Robotic Auxiliary Losses

This is a baselines DDPG implementation with added Robotic Auxiliary Losses

To install, first set up the baselines repository on your workstation (https://github.com/openai/baselines/)

Afterwards, go to baselines/baselines/ddpg and clone this repository (NOTE: this will overwrite the original ddpg implementation!)

Example training run:

python -m baselines.ddpg.main --env-id Humanoid-v2 --aux-tasks 'repeat' 'caus' --log-dir /tmp/RobAuxLossTest

DDPG

• Original paper: https://arxiv.org/abs/1509.02971
• Baselines post: https://blog.openai.com/better-exploration-with-parameter-noise/
• python -m baselines.ddpg.main runs the algorithm for 1M frames = 10M timesteps on a Mujoco environment. See help (-h) for more options.