Vishal Soni and Satinder Singh
Humans frequently engage in activities for their own sake rather than as a step towards solving a specific task. During such behavior, which psychologists refer to as being intrinsically motivated, we often develop skills that allow us to exercise mastery over our environment. Singh, Barto, and Chentanez (2004) have recently proposed an algorithm for intrinsically motivated reinforce- ment learning (IMRL) aimed at constructing hierarchies of skills through self-motivated interaction of an agent with its environment. While they were able to successfully demonstrate the utility of IMRL in simulation, we present the first realization of this approach on a real robot. To this end, we implemented a control architec- ture for the Sony-AIBO robot that extends the IMRL algorithm to this platform. Through experiments, we examine whether the Aibo is indeed able to learn useful skill hierarchies.