Added Docstrings and Updated manual_control.py

README.md

@@ -118,3 +118,4 @@ Alternatively, if this doesn't work, you can also try running MiniWorld with `xv
 ```
 xvfb-run -a -s "-screen 0 1024x768x24 -ac +extension GLX +render -noreset" python3 your_script.py
 ```
+

gym_miniworld/envs/collecthealth.py

@@ -4,15 +4,45 @@
 
 class CollectHealth(MiniWorldEnv):
     """
+    ## Description
+
     Environment where the agent has to collect health kits and stay
     alive as long as possible. This is inspired from the VizDoom
-    HealthGathering environment. Please note, however, that the rewards
+    `HealthGathering` environment. Please note, however, that the rewards
     produced by this environment are not directly comparable to those
     of the VizDoom environment.
 
-    reward:
+    ## Action Space
+
+    | Num | Action                      |
+    |-----|-----------------------------|
+    | 0   | turn left                   |
+    | 1   | turn right                  |
+    | 2   | move forward                |
+    | 3   | move back                   |
+    | 4   | pick up                     |
+    | 5   | drop                        |
+    | 6   | toggle / activate an object |
+    | 7   | complete task               |
+
+    ## Observation Space
+
+    The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
+    representing the view the agents sees.
+
+    ## Rewards:
+
     +2 for each time step
     -100 for dying
+
+    ## Arguments
+
+    ```python
+    CollectHealth(size=16)
+    ```
+
+    `size`: size of the room
+
     """
 
     def __init__(self, size=16, **kwargs):

gym_miniworld/envs/fourrooms.py

@@ -6,8 +6,33 @@
 
 class FourRooms(MiniWorldEnv):
     """
-    Classic four rooms environment.
-    The agent must reach the red box to get a reward.
+    ## Description
+
+    Classic four rooms environment. The agent must reach the red box to get a reward.
+
+    ## 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 the view the agents sees.
+
+    ## Rewards:
+
+    +(1 - 0.2 * (step_count / max_episode_steps)) when red box reached
+
+    ## Arguments
+
+    ```python
+    FourRooms()
+    ```
+
     """
 
     def __init__(self, **kwargs):

gym_miniworld/envs/hallway.py

@@ -8,8 +8,36 @@
 
 class Hallway(MiniWorldEnv):
     """
+    ## Description
+
     Environment in which the goal is to go to a red box
     at the end of a hallway
+
+    ## 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 the view the agents sees.
+
+    ## Rewards:
+
+    +(1 - 0.2 * (step_count / max_episode_steps)) when red box reached
+
+    ## Arguments
+
+    ```python
+    FourRooms(length=12)
+    ```
+
+    `length`: length of the entire space
+
     """
 
     def __init__(self, length=12, **kwargs):

gym_miniworld/envs/maze.py

@@ -7,7 +7,37 @@
 
 class Maze(MiniWorldEnv):
     """
+    ## Description
+
     Maze environment in which the agent has to reach a red box
+
+    ## 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 the view the agents sees.
+
+    ## Rewards:
+
+    +(1 - 0.2 * (step_count / max_episode_steps)) when red box reached
+
+    ## Arguments
+
+    ```python
+    MazeS2()
+    # or
+    MazeS3()
+    # or
+    MazeS3Fast()
+    ```
+
     """
 
     def __init__(

gym_miniworld/envs/oneroom.py

@@ -7,8 +7,36 @@
 
 class OneRoom(MiniWorldEnv):
     """
+    ## Description
+
     Environment in which the goal is to go to a red box
     placed randomly in one big room.
+
+    ## 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 the view the agents sees.
+
+    ## Rewards:
+
+    +(1 - 0.2 * (step_count / max_episode_steps)) when red box reached
+
+    ## Arguments
+
+    ```python
+    OneRoomS6()
+    # or
+    OneRoomS6Fast()
+    ```
+
     """
 
     def __init__(self, size=10, max_episode_steps=180, **kwargs):

gym_miniworld/envs/pickupobjs.py

@@ -6,8 +6,40 @@
 
 class PickupObjs(MiniWorldEnv):
     """
+    ## Description
+
     Room with multiple objects. The agent collects +1 reward for picking up
     each object. Objects disappear when picked up.
+
+    ## Action Space
+
+    | Num | Action                      |
+    |-----|-----------------------------|
+    | 0   | turn left                   |
+    | 1   | turn right                  |
+    | 2   | move forward                |
+    | 3   | move_back                   |
+    | 4   | pickup                      |
+
+    ## Observation Space
+
+    The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
+    representing the view the agents sees.
+
+    ## Rewards:
+
+    +1 when agent picked up object
+
+    ## Arguments
+
+    ```python
+    PickupObjs(size=12, num_objs=5)
+    ```
+
+    `size`: size of world
+
+    `num_objs`: number of objects
+
     """
 
     def __init__(self, size=12, num_objs=5, **kwargs):

gym_miniworld/envs/putnext.py

@@ -4,8 +4,41 @@
 
 class PutNext(MiniWorldEnv):
     """
+    ## Description
+
     Single-room environment where a red box must be placed next
     to a yellow box.
+
+    ## Action Space
+
+    | Num | Action                      |
+    |-----|-----------------------------|
+    | 0   | turn left                   |
+    | 1   | turn right                  |
+    | 2   | move forward                |
+    | 3   | move back                   |
+    | 4   | pick up                     |
+    | 5   | drop                        |
+    | 6   | toggle / activate an object |
+    | 7   | complete task               |
+
+    ## Observation Space
+
+    The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
+    representing the view the agents sees.
+
+    ## Rewards:
+
+    +(1 - 0.2 * (step_count / max_episode_steps)) when red box is next to yellow box
+
+    ## Arguments
+
+    ```python
+    PutNext(size=12)
+    ```
+
+    `size`: size of world
+
     """
 
     def __init__(self, size=12, **kwargs):

gym_miniworld/envs/remotebot.py

@@ -23,9 +23,6 @@
 
 from gym_miniworld.miniworld import MiniWorldEnv
 
-# from gym.utils import seeding
-
-
 try:
     import zmq
 except ImportError:

gym_miniworld/envs/roomobjs.py

@@ -6,10 +6,43 @@
 
 class RoomObjs(MiniWorldEnv):
     """
+    ## Description
+
     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
+
+    | Num | Action                      |
+    |-----|-----------------------------|
+    | 0   | turn left                   |
+    | 1   | turn right                  |
+    | 2   | move forward                |
+    | 3   | move back                   |
+    | 4   | pick up                     |
+    | 5   | drop                        |
+    | 6   | toggle / activate an object |
+    | 7   | complete task               |
+
+    ## Observation Space
+
+    The observation space is an `ndarray` with shape `(obs_height, obs_width, 3)`
+    representing the view the agents sees.
+
+    ## Rewards:
+
+    +0
+
+    ## Arguments
+
+    ```python
+    RoomObjs(size=16)
+    ```
+
+    `size`: size of world
+
     """
 
     def __init__(self, size=10, **kwargs):

gym_miniworld/envs/sidewalk.py

@@ -9,8 +9,34 @@
 
 class Sidewalk(MiniWorldEnv):
     """
+    ## Description
+
     Walk on a sidewalk up to an object to be collected.
     Don't walk into the street.
+
+    ## 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 the view the agents sees.
+
+    ## Rewards:
+
+    +(1 - 0.2 * (step_count / max_episode_steps)) when object reached
+
+    ## Arguments
+
+    ```python
+    Sidewalk()
+    ```
+
     """
 
     def __init__(self, **kwargs):