robopal.envs.manipulation_tasks.robot_manipulate module

class robopal.envs.manipulation_tasks.robot_manipulate.ManipulateEnv(robot=None, render_mode='human', control_freq=20, controller='CARTIK', is_interpolate=False, action_type='velocity', is_randomize_end=True, is_randomize_object=True, is_show_camera_in_cv=False, is_render_camera_offscreen=False, camera_in_render='frontview', camera_in_window='free')

基类：RobotEnv

The control frequency of the robot is of f = 20 Hz. This is achieved by applying the same action in 50 subsequent simulator step (with a time step of dt = 0.0005 s) before returning the control to the robot.

compute_end_position(input) → Tuple[ndarray, Any]

Map to target action space bounds

compute_rewards(achieved_goal: ndarray = array([0., 0., 0.]), desired_goal: ndarray = array([0., 0., 0.]), info: dict = None, **kwargs)

Sparse Reward: the returned reward can have two values: -1 if the block hasn’t reached its final target position, and 0 if the block is in the final target position (the block is considered to have reached the goal if the Euclidean distance between both is lower than 0.05 m).

static goal_distance(goal_a, goal_b)

reset(seed=None, options=None)

Reset the simulate environment, in order to execute next episode.

reset_object()

Reset the object to a random pose within the workspace.

set_random_init_position()

Set the initial position of the end effector to a random position within the workspace.

step(action) → Tuple

Take one step in the environment.

参数:

action -- The action space is 4-dimensional, with the first 3 dimensions corresponding to the desired

linear velocities of the end effector in Cartesian coordinates, and the last dimension corresponding to the desired gripper state (0 denotes closed, 1 denotes open). :return: obs, reward, terminated, truncated, info

update_init_pose_to_current()