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

Use ant=..., for arguments of Scaling

Browse Source
This commit is contained in:
Yuji Kanagawa 2021-09-23 13:32:35 +09:00
parent f966e436c4
commit ef088d848d
1 changed files with 9 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
mujoco_maze/maze_task.py
View File

@ -56,7 +56,7 @@ class Scaling(NamedTuple):
class MazeTask(ABC): class MazeTask(ABC):
REWARD_THRESHOLD: float REWARD_THRESHOLD: float
PENALTY: Optional[float] = None PENALTY: Optional[float] = None
MAZE_SIZE_SCALING: Scaling = Scaling(8.0, 4.0, 4.0) MAZE_SIZE_SCALING: Scaling = Scaling(ant=8.0, point=4.0, swimmer=4.0)
INNER_REWARD_SCALING: float = 0.01 INNER_REWARD_SCALING: float = 0.01
# For Fall/Push/BlockMaze # For Fall/Push/BlockMaze
OBSERVE_BLOCKS: bool = False OBSERVE_BLOCKS: bool = False
@ -146,7 +146,7 @@ class DistRewardSimpleRoom(GoalRewardSimpleRoom, DistRewardMixIn):
class GoalRewardSquareRoom(GoalRewardUMaze): class GoalRewardSquareRoom(GoalRewardUMaze):
MAZE_SIZE_SCALING: Scaling = Scaling(2.0, 4.0, 2.0) MAZE_SIZE_SCALING: Scaling = Scaling(ant=2.0, point=4.0, swimmer=2.0)
def __init__(self, scale: float, goal: Tuple[float, float] = (1.0, 0.0)) -> None: def __init__(self, scale: float, goal: Tuple[float, float] = (1.0, 0.0)) -> None:
super().__init__(scale) super().__init__(scale)
@ -227,7 +227,7 @@ class DistRewardFall(GoalRewardFall, DistRewardMixIn):
class GoalReward2Rooms(MazeTask): class GoalReward2Rooms(MazeTask):
REWARD_THRESHOLD: float = 0.9 REWARD_THRESHOLD: float = 0.9
PENALTY: float = -0.0001 PENALTY: float = -0.0001
MAZE_SIZE_SCALING: Scaling = Scaling(4.0, 4.0, 4.0) MAZE_SIZE_SCALING: Scaling = Scaling(ant=4.0, point=4.0, swimmer=4.0)
def __init__(self, scale: float, goal: Tuple[int, int] = (4.0, -2.0)) -> None: def __init__(self, scale: float, goal: Tuple[int, int] = (4.0, -2.0)) -> None:
super().__init__(scale) super().__init__(scale)
@ -274,7 +274,7 @@ class SubGoal2Rooms(GoalReward2Rooms):
class GoalReward4Rooms(MazeTask): class GoalReward4Rooms(MazeTask):
REWARD_THRESHOLD: float = 0.9 REWARD_THRESHOLD: float = 0.9
PENALTY: float = -0.0001 PENALTY: float = -0.0001
MAZE_SIZE_SCALING: Scaling = Scaling(4.0, 4.0, 4.0) MAZE_SIZE_SCALING: Scaling = Scaling(ant=4.0, point=4.0, swimmer=4.0)
def __init__(self, scale: float) -> None: def __init__(self, scale: float) -> None:
super().__init__(scale) super().__init__(scale)
@ -318,7 +318,7 @@ class SubGoal4Rooms(GoalReward4Rooms):
class GoalRewardTRoom(MazeTask): class GoalRewardTRoom(MazeTask):
REWARD_THRESHOLD: float = 0.9 REWARD_THRESHOLD: float = 0.9
PENALTY: float = -0.0001 PENALTY: float = -0.0001
MAZE_SIZE_SCALING: Scaling = Scaling(4.0, 4.0, 4.0) MAZE_SIZE_SCALING: Scaling = Scaling(ant=4.0, point=4.0, swimmer=4.0)
def __init__(self, scale: float, goal: Tuple[float, float] = (2.0, -3.0)) -> None: def __init__(self, scale: float, goal: Tuple[float, float] = (2.0, -3.0)) -> None:
super().__init__(scale) super().__init__(scale)
@ -362,7 +362,7 @@ class SubGoalTRoom(GoalRewardTRoom):
class NoRewardCorridor(MazeTask): class NoRewardCorridor(MazeTask):
REWARD_THRESHOLD: float = 0.0 REWARD_THRESHOLD: float = 0.0
MAZE_SIZE_SCALING: Scaling = Scaling(4.0, 4.0, 1.0) MAZE_SIZE_SCALING: Scaling = Scaling(ant=4.0, point=4.0, swimmer=1.0)
def reward(self, _obs: np.ndarray) -> float: def reward(self, _obs: np.ndarray) -> float:
return 0.0 return 0.0
@ -403,7 +403,7 @@ class DistRewardCorridor(GoalRewardCorridor, DistRewardMixIn):
class GoalRewardLongCorridor(GoalRewardUMaze): class GoalRewardLongCorridor(GoalRewardUMaze):
MAZE_SIZE_SCALING: Scaling = Scaling(2.0, 4.0, 2.0) MAZE_SIZE_SCALING: Scaling = Scaling(ant=2.0, point=4.0, swimmer=2.0)
def __init__(self, scale: float, goal: Tuple[float, float] = (1.0, 3.0)) -> None: def __init__(self, scale: float, goal: Tuple[float, float] = (1.0, 3.0)) -> None:
super().__init__(scale) super().__init__(scale)
@ -427,7 +427,7 @@ class DistRewardLongCorridor(GoalRewardLongCorridor, DistRewardMixIn):
class GoalRewardBlockMaze(GoalRewardUMaze): class GoalRewardBlockMaze(GoalRewardUMaze):
MAZE_SIZE_SCALING: Scaling = Scaling(8.0, 4.0, None) MAZE_SIZE_SCALING: Scaling = Scaling(ant=8.0, point=4.0, swimmer=None)
OBSERVE_BLOCKS: bool = True OBSERVE_BLOCKS: bool = True
def __init__(self, scale: float) -> None: def __init__(self, scale: float) -> None:
@ -455,7 +455,7 @@ class DistRewardBlockMaze(GoalRewardBlockMaze, DistRewardMixIn):
class GoalRewardBilliard(MazeTask): class GoalRewardBilliard(MazeTask):
REWARD_THRESHOLD: float = 0.9 REWARD_THRESHOLD: float = 0.9
PENALTY: float = -0.0001 PENALTY: float = -0.0001
MAZE_SIZE_SCALING: Scaling = Scaling(None, 3.0, None) MAZE_SIZE_SCALING: Scaling = Scaling(ant=None, point=3.0, swimmer=None)
OBSERVE_BALLS: bool = True OBSERVE_BALLS: bool = True
GOAL_SIZE: float = 0.3 GOAL_SIZE: float = 0.3