In order to solve the partially observed LunarLander-v2 OpenAI Gym environment, We tried four different methods :

1. SARSA
2. Q-Learning
3. DQN
4. DQN bonus

And we have tested their performance in two different situation, blind are width = 0.2 / 0.4.

You may encounter some problem with the environment LunarLander-v2. This link could be helpful.

https://stackoverflow.com/questions/64161280/rl-problem-on-colab-for-gym-envs-box2d-has-no-attribute-lunarlander