Cross-Entropy open loop planning applied to noisy humanoid walking

In this experiment, both the domain and the generative model are noisy.  Noise is introduced by applying  a uniformly distributed noise to one randomly selected joint at each time step.  The noise is not “well-behaved” as it is not gaussian, and not zero-mean.  I believe the results look better than the previous post, where the domain was noisy, but the generative model was not.