Evolving Developmental, Recurrent And Convolutional Neural Networks For Deliberate Motion Planning In Sparse Reward Tasks