Just writing a quick note to myself after a cursory read yesterday. Paper at a very high level discusses the idea that trajectory libraries that are generated offline can be used for rapid decision making online. The concept here is used by the impressive work by Tassa and Erez for their different forms of real time (or near real time) planning.\
As I said, there isn’t really a particular algorithm that is discussed, and the experimental results are pretty weak. If I read it myself without reading the previous work I mentioned I would probably just dismiss it. But the proof is in the pudding. I should try it out on my stuff.