Learning a Symbolic Planning Domain through the Interaction with Continuous Environments