Letícia Maria Friske, Carlos Henrique Costa Ribeiro.
In this paper, we present a study about the use of options policies in Reinforcement Learning for navigation tasks of autonomous mobiles robots. We considered the use of options based on action policies and sequences, respectively O.... and OS options. The first ones, also referred in the literature as macro-actions or macro-operators, correspond to sequences of low-level actions, characterized by an exclusive dependency on the visited state, whilst OS options are sequences of actions dependent on the history of states visited since its activation. Experimental results show that the actionÂs degree of redundancy in the options influences the exploratory capability, an essential feature in Reinforcement Learning. Clique no link abaixo para buscar o texto completo deste trabalho na Web: Buscar na Web