v1.0.0

What's Changed

• Support for Unitree H1
• Support tensorboard logging for PPO
• Support for learning exploration noise
• Configuration yaml for environments
• Create RobotBase class that supports
  • PD gains randomization
  • Adding "back-emf" effect
  • "Actuator nets"
• Use pyMJCF for generating scene XML
• Support for using observation history
• Support for imitating a pre-trained policy
• EvaluateEnv class
  • Evaluate any trained policy by using `python run_experiment.py eval --path
  • Episode will be rendered and recorded automatically

... and more

TODO

JVRCWalkEnv and JVRCStepEnv may need some tuning (and bug fixes?) after these changes.

Full Changelog: v0.0.1...v1.0.0