def __init__(self, goal_direction=None):
self.goal_direction = goal_direction if goal_direction else 1.0
MujocoEnv.__init__(self, 'ant.xml', 5)
gym.utils.EzPickle.__init__(self)
def __init__(self, task=None):
MetaEnv.__init__(self, task)
MujocoEnv.__init__(self, 'humanoid.xml', 5)
gym.utils.EzPickle.__init__(self)
def __init__(self):
self.set_task(self.sample_tasks(1)[0])
MujocoEnv.__init__(self, 'swimmer.xml', 4)
gym.utils.EzPickle.__init__(self)
def __init__(self):
self.goal = np.ones((3, ))
MujocoEnv.__init__(
self, os.path.join(os.path.dirname(__file__), "assets", "pr2.xml"),
2)
gym.utils.EzPickle.__init__(self)
def __init__(self):
MujocoEnv.__init__(self, 'humanoid.xml', 5)
gym.utils.EzPickle.__init__(self)
def __init__(self):
utils.EzPickle.__init__(self)
MujocoEnv.__init__(self, 'inverted_pendulum.xml', 2)
def __init__(self,
#goal=(0, -1),
eef_to_puck_distance_cost_coeff=0,
goal_to_puck_distance_cost_coeff=1.0,
ctrl_cost_coeff=0.1,
puck_initial_x_range=(0, 1),
puck_initial_y_range=(-1, -0.5),
goal_x_range=(-1, 0),
goal_y_range=(-1, 1),
num_goals=-1,
swap_goal_upon_completion=True,
reset_mode="random",
initial_distribution_path="",
goal_mode="random",
goal_sampling_radius_increment=0,
):
"""
goal (`list`): List of two elements denoting the x and y coordinates of
the goal location. Either of the coordinate can also be a string
'any' to make the reward not to depend on the corresponding
coordinate.
arm_distance_coeff ('float'): Coefficient for the arm-to-object distance
cost.
goal_distance_coeff ('float'): Coefficient for the object-to-goal
distance cost.
goal_mode ('string'): Scheme for goal sampling.
- "random", randomly sample a goal within the goal range
- "curriculum-radius", sample a goal from a gradually increasing
region around the puck
"""
self._Serializable__initialize(locals())
self._eef_to_puck_distance_cost_coeff = eef_to_puck_distance_cost_coeff
self._goal_to_puck_distance_cost_coeff = goal_to_puck_distance_cost_coeff
self._ctrl_cost_coeff = ctrl_cost_coeff
self._first_step = True
self._puck_initial_x_range = puck_initial_x_range
self._puck_initial_y_range = puck_initial_y_range
self._goal_x_range = goal_x_range
self._goal_y_range = goal_y_range
self._num_goals = num_goals
self._swap_goal_upon_completion = swap_goal_upon_completion
# For multigoal setting, sample goals
if self._num_goals > 0:
self._goals = list(np.random.uniform(
low=(goal_x_range[0], goal_y_range[0]),
high=(goal_x_range[1], goal_y_range[1]),
size=(num_goals, 2)
))
if self._swap_goal_upon_completion:
self._current_goal_index = 0
else:
self._swap_goal_upon_completion = False
self._reset_mode = reset_mode
if self._reset_mode == "distribution":
self._init_states = self._get_init_pool(initial_distribution_path)
self._goal_mode = goal_mode
if self._goal_mode == "curriculum-radius":
self.goal_sampling_radius = 0
self._goal_sampling_radius_increment = goal_sampling_radius_increment
MujocoEnv.__init__(self, model_path=self.MODEL_PATH, frame_skip=5)
self.model.stat.extent = 10
self._last_qpos = np.concatenate([
self.sim.data.qpos.flat[self.QPOS_JOINT_INDS].copy(),
[np.mean(puck_initial_x_range), np.mean(puck_initial_y_range)],
self.init_qpos.squeeze()[self.QPOS_GOAL_INDS],
])
def __init__(self):
MujocoEnv.__init__(self, 'hopper.xml', 4)
gym.utils.EzPickle.__init__(self)
def __init__(self, file_path: str, frame_skip: int) -> None:
MujocoEnv.__init__(self, file_path, frame_skip)
EzPickle.__init__(self)