living support for elderly users in basic activities such as mobility or household maintenance (Heerink et al. 2006). Security tasks can also be performed by surveillance robots like AIBO watchdogs (Yang et al. 2006). Studies conducted by Masahiro Fujita prove that AIBO seems to be a good partner for users, having a positive effect on their emotional state (Fujita 2004). Robots have shown their utility even in medical domains by aiding the diagnosis and therapy of diseases as autism (Scassellati 2005). According to Fujita the current implementation of AIBO software provided by Sony uses behaviors that come from a “manually designed database” (Arkin et al. 2003). The next steps would be a learning-based systems or evolving systems able to create new behaviors through the continuous interaction with both human beings and environment. Mature systems would play also an important role in the future of entertainment robotics. Figure 1: AIBO robot as a companion dog This project focuses on the idea of improving interaction between man and machine (Figure 1) through the creation of a biological inspired model of human mind. This artificial brain must act coherently with the environment, drawing conclusions according to the different situations that can arise during a day. To achieve this, it is necessary to create a personality model capable of reasoning about a certain situation to perform complex actions in every context. So there must be a whole coherence between events, actions, and emotions. The situation awareness process would include a reasoning system that combines the different personality parameters (PME) with the levels of certain needs (n). Triggered by incoming events it should be able to perform a set of actions to satisfy the needs, in the similar way a real dog would do. This paper is organized as follows: first we will introduce the original nPME model (Dobai and Rothkrantz 2005); then our extension will be explained; next section will describe the architecture implemented to test the model; and finally, a summary of the evaluation method and results in the user study will be provided.