We implemented many classes of Sparse Reward algorithms in Gym Fetch environment including Reward Shaping, Imitation Learning, Curriculum Learning, Hindsight Experience Replay, Curiosity-Driven Exploration, Hierachical Reinforcement Learning. This work is for better understanding of sparse reward algorithms.
Our code is based on https://github.com/andrew-j-levy/Hierarchical-Actor-Critc-HAC- and we have changed a lot on code simplification and content richness.
- DDPG:
python main.py --retrain
- Reward Shaping:
python main.py --retrain --rtype dense
- Curriculum Learning:
python main.py --retrain --curriculum 2
- Imitation Learning:
python main.py --retrain --imitation --imit_ratio 1
- Hindsight Experience Replay:
python main.py --retrain --her
- Forward Dynamic:
python main.py --retrain --curiosity
- Hierachical DDPG:
python main.py --retrain --layers 2
- Test the latest saved checkpoint:
python main.py --test
if using HDDPG, you should use :
python main.py --test --layers 2
- Save demostrations for imitation learning:
python main.py --retrain --her --save_experience