I think in this context (particularly with a very quick skim of the paper for some additional context), it it might be more helpful to think of air "powering" this design in the same way that electricity "powers" things. The focus isn't on the energy source, it's on the structural design of the "robot" itself.
Consider it another way: if their system/model/whatever designed a conventional electrically-powered robot without also designing an electrical generator or batteries etc, would you still discount it as "not being a robot"? The problem might be in our expectation based on the language being used. I might also be full of crap haha, but hopefully that's another perspective to consider.