Maze¶


Action Space	`Discrete(3)`
Observation Space	`Box(0, 255, (60, 80, 3), uint8)`
Creation	`gymnasium.make("MiniWorld-Maze-v0")`

Description¶

Maze environment in which the agent has to reach a red box. There are a few variants of the Maze environment. The MazeS2 environment gives you a 2x2 maze and the MazeS3 environment gives you a 3x3 maze. The MazeS3Fast also gives you a 2x2 maze, but the turning and moving motion per action is larger.

Action Space¶

Num	Action
0	turn left
1	turn right
2	move forward

Observation Space¶

The observation space is an ndarray with shape (obs_height, obs_width, 3) representing an RGB image of what the agents see.

Rewards¶

+(1 - 0.2 * (step_count / max_episode_steps)) when red box reached and zero otherwise.

Arguments¶

env = gymnasium.make("MiniWorld-Maze-v0")
# or
env = gymnasium.make("MiniWorld-MazeS2-v0")
# or
env = gymnasium.make("MiniWorld-MazeS3-v0")
# or
env = gymnasium.make("MiniWorld-MazeS3Fast-v0")