import numpy as np
from robopal.envs.manipulation_tasks.robot_manipulate import ManipulateEnv
import robopal.commons.transform as trans
from robopal.robots.diana_med import DianaDrawerCube
from robopal.wrappers import GoalEnvWrapper
[文档]
class DrawerCubeEnv(ManipulateEnv):
name = 'DrawerBox-v1'
def __init__(self,
robot=DianaDrawerCube,
render_mode='human',
control_freq=20,
is_show_camera_in_cv=False,
controller='CARTIK',
):
super().__init__(
robot=robot,
render_mode=render_mode,
control_freq=control_freq,
is_show_camera_in_cv=is_show_camera_in_cv,
controller=controller,
)
self.obs_dim = (35,)
self.goal_dim = (9,)
self.action_dim = (4,)
self.max_action = 1.0
self.min_action = -1.0
self.max_episode_steps = 50
self.TASK_FLAG = 0
self.pos_max_bound = np.array([0.6, 0.2, 0.37])
self.pos_min_bound = np.array([0.3, -0.2, 0.12])
self.grip_max_bound = 0.02
self.grip_min_bound = -0.01
def _get_obs(self) -> dict:
""" The observation space is 16-dimensional, with the first 3 dimensions corresponding to the position
of the block, the next 3 dimensions corresponding to the position of the goal, the next 3 dimensions
corresponding to the position of the gripper, the next 3 dimensions corresponding to the vector
between the block and the gripper, and the last dimension corresponding to the current gripper opening.
"""
obs = np.zeros(self.obs_dim)
# gripper state
obs[0:8] = np.concatenate((
# gripper position in global coordinates
end_pos := self.get_site_pos('0_grip_site'),
# gripper linear velocity
end_vel := self.get_site_xvelp('0_grip_site') * self.dt,
self.mj_data.joint('0_r_finger_joint').qpos,
self.mj_data.joint('0_r_finger_joint').qvel * self.dt
))
if self.TASK_FLAG == 0:
obs[11:14] = ( # handle position in global coordinates
handle_pos := self.get_site_pos('drawer')
)
obs[14:17] = end_pos - handle_pos # distance between the handle and the end
# velocity with respect to the gripper
handle_velp = self.get_site_xvelp('drawer') * self.dt
obs[17:20] = ( # velocity with respect to the gripper
handle_velp - end_vel
)
# drawer state
if self.TASK_FLAG == 1:
obs[20:23] = ( # block position in global coordinates
block_pos := self.get_body_pos('green_block')
)
obs[23:26] = end_pos - block_pos # distance between the block and the end
obs[26:29] = ( # block rotation
trans.mat_2_euler(self.get_body_rotm('green_block'))
)
block_velp = self.get_body_xvelp('green_block') * self.dt
obs[29:32] = ( # velocity with respect to the gripper
block_velp # - end_vel
)
return obs.copy()
def _get_achieved_goal(self):
achieved_goal = np.concatenate([
self.get_site_pos('0_grip_site'),
self.get_site_pos('drawer'),
self.get_body_pos('green_block')
], axis=0)
return achieved_goal.copy()
def _get_desired_goal(self):
if self._is_success(self.get_site_pos('drawer'), self.get_site_pos('drawer_goal')) == 0:
reach_goal = self.get_site_pos('drawer')
else:
reach_goal = self.get_body_pos('green_block')
desired_goal = np.concatenate([
reach_goal,
self.get_site_pos('drawer_goal'),
self.get_site_pos('cube_goal'),
], axis=0)
return desired_goal.copy()
def _get_info(self) -> dict:
return {
'is_drawer_success': self._is_success(self.get_site_pos('drawer'), self.get_site_pos('drawer_goal')),
'is_place_success': self._is_success(self.get_body_pos('green_block'), self.get_site_pos('cube_goal'))
}
[文档]
def reset_object(self):
if self.TASK_FLAG == 1:
self.mj_data.joint('drawer:joint').qpos[0] = 0.14
# reset object position
random_x_pos = np.random.uniform(0.35, 0.4)
random_y_pos = np.random.uniform(-0.15, 0.15)
self.set_object_pose('green_block:joint', np.array([random_x_pos, random_y_pos, 0.44, 1.0, 0.0, 0.0, 0.0]))
self.set_site_pos('cube_goal', np.array([0.59, 0.0, 0.478]))
return super().reset_object()
if __name__ == "__main__":
env = DrawerCubeEnv()
env = GoalEnvWrapper(env)
env.reset()
for t in range(int(1e5)):
action = np.random.uniform(env.min_action, env.max_action, env.action_dim)
s_, r, terminated, truncated, info = env.step(action)
if truncated:
env.reset()
env.close()
