INI MultiGrid

Cooperative exploration multi-agent environments with view_size=7.

Install:

pip install -e 3rd_party/environments/multigrid-ini/

Paradigm:

Multi-agent (simultaneous)

Stepping:

SIMULTANEOUS

Environment	Description
MultiGrid-Empty (5×5 to 16×16)	Cooperative navigation in empty grids
MultiGrid-BlockedUnlockPickup	Blocked door puzzle requiring cooperation
MultiGrid-LockedHallway (2/4/6 Rooms)	Multi-room exploration with locked doors
MultiGrid-Playground	Playground with various interactive objects
MultiGrid-RedBlueDoors (6×6, 8×8)	Cooperative door-colour sequencing

Citation

@article{oguntola2023theory,
  title        = {Theory of mind as intrinsic motivation for multi-agent reinforcement learning},
  author       = {Ini Oguntola and Joseph Campbell and Simon Stepputtis and Katia Sycara},
  journal      = {arXiv preprint arXiv:2307.01158},
  year         = {2023}
}