The source code and dockerfile for the GSW2024 AI Lab.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
This repo is archived. You can view files and clone it, but cannot push or open issues/pull-requests.
 
 
 
 
 
 

161 lines
5.0 KiB

from __future__ import annotations
from minigrid.core.constants import COLOR_NAMES
from minigrid.core.grid import Grid
from minigrid.core.mission import MissionSpace
from minigrid.core.world_object import Ball, Box, Key
from minigrid.minigrid_env import MiniGridEnv
class GoToObjectEnv(MiniGridEnv):
"""
## Description
This environment is a room with colored objects. The agent
receives a textual (mission) string as input, telling it which colored object to go
to, (eg: "go to the red key"). It receives a positive reward for performing
the `done` action next to the correct object, as indicated in the mission
string.
## Mission Space
"go to the {color} {obj_type}"
{color} is the color of the object. Can be "red", "green", "blue", "purple",
"yellow" or "grey".
{obj_type} is the type of the object. Can be "key", "ball", "box".
## 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 | Unused |
| 6 | done | Done completing task |
## 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/minigrid.py](minigrid/minigrid.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 stands next the correct door performing the `done` action.
2. Timeout (see `max_steps`).
## Registered Configurations
- `MiniGrid-GoToObject-6x6-N2-v0`
- `MiniGrid-GoToObject-8x8-N2-v0`
"""
def __init__(self, size=6, numObjs=2, max_steps: int | None = None, **kwargs):
self.numObjs = numObjs
self.size = size
# Types of objects to be generated
self.obj_types = ["key", "ball", "box"]
mission_space = MissionSpace(
mission_func=self._gen_mission,
ordered_placeholders=[COLOR_NAMES, self.obj_types],
)
if max_steps is None:
max_steps = 5 * size**2
super().__init__(
mission_space=mission_space,
width=size,
height=size,
# Set this to True for maximum speed
see_through_walls=True,
max_steps=max_steps,
**kwargs,
)
@staticmethod
def _gen_mission(color: str, obj_type: str):
return f"go to the {color} {obj_type}"
def _gen_grid(self, width, height):
self.grid = Grid(width, height)
# Generate the surrounding walls
self.grid.wall_rect(0, 0, width, height)
# Types and colors of objects we can generate
types = ["key", "ball", "box"]
objs = []
objPos = []
# Until we have generated all the objects
while len(objs) < self.numObjs:
objType = self._rand_elem(types)
objColor = self._rand_elem(COLOR_NAMES)
# If this object already exists, try again
if (objType, objColor) in objs:
continue
if objType == "key":
obj = Key(objColor)
elif objType == "ball":
obj = Ball(objColor)
elif objType == "box":
obj = Box(objColor)
else:
raise ValueError(
"{} object type given. Object type can only be of values key, ball and box.".format(
objType
)
)
pos = self.place_obj(obj)
objs.append((objType, objColor))
objPos.append(pos)
# Randomize the agent start position and orientation
self.place_agent()
# Choose a random object to be picked up
objIdx = self._rand_int(0, len(objs))
self.targetType, self.target_color = objs[objIdx]
self.target_pos = objPos[objIdx]
descStr = f"{self.target_color} {self.targetType}"
self.mission = "go to the %s" % descStr
# print(self.mission)
def step(self, action):
obs, reward, terminated, truncated, info = super().step(action)
ax, ay = self.agent_pos
tx, ty = self.target_pos
# Toggle/pickup action terminates the episode
if action == self.actions.toggle:
terminated = True
# Reward performing the done action next to the target object
if action == self.actions.done:
if (ax == tx and abs(ay - ty) == 1) or (ay == ty and abs(ax - tx) == 1):
reward = self._reward()
terminated = True
return obs, reward, terminated, truncated, info