Unlock#

Unlock

Action Space

Discrete(7)

Observation Space

Dict('direction': Discrete(4), 'image': Box(0, 255, (7, 7, 3), uint8), 'mission': MissionSpace(<function UnlockEnv._gen_mission at 0x7f813f47d040>, None))

Reward Range

(0, 1)

Creation

gymnasium.make("MiniGrid-Unlock-v0")

Description#

The agent has to open a locked door. This environment can be solved without relying on language.

Mission Space#

“open the door”

Action Space#

Num

Name

Action

0

left

Turn left

1

right

Turn right

2

forward

Move forward

3

pickup

Unused

4

drop

Unused

5

toggle

Toggle/activate an object

6

done

Unused

Observation Encoding#

  • Each tile is encoded as a 3 dimensional tuple: (OBJECT_IDX, COLOR_IDX, STATE)

  • OBJECT_TO_IDX and COLOR_TO_IDX mapping can be found in minigrid/core/constants.py

  • STATE refers to the door state with 0=open, 1=closed and 2=locked

Rewards#

A reward of ‘1 - 0.9 * (step_count / max_steps)’ is given for success, and ‘0’ for failure.

Termination#

The episode ends if any one of the following conditions is met:

  1. The agent opens the door.

  2. Timeout (see max_steps).

Registered Configurations#

  • MiniGrid-Unlock-v0