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.
 
 
 
 
 
 

200 lines
6.2 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 PutNearEnv(MiniGridEnv):
"""
## Description
The agent is instructed through a textual string to pick up an object and
place it next to another object. This environment is easy to solve with two
objects, but difficult to solve with more, as it involves both textual
understanding and spatial reasoning involving multiple objects.
## Mission Space
"put the {move_color} {move_type} near the {target_color} {target_type}"
{move_color} and {target_color} can be "red", "green", "blue", "purple",
"yellow" or "grey".
{move_type} and {target_type} Can be "box", "ball" or "key".
## Action Space
| Num | Name | Action |
|-----|--------------|-------------------|
| 0 | left | Turn left |
| 1 | right | Turn right |
| 2 | forward | Move forward |
| 3 | pickup | Pick up an object |
| 4 | drop | Drop an object |
| 5 | toggle | Unused |
| 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/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 picks up the wrong object.
2. The agent drop the correct object near the target.
3. Timeout (see `max_steps`).
## Registered Configurations
N: number of objects.
- `MiniGrid-PutNear-6x6-N2-v0`
- `MiniGrid-PutNear-8x8-N3-v0`
"""
def __init__(self, size=6, numObjs=2, max_steps: int | None = None, **kwargs):
self.size = size
self.numObjs = numObjs
self.obj_types = ["key", "ball", "box"]
mission_space = MissionSpace(
mission_func=self._gen_mission,
ordered_placeholders=[
COLOR_NAMES,
self.obj_types,
COLOR_NAMES,
self.obj_types,
],
)
if max_steps is None:
max_steps = 5 * size
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(
move_color: str, move_type: str, target_color: str, target_type: str
):
return f"put the {move_color} {move_type} near the {target_color} {target_type}"
def _gen_grid(self, width, height):
self.grid = Grid(width, height)
# Generate the surrounding walls
self.grid.horz_wall(0, 0)
self.grid.horz_wall(0, height - 1)
self.grid.vert_wall(0, 0)
self.grid.vert_wall(width - 1, 0)
# Types and colors of objects we can generate
types = ["key", "ball", "box"]
objs = []
objPos = []
def near_obj(env, p1):
for p2 in objPos:
dx = p1[0] - p2[0]
dy = p1[1] - p2[1]
if abs(dx) <= 1 and abs(dy) <= 1:
return True
return False
# 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, reject_fn=near_obj)
objs.append((objType, objColor))
objPos.append(pos)
# Randomize the agent start position and orientation
self.place_agent()
# Choose a random object to be moved
objIdx = self._rand_int(0, len(objs))
self.move_type, self.moveColor = objs[objIdx]
self.move_pos = objPos[objIdx]
# Choose a target object (to put the first object next to)
while True:
targetIdx = self._rand_int(0, len(objs))
if targetIdx != objIdx:
break
self.target_type, self.target_color = objs[targetIdx]
self.target_pos = objPos[targetIdx]
self.mission = "put the {} {} near the {} {}".format(
self.moveColor,
self.move_type,
self.target_color,
self.target_type,
)
def step(self, action):
preCarrying = self.carrying
obs, reward, terminated, truncated, info = super().step(action)
u, v = self.dir_vec
ox, oy = (self.agent_pos[0] + u, self.agent_pos[1] + v)
tx, ty = self.target_pos
# If we picked up the wrong object, terminate the episode
if action == self.actions.pickup and self.carrying:
if (
self.carrying.type != self.move_type
or self.carrying.color != self.moveColor
):
terminated = True
# If successfully dropping an object near the target
if action == self.actions.drop and preCarrying:
if self.grid.get(ox, oy) is preCarrying:
if abs(ox - tx) <= 1 and abs(oy - ty) <= 1:
reward = self._reward()
terminated = True
return obs, reward, terminated, truncated, info