RL Based Walking

[pejotejo]

Try to rethink our walking using Residual Policy Learning.

• one could set up a Mujoco Enviroment with an exact controller model (system identification)
• the alternative approach is to use the Walking engine as Animation Guidance and give penalties on instability