NetHack-2021

AIM

Designing an agent which can navigate the procedurally generated, ascii dungeons of Nethack :trophy:

Todos (Feel free to add if you feel anything is missing):heavy_plus_sign:

• [ ] Develop a policy that favors the agent almost all the time.
• [ ] Model based vs Model free, battle for best
• [ ] Mix N Match different policies
• [ ] Make three independent agents where the third is a mix of the two.
• [ ] If using NN, optimize it using Neural Magic - Reduces inference time
• [ ] Maybe use stable baselines

In Progress :pick:

• Working on creating a player using path finding algorithms (SSSP, ASSP etc) :boom:

Done :heavy_check_mark:

• Not done

  Gotchas

• Figure out how to deal with secret passage (Tentatively dealt with!) :black_small_square: :small_red_triangle_down:

Algorithms (Add picture)

Using

Testings

• Add tests