/imitation-learning

Imitation learning algorithms

Primary LanguagePythonMIT LicenseMIT

IL

MIT License

Imitation learning algorithms (with PPO [1]):

python main.py --imitation [AIRL|BC|GAIL|GMMIL|RED]

Results

PPO

Train Test
ppo_train_returns ppo_test_returns

AIRL(s, a)

Train Test
airl_sa_train_returns airl_sa_test_returns

AIRL(s)

Train Test
airl_s_train_returns airl_s_test_returns

BC

Train Test
bc_test_returns bc_test_returns

GAIL(s, a)

Train Test
gail_sa_train_returns gail_sa_test_returns

GAIL(s)

Train Test
gail_s_train_returns gail_s_test_returns

GMMIL(s, a)

Train Test
gmmil_sa_train_returns gmmil_sa_test_returns

GMMIL(s)

Train Test
gmmil_s_train_returns gmmil_s_test_returns

RED(s, a)

Train Test
red_sa_train_returns red_sa_test_returns

RED(s)

Train Test
red_s_train_returns red_s_test_returns

Acknowledgements

References

[1] Proximal Policy Optimization Algorithms
[2] Adversarial Behavioral Cloning
[3] Learning Robust Rewards with Adversarial Inverse Reinforcement Learning
[4] Efficient Training of Artificial Neural Networks for Autonomous Navigation
[5] Generative Adversarial Imitation Learning
[6] Imitation Learning via Kernel Mean Embedding
[7] InfoGAIL: Interpretable Imitation Learning from Visual Demonstrations
[8] Primal Wasserstein Imitation Learning
[9] Random Expert Distillation: Imitation Learning via Expert Policy Support Estimation