RST video: Above is a video showing an episode that resulted in a return of of 2562, which is near the reasonable solution threshold (RST).  This policy exhibits a reasonable gait and an ability to move the simulated robot forward at a reasonable and consistent speed.

To build more intuition about the choice of RST, above is a video showing an episode (from a different policy) that resulted in a return of of 1571--below the RST but well above a random policy's objective.  This policy exhibits an ability to consistently move the agent forward, but it is visibly inefficient and awkward.  A policy which gives returns near the RST visibly and significantly outperforms a policy like this one.