StARformer: Transformer With State-Action-Reward Representations for Robot Learning | IEEE Journals & Magazine | IEEE Xplore