dodox/mujoco_maze
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[Experimental] MultiPush

Browse Source
This commit is contained in:
Yuji Kanagawa 2021-09-24 21:13:02 +09:00
parent 9ddfecc846
commit 2e0c9da3d1
2 changed files with 35 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
mujoco_maze/maze_env.py
View File

@ -451,7 +451,12 @@ class MazeEnv(gym.Env):
def _add_object_ball( def _add_object_ball(
worldbody: ET.Element, i: str, j: str, x: float, y: float, size: float worldbody: ET.Element,
i: str,
j: str,
x: float,
y: float,
size: float,
) -> None: ) -> None:
body = ET.SubElement(worldbody, "body", name=f"objball_{i}_{j}", pos=f"{x} {y} 0") body = ET.SubElement(worldbody, "body", name=f"objball_{i}_{j}", pos=f"{x} {y} 0")
mass = 0.0001 * (size ** 3) mass = 0.0001 * (size ** 3)

31
mujoco_maze/maze_task.py
View File

@ -181,7 +181,7 @@ class GoalRewardPush(GoalRewardUMaze):
def __init__(self, scale: float) -> None: def __init__(self, scale: float) -> None:
super().__init__(scale) super().__init__(scale)
self.goals = [MazeGoal(np.array([0.0, 2.375 * scale]))] self.goals = [MazeGoal(np.array([0.0, 2.375]) * scale)]
@staticmethod @staticmethod
def create_maze() -> List[List[MazeCell]]: def create_maze() -> List[List[MazeCell]]:
@ -199,12 +199,38 @@ class DistRewardPush(GoalRewardPush, DistRewardMixIn):
pass pass
class GoalRewardMultiPush(GoalRewardUMaze):
OBSERVE_BLOCKS: bool = True
MAZE_SIZE_SCALING: Scaling = Scaling(ant=3.0, point=4.0, swimmer=None)
def __init__(self, scale: float, goal: Tuple[float, float] = (1.0, -2)) -> None:
super().__init__(scale)
self.goals = [MazeGoal(np.array(goal) * scale)]
@staticmethod
def create_maze() -> List[List[MazeCell]]:
E, B, R, M = MazeCell.EMPTY, MazeCell.BLOCK, MazeCell.ROBOT, MazeCell.XY_BLOCK
return [
[B, B, B, B, B],
[B, B, E, B, B],
[B, E, M, E, B],
[B, R, E, B, B],
[B, E, M, E, B],
[B, B, E, B, B],
[B, B, B, B, B],
]
class DistRewardMultiPush(GoalRewardMultiPush, DistRewardMixIn):
pass
class GoalRewardFall(GoalRewardUMaze): class GoalRewardFall(GoalRewardUMaze):
OBSERVE_BLOCKS: bool = True OBSERVE_BLOCKS: bool = True
def __init__(self, scale: float) -> None: def __init__(self, scale: float) -> None:
super().__init__(scale) super().__init__(scale)
self.goals = [MazeGoal(np.array([0.0, 3.375 * scale, 4.5]))] self.goals = [MazeGoal(np.array([0.0, 3.375, 4.5]) * scale)]
@staticmethod @staticmethod
def create_maze() -> List[List[MazeCell]]: def create_maze() -> List[List[MazeCell]]:
@ -561,6 +587,7 @@ class TaskRegistry:
"SquareRoom": [DistRewardSquareRoom, GoalRewardSquareRoom, NoRewardSquareRoom], "SquareRoom": [DistRewardSquareRoom, GoalRewardSquareRoom, NoRewardSquareRoom],
"UMaze": [DistRewardUMaze, GoalRewardUMaze], "UMaze": [DistRewardUMaze, GoalRewardUMaze],
"Push": [DistRewardPush, GoalRewardPush], "Push": [DistRewardPush, GoalRewardPush],
"MultiPush": [DistRewardMultiPush, GoalRewardMultiPush],
"Fall": [DistRewardFall, GoalRewardFall], "Fall": [DistRewardFall, GoalRewardFall],
"2Rooms": [DistReward2Rooms, GoalReward2Rooms, SubGoal2Rooms], "2Rooms": [DistReward2Rooms, GoalReward2Rooms, SubGoal2Rooms],
"4Rooms": [DistReward4Rooms, GoalReward4Rooms, SubGoal4Rooms], "4Rooms": [DistReward4Rooms, GoalReward4Rooms, SubGoal4Rooms],