esalcort / project_rl

Try RL policies for duckietown

project_rl

Try RL policies for duckietown

We trained the policies on google colab. See how at Run_policies.ipynb.

PPO

The PPO models are saved at:

  ./PPO-PyTorch/

To run PPO for testing, download from drive (if trained with Run_policies.ipynb) or copy into preTrained:

cd PPO-PyTorch
cp PPO_continuous_Duckietown-loop_empty-v0.pth ./preTrained/
python3 test_continuous.py

DDPG and SAC:

Follow Duckietown instructions to install all the necessary dependencies To run SAC for training: change directory to gym-duckietown/learning

  python3 -m reinforcement.pytorch.train_reinforcement --model sac

To run DDPG for training: change directory to gym-duckietown/learning

  python3 -m reinforcement.pytorch.train_reinforcement --model ddpg

The SAC and DDPG models are saved at:

  gym-duckietown/learning/reinforcement/pytorch/models

To run SAC for testing: change directory to gym-duckietown/learning

python3 -m reinforcement.pytorch.enjoy_reinforcement --model sac

To run DDPG for training: change directory to gym-duckietown/learning

  python3 -m reinforcement.pytorch.enjoy_reinforcement --model ddpg