/RLSSContinuousControlTutorial

Tutorial on continuous control at Reinforcement Learning Summer School 2017.

Primary LanguagePython

Reinforcement Learning Summer School : Practical Tutorial on RL for Continuous Control

Here we go over:

  • How to setup MuJoCo and openai/rllab
  • How to run basic TRPO and DDPG code
  • The core code snippets in TRPO and DDPG so you can build on top of these algorithms
  • How to create your own modified MuJoCo environment (Multi-task modifications can be pull-requested into gym-extensions)

How to run examples

Run TRPO

cd code; source activate rllab3; python run_trpo.py Hopper-v1

Run DDPG

cd code; source activate rllab3; python run_ddpg.py Hopper-v1

Plotting Results

cd code; python plot_results.py data/progress.csv Hopper-v1 --labels "trpo"

Manual testing of an env and custom env

cd code; python test_manual Hopper-v1
cd code; python test_modified_hopper_env_manually.py