I think the major iteration that will come next is "NEAT in a dream" where a robot or agent online-trains a library of behaviors on an model of the environment constantly updated with the new experiences generated by the dominate behaviors interacting with reality.