Updated docstrings

miniworld/envs/collecthealth.py

@@ -30,7 +30,7 @@ class CollectHealth(MiniWorldEnv, utils.EzPickle):
     ## Observation Space
 
     The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
-    representing the view the agents sees.
+    representing a RGB image of what the agents sees.
 
     ## Rewards:
 

miniworld/envs/fourrooms.py

@@ -8,7 +8,8 @@ class FourRooms(MiniWorldEnv, utils.EzPickle):
     """
     ## Description
 
-    Classic four rooms environment. The agent must reach the red box to get a reward.
+    Classic four rooms environment. The goal is to reach the red box to get a
+    reward in as few steps as possible.
 
     ## Action Space
 
@@ -21,16 +22,16 @@ class FourRooms(MiniWorldEnv, utils.EzPickle):
     ## Observation Space
 
     The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
-    representing the view the agents sees.
+    representing a RGB image of what the agents sees.
 
     ## Rewards:
 
-    +(1 - 0.2 * (step_count / max_episode_steps)) when red box reached
+    +(1 - 0.2 * (step_count / max_episode_steps)) when red box reached and zero otherwise.
 
     ## Arguments
 
     ```python
-    FourRooms()
+    env = gym.make("MiniWorld-FourRooms-v0")
     ```
 
     """

miniworld/envs/hallway.py

@@ -10,8 +10,8 @@ class Hallway(MiniWorldEnv, utils.EzPickle):
     """
     ## Description
 
-    Environment in which the goal is to go to a red box
-    at the end of a hallway
+    Environment in which the goal is to go to a red box at the end of a
+    hallway within as few steps as possible.
 
     ## Action Space
 
@@ -24,7 +24,7 @@ class Hallway(MiniWorldEnv, utils.EzPickle):
     ## Observation Space
 
     The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
-    representing the view the agents sees.
+    representing a RGB image of what the agents sees.
 
     ## Rewards:
 
@@ -33,7 +33,7 @@ class Hallway(MiniWorldEnv, utils.EzPickle):
     ## Arguments
 
     ```python
-    FourRooms(length=12)
+    Hallway(length=12)
     ```
 
     `length`: length of the entire space

miniworld/envs/maze.py

@@ -9,7 +9,11 @@ class Maze(MiniWorldEnv, utils.EzPickle):
     """
     ## Description
 
-    Maze environment in which the agent has to reach a red box
+    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
 
@@ -22,20 +26,22 @@ class Maze(MiniWorldEnv, utils.EzPickle):
     ## Observation Space
 
     The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
-    representing the view the agents sees.
+    representing a RGB image of what the agents sees.
 
     ## Rewards:
 
-    +(1 - 0.2 * (step_count / max_episode_steps)) when red box reached
+    +(1 - 0.2 * (step_count / max_episode_steps)) when red box reached and zero otherwise.
 
     ## Arguments
 
     ```python
-    MazeS2()
+    env = gym.make("MiniWorld-Maze-v0")
     # or
-    MazeS3()
+    env = gym.make("MiniWorld-MazeS2-v0")
     # or
-    MazeS3Fast()
+    env = gym.make("MiniWorld-MazeS3-v0")
+    # or
+    env = gym.make("MiniWorld-MazeS3Fast-v0")
     ```
 
     """

miniworld/envs/oneroom.py

@@ -9,8 +9,11 @@ class OneRoom(MiniWorldEnv, utils.EzPickle):
     """
     ## Description
 
-    Environment in which the goal is to go to a red box
-    placed randomly in one big room.
+    Environment in which the goal is to go to a red box placed randomly in one big room.
+    The `OneRoom` environment has two variants. The `OneRoomS6` environment gives you
+    a room with size 6 (the `OneRoom` environment has size 10). The `OneRoomS6Fast`
+    environment also is using a room with size 6, but the turning and moving motion
+    is larger.
 
     ## Action Space
 
@@ -23,18 +26,20 @@ class OneRoom(MiniWorldEnv, utils.EzPickle):
     ## Observation Space
 
     The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
-    representing the view the agents sees.
+    representing a RGB image of what the agents sees.
 
     ## Rewards:
 
-    +(1 - 0.2 * (step_count / max_episode_steps)) when red box reached
+    +(1 - 0.2 * (step_count / max_episode_steps)) when red box reached and zero otherwise.
 
     ## Arguments
 
     ```python
-    OneRoomS6()
+    env = gym.make("MiniWorld-OneRoom-v0")
     # or
-    OneRoomS6Fast()
+    env = gym.make("MiniWorld-OneRoomS6-v0")
+    # or
+    env = gym.make("MiniWorld-OneRoomS6Fast-v0")
     ```
 
     """

miniworld/envs/pickupobjects.py

@@ -24,7 +24,7 @@ class PickupObjects(MiniWorldEnv, utils.EzPickle):
     ## Observation Space
 
     The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
-    representing the view the agents sees.
+    representing a RGB image of what the agents sees.
 
     ## Rewards:
 

miniworld/envs/putnext.py

@@ -9,7 +9,8 @@ class PutNext(MiniWorldEnv, utils.EzPickle):
     ## Description
 
     Single-room environment where a red box must be placed next
-    to a yellow box.
+    to a yellow box. The goal is to perform this task in as few
+    steps as possible.
 
     ## Action Space
 
@@ -27,7 +28,7 @@ class PutNext(MiniWorldEnv, utils.EzPickle):
     ## Observation Space
 
     The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
-    representing the view the agents sees.
+    representing a RGB image of what the agents sees.
 
     ## Rewards:
 

miniworld/envs/roomobjects.py

@@ -10,9 +10,8 @@ class RoomObjects(MiniWorldEnv, utils.EzPickle):
     """
     ## Description
 
-    Single room with multiple objects
-    Inspired by the single room environment of
-    the Generative Query Networks paper:
+    Single room with multiple objects. Inspired by the single room environment
+    of the Generative Query Networks paper:
     https://deepmind.com/blog/neural-scene-representation-and-rendering/
 
     ## Action Space
@@ -31,11 +30,11 @@ class RoomObjects(MiniWorldEnv, utils.EzPickle):
     ## Observation Space
 
     The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
-    representing the view the agents sees.
+    representing a RGB image of what the agents sees.
 
     ## Rewards:
 
-    +0
+    None
 
     ## Arguments
 

miniworld/envs/sidewalk.py

@@ -11,8 +11,8 @@ class Sidewalk(MiniWorldEnv, utils.EzPickle):
     """
     ## Description
 
-    Walk on a sidewalk up to an object to be collected.
-    Don't walk into the street.
+    Walk on a sidewalk up to an object to be collected. Don't walk into the
+    street. The goal is to reach the object in as few steps as possible.
 
     ## Action Space
 
@@ -25,7 +25,7 @@ class Sidewalk(MiniWorldEnv, utils.EzPickle):
     ## Observation Space
 
     The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
-    representing the view the agents sees.
+    representing a RGB image of what the agents sees.
 
     ## Rewards:
 
@@ -34,7 +34,7 @@ class Sidewalk(MiniWorldEnv, utils.EzPickle):
     ## Arguments
 
     ```python
-    Sidewalk()
+    env = gym.make("MiniWorld-Sidewalk-v0")
     ```
 
     """

miniworld/envs/sign.py

@@ -52,7 +52,7 @@ class Sign(MiniWorldEnv, utils.EzPickle):
     ## Observation Space
 
     The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
-    representing the view the agents sees.
+    representing a RGB image of what the agents sees.
 
     ## Rewards:
 
@@ -75,7 +75,7 @@ class Sign(MiniWorldEnv, utils.EzPickle):
 
     """
 
-    def __init__(self, size=10, max_episode_steps=20, color_index=0, goal=0):
+    def __init__(self, size=10, max_episode_steps=20, color_index=0, goal=0, **kwargs):
         if color_index not in [0, 1, 2]:
             raise ValueError("Only supported values for color_index are 0, 1, 2.")
 
@@ -91,9 +91,15 @@ def __init__(self, size=10, max_episode_steps=20, color_index=0, goal=0):
         self._color_index = color_index
 
         MiniWorldEnv.__init__(
-            self, params=params, max_episode_steps=max_episode_steps, domain_rand=False
+            self,
+            params=params,
+            max_episode_steps=max_episode_steps,
+            domain_rand=False,
+            **kwargs,
+        )
+        utils.EzPickle.__init__(
+            self, size, max_episode_steps, color_index, goal, **kwargs
         )
-        utils.EzPickle.__init__(self, size, max_episode_steps, color_index, goal)
 
         self.observation_space = Dict(obs=self.observation_space, goal=Discrete(2))
 

miniworld/envs/threerooms.py

@@ -10,7 +10,9 @@ class ThreeRooms(MiniWorldEnv, utils.EzPickle):
     """
     ## Description
 
-    Two small rooms connected to one large room
+    Two small rooms connected to one large room, with five different items
+    placed on the ground: a red box, a green box, a white ball, a key, and
+    a rubber duck.
 
     ## Action Space
 
@@ -23,7 +25,7 @@ class ThreeRooms(MiniWorldEnv, utils.EzPickle):
     ## Observation Space
 
     The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
-    representing the view the agents sees.
+    representing a RGB image of what the agents sees.
 
     ## Rewards:
 
@@ -32,7 +34,7 @@ class ThreeRooms(MiniWorldEnv, utils.EzPickle):
     ## Arguments
 
     ```python
-    ThreeRooms()
+    env = gym.make("MiniWorld-ThreeRooms-v0")
     ```
 
     """

miniworld/envs/tmaze.py

@@ -10,7 +10,11 @@ class TMaze(MiniWorldEnv, utils.EzPickle):
     """
     ## Description
 
-    Two hallways connected in a T-junction
+    Two hallways connected in a T-junction, the goal is to move the agent
+    towards a red box within as few steps as possible. In
+    `MiniWorld-TMazeLeft-v0`, the red box is located on the left wing of
+    the T-shaped junction. In `MiniWorld-TMazeRight-v0`,  the red box is
+    located on the right wing of the T-shaped junction.
 
     ## Action Space
 
@@ -23,21 +27,19 @@ class TMaze(MiniWorldEnv, utils.EzPickle):
     ## Observation Space
 
     The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
-    representing the view the agents sees.
+    representing a RGB image of what the agents sees.
 
     ## Rewards:
 
-    +(1 - 0.2 * (step_count / max_episode_steps)) when box reached
+    +(1 - 0.2 * (step_count / max_episode_steps)) when box reached and zero otherwise.
 
     ## Arguments
 
     ```python
-    TMazeLeft()
+    env = gym.make("MiniWorld-TMazeLeft-v0")
     # or
-    TMazeRight()
+    env = gym.make("MiniWorld-TMazeRight-v0")
     ```
-
-
     """
 
     def __init__(self, goal_pos=None, **kwargs):

miniworld/envs/wallgap.py

@@ -11,7 +11,8 @@ class WallGap(MiniWorldEnv, utils.EzPickle):
     """
     ## Description
 
-    Outside environment with two rooms connected by a gap in a wall
+    Outside environment with two rooms connected by a gap in a wall. The
+    goal is to go to a red box within as little steps as possible.
 
     ## Action Space
 
@@ -24,7 +25,7 @@ class WallGap(MiniWorldEnv, utils.EzPickle):
     ## Observation Space
 
     The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
-    representing the view the agents sees.
+    representing a RGB image of what the agents sees.
 
     ## Rewards:
 
@@ -33,7 +34,7 @@ class WallGap(MiniWorldEnv, utils.EzPickle):
     ## Arguments
 
     ```python
-    WallGap()
+    env = gym.make("MiniWorld-WallGap-v0")
     ```
     """
 

miniworld/envs/ymaze.py

@@ -12,7 +12,11 @@ class YMaze(MiniWorldEnv, utils.EzPickle):
     """
     ## Description
 
-    Two hallways connected in a Y-junction
+    Two hallways connected in a Y-junction. the goal is to move the agent
+    towards a red box within as little steps as possible. In
+    `MiniWorld-YMazeLeft-v0`, the red box is located on the left wing of
+    the Y-shaped junction. In `MiniWorld-YMazeRight-v0`,  the red box is
+    located on the right wing of the Y-shaped junction.
 
     ## Action Space
 
@@ -25,7 +29,7 @@ class YMaze(MiniWorldEnv, utils.EzPickle):
     ## Observation Space
 
     The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
-    representing the view the agents sees.
+    representing a RGB image of what the agents sees.
 
     ## Rewards:
 
@@ -34,9 +38,9 @@ class YMaze(MiniWorldEnv, utils.EzPickle):
     ## Arguments
 
     ```python
-    YMazeLeft()
+    env = gym.make("MiniWorld-YMazeLeft-v0")
     # or
-    YMazeRight()
+    env = gym.make("MiniWorld-YMazeRight-v0")
     ```
 
     """