1 The role of Surprise, Curiosity and Hunger on Exploration of Unknown Environments Populated with Entities Luís Macedo, Amílcar Cardoso Abstract— This paper describes an approach based on affect to the problem of exploring unknown environments populated with entities by agents. To this end, a multi-agent system based on the notion of affect as well as on the Belief-Desire-Intention (BDI) model was used. The affective component of the agents is confined to the motivations that are usually associated to exploratory behavior: surprise, curiosity and hunger. An experiment that evaluates the role of these motivations in exploration performance is presented. Index Terms—Exploration of unknown environments, Curiosity, Hunger, Surprise. I. INTRODUCTION XPLORATION gathers information about the unknown. Exploration of unknown environments by artificial agents (usually mobile robots) has actually been an active research field. The exploration domains include planetary exploration (e.g., Mars or lunar exploration), search for meteorites in Antarctica, volcano exploration, map-building of interiors, etc. The main advantage of using artificial agents in those domains instead of humans is that most of them are extreme environments making exploration a dangerous task for human agents. However, there is still much to be done especially in dynamic environments as those real environments mentioned above. Those real environments consist of objects. For example, office environments possess chairs, doors, garbage cans, etc., cities comprise several kinds of buildings (houses, offices, hospitals, churches, etc.), cars, etc. Many of these objects are non-stationary, that is, their locations may change over time. This observation motivates research on a new generation of mapping algorithms, which represent environments as collections of objects. Moreover, the autonomy of agents still needs to be improved, as happens for instance in planetary exploration which is still too human dependent. Several exploration techniques have been proposed and tested either in simulated and real, indoor and outdoor environments, using single or multiple agents (for an overview see e.g., [1-3]). In human beings, exploration has been closely connected with motivation (including emotion and drives). This relationship between exploration and motivation has been defended for a long time in the realms of psychology and ethology. James’ concept of selective attention [4], Freud’s term cathexis [5], and McDougall’s notion of curiosity instinct [6] are foundation thoughts for the relationship between motivation and exploratory behaviour. Therefore, a reasonable approach is to model artificial agent’s exploration according to humans, i.e., in a human-like fashion by assigning artificial agents mentalistic qualities such as emotion and motivation, beliefs, intentions, and desires. Actually, there is one primary reason for taking the way humans explore the environment as a reference: the problem of modelling exploration in humans has already been successfully solved by millions of years of evolution. Yet, in general, a lot of barriers have been found to incorporate models of emotion in artificial agents. Research in AI has almost ignored this significant role of emotions on reasoning, and only recently this issue was taken seriously (e.g., [7-18]) mainly because of the recent advances in neuroscience, which have given evidence that cognitive tasks of humans, and particularly planning and decision-making, are influenced by emotion [19]. This work was supported in part by the PRODEP. L. Macedo is with the Instituto Superior de Engenharia de Coimbra, Quinta da Nora, 3030 Coimbra Portugal (e-mail: lmacedo@isec.pt) and with the Centre for Informatics and Systems of the University of Coimbra, Department of Informatics Engineering, Polo II, 3000 Coimbra Portugal (e-mail: macedo@dei.uc.pt). A. Cardoso is with the Centre for Informatics and Systems of the University of Coimbra, Department of Informatics Engineering, Polo II, 3000 Coimbra Portugal (e-mail: amilcar@dei.uc.pt). In this paper we describe an approach based on affect to the problem of exploring unknown environments by agents. We developed a multi-agent system based on the notion of affect as well as on the BDI model, which was used as a platform to develop the application to exploration of unknown environments with affective agents. Primary relevance is given to the architecture of an affective agent and especially to its affective module and its influence on exploratory behavior. We confined the set of motivations to those that are more related with exploratory behavior in humans [20]. The next section describes the approach for exploring unknown environments with affective agents. Section 3 presents an experiment that was conducted to evaluate that approach. Finally, we present conclusions. E