One of the main challenges in Artificial Intelligence is the problem of abstracting high-level models directly leveraging the interaction between the agent and the environment, where such interaction is typically performed at low-level through the agent’s sensing and actuating capabilities. Such information abstraction process indeed reveals invaluable for high-level planning, as it allows to make explicit the causal relations existing at the high-level which would otherwise remain hidden at low-level. In this respect, some interesting work has been done in the recent literature. For instance, in [3] an algorithm is presented for automatically producing symbolic domains based on the Planning Domain Definition Language (PDDL, see [2]), starting from a set of low-level skills represented in the form of abstract subgoal options. The contribution of this work is the following. First, we extend the scope of the information abstraction procedure proposed in [3] by directly linking it to a robotic architecture (GRAIL – Goal-Discovering Robotic Architecture for Intrinsically-Motivated Learning; [6]) able to autonomously discover goals and learn skills based on intrinsically motivated learning algorithms [1]. Such skills are then used as input for the subsequent abstraction process, thus creating an automated information processing pipeline from the low-level direct inter- action of the agent with the environment, to the corresponding high-level PDDL domain representation of the environment. Second, given a set of low-level domains in which GRAIL operates, we carry out an analysis on the features of the produced abstract PDDL representations depending on the categorization capabilities of the classifiers used for the production of the symbolic vocabulary, thus shedding some light on a number of interesting correlations between low-level generalization capabilities of the abstraction procedure and the quality of the produced PDDL high-level representations. Third, we have tested the overall system within the context of an ESA project called IMPACT (see the acknowledgement footnote for more details), in particular within two different space exploration scenarios, demonstrating the advantages of enhancing the wellknown Sense-Plan-Act (SPA) paradigm for controlling robotic systems [4] with autonomous skill-learning capabilities in the general context of space exploration.