Closed mkbera closed 6 years ago
This question is from a theoretical perspective. Do the roboschool environments have Markov property? I mean are they MDP? If yes, is the state representation returned by env.step() the complete state representation?
Mostly yes, you have x and x dot. But some things are not observable: coordinates, absolute orientation.
This question is from a theoretical perspective. Do the roboschool environments have Markov property? I mean are they MDP? If yes, is the state representation returned by env.step() the complete state representation?