This is based on the opendog from here:
https://github.com/robotlearn/pyrobolearn