Adroit Relocate

Description

This environment was introduced in “Learning Complex Dexterous Manipulation with Deep Reinforcement Learning and Demonstrations” by Aravind Rajeswaran, Vikash Kumar, Abhishek Gupta, Giulia Vezzani, John Schulman, Emanuel Todorov, and Sergey Levine.

The environment is based on the Adroit manipulation platform, a30 degree of freedom system which consists of a 24 degrees of freedom ShadowHand and a 6 degree of freedom arm. The task to be completed consists on moving the blue ball to the green target. The positions of the ball and target are randomized over the entire workspace. The task will be considered successful when the object is within epsilon-ball of the target.

Action Space

The action space is a Box(-1.0, 1.0, (30,), float32). The control actions are absolute angular positions of the Adroit hand joints. The input of the control actions is set to a range between -1 and 1 by scaling the real actuator angle ranges in radians. The elements of the action array are the following:

Num	Action	Control Min	Control Max	Angle Min	Angle Max	Name (in corresponding XML file)	Joint	Unit
0	Linear translation of the full arm in x direction	-1	1	-0.3 (m)	0.5 (m)	A_ARTx	slide	position (m)
1	Linear translation of the full arm in y direction	-1	1	-0.3 (m)	0.5 (m)	A_ARTy	slide	position (m)
2	Linear translation of the full arm in z direction	-1	1	-0.3 (m)	0.5 (m)	A_ARTz	slide	position (m)
3	Angular up and down movement of the full arm	-1	1	-0.4 (rad)	0.25 (rad)	A_ARRx	hinge	angle (rad)
4	Angular left and right and down movement of the full arm	-1	1	-0.3 (rad)	0.3 (rad)	A_ARRy	hinge	angle (rad)
5	Roll angular movement of the full arm	-1	1	-1.0 (rad)	2.0 (rad)	A_ARRz	hinge	angle (rad)
6	Angular position of the horizontal wrist joint (radial/ulnar deviation)	-1	1	-0.524 (rad)	0.175 (rad)	A_WRJ1	hinge	angle (rad)
7	Angular position of the horizontal wrist joint (flexion/extension)	-1	1	-0.79 (rad)	0.61 (rad)	A_WRJ0	hinge	angle (rad)
8	Horizontal angular position of the MCP joint of the forefinger (adduction/abduction)	-1	1	-0.44 (rad)	0.44(rad)	A_FFJ3	hinge	angle (rad)
9	Vertical angular position of the MCP joint of the forefinger (flexion/extension)	-1	1	0 (rad)	1.6 (rad)	A_FFJ2	hinge	angle (rad)
10	Angular position of the PIP joint of the forefinger (flexion/extension)	-1	1	0 (rad)	1.6 (rad)	A_FFJ1	hinge	angle (rad)
11	Angular position of the DIP joint of the forefinger	-1	1	0 (rad)	1.6 (rad)	A_FFJ0	hinge	angle (rad)
12	Horizontal angular position of the MCP joint of the middle finger (adduction/abduction)	-1	1	-0.44 (rad)	0.44(rad)	A_MFJ3	hinge	angle (rad)
13	Vertical angular position of the MCP joint of the middle finger (flexion/extension)	-1	1	0 (rad)	1.6 (rad)	A_MFJ2	hinge	angle (rad)
14	Angular position of the PIP joint of the middle finger (flexion/extension)	-1	1	0 (rad)	1.6 (rad)	A_MFJ1	hinge	angle (rad)
15	Angular position of the DIP joint of the middle finger	-1	1	0 (rad)	1.6 (rad)	A_MFJ0	hinge	angle (rad)
16	Horizontal angular position of the MCP joint of the ring finger (adduction/abduction)	-1	1	-0.44 (rad)	0.44(rad)	A_RFJ3	hinge	angle (rad)
17	Vertical angular position of the MCP joint of the ring finger (flexion/extension)	-1	1	0 (rad)	1.6 (rad)	A_RFJ2	hinge	angle (rad)
18	Angular position of the PIP joint of the ring finger	-1	1	0 (rad)	1.6 (rad)	A_RFJ1	hinge	angle (rad)
19	Angular position of the DIP joint of the ring finger	-1	1	0 (rad)	1.6 (rad)	A_RFJ0	hinge	angle (rad)
20	Angular position of the CMC joint of the little finger	-1	1	0 (rad)	0.7(rad)	A_LFJ4	hinge	angle (rad)
21	Horizontal angular position of the MCP joint of the little finger (adduction/abduction)	-1	1	-0.44 (rad)	0.44(rad)	A_LFJ3	hinge	angle (rad)
22	Vertical angular position of the MCP joint of the little finger (flexion/extension)	-1	1	0 (rad)	1.6 (rad)	A_LFJ2	hinge	angle (rad)
23	Angular position of the PIP joint of the little finger (flexion/extension)	-1	1	0 (rad)	1.6 (rad)	A_LFJ1	hinge	angle (rad)
24	Angular position of the DIP joint of the little finger	-1	1	0 (rad)	1.6 (rad)	A_LFJ0	hinge	angle (rad)
25	Horizontal angular position of the CMC joint of the thumb finger	-1	1	-1.047 (rad)	1.047 (rad)	A_THJ4	hinge	angle (rad)
26	Vertical Angular position of the CMC joint of the thumb finger	-1	1	0 (rad)	1.3 (rad)	A_THJ3	hinge	angle (rad)
27	Horizontal angular position of the MCP joint of the thumb finger (adduction/abduction)	-1	1	-0.26 (rad)	0.26(rad)	A_THJ2	hinge	angle (rad)
28	Vertical angular position of the MCP joint of the thumb finger (flexion/extension)	-1	1	-0.52 (rad)	0.52 (rad)	A_THJ1	hinge	angle (rad)
29	Angular position of the IP joint of the thumb finger (flexion/extension)	-1	1	-1.571 (rad)	0 (rad)	A_THJ0	hinge	angle (rad)

Observation Space

The observation space is of the type Box(-inf, inf, (39,), float64). It contains information about the angular position of the finger joints, the pose of the palm of the hand, as well as kinematic information about the ball and target.

Num	Observation	Min	Max	Joint Name (in corresponding XML file)	Site/Body Name (in corresponding XML file)	Joint Type	Unit
0	Translation of the arm in the x direction	-Inf	Inf	ARTx	-	slide	position (m)
1	Translation of the arm in the y direction	-Inf	Inf	ARTy	-	slide	position (m)
2	Translation of the arm in the z direction	-Inf	Inf	ARTz	-	slide	position (m)
3	Angular position of the vertical arm joint	-Inf	Inf	ARRx	-	hinge	angle (rad)
4	Angular position of the horizontal arm joint	-Inf	Inf	ARRy	-	hinge	angle (rad)
5	Roll angular value of the arm	-Inf	Inf	ARRz	-	hinge	angle (rad)
6	Angular position of the horizontal wrist joint	-Inf	Inf	WRJ1	-	hinge	angle (rad)
7	Angular position of the vertical wrist joint	-Inf	Inf	WRJ0	-	hinge	angle (rad)
8	Horizontal angular position of the MCP joint of the forefinger	-Inf	Inf	FFJ3	-	hinge	angle (rad)
9	Vertical angular position of the MCP joint of the forefinge	-Inf	Inf	FFJ2	-	hinge	angle (rad)
10	Angular position of the PIP joint of the forefinger	-Inf	Inf	FFJ1	-	hinge	angle (rad)
11	Angular position of the DIP joint of the forefinger	-Inf	Inf	FFJ0	-	hinge	angle (rad)
12	Horizontal angular position of the MCP joint of the middle finger	-Inf	Inf	MFJ3	-	hinge	angle (rad)
13	Vertical angular position of the MCP joint of the middle finger	-Inf	Inf	MFJ2	-	hinge	angle (rad)
14	Angular position of the PIP joint of the middle finger	-Inf	Inf	MFJ1	-	hinge	angle (rad)
15	Angular position of the DIP joint of the middle finger	-Inf	Inf	MFJ0	-	hinge	angle (rad)
16	Horizontal angular position of the MCP joint of the ring finger	-Inf	Inf	RFJ3	-	hinge	angle (rad)
17	Vertical angular position of the MCP joint of the ring finger	-Inf	Inf	RFJ2	-	hinge	angle (rad)
18	Angular position of the PIP joint of the ring finger	-Inf	Inf	RFJ1	-	hinge	angle (rad)
19	Angular position of the DIP joint of the ring finger	-Inf	Inf	RFJ0	-	hinge	angle (rad)
20	Angular position of the CMC joint of the little finger	-Inf	Inf	LFJ4	-	hinge	angle (rad)
21	Horizontal angular position of the MCP joint of the little finger	-Inf	Inf	LFJ3	-	hinge	angle (rad)
22	Vertical angular position of the MCP joint of the little finger	-Inf	Inf	LFJ2	-	hinge	angle (rad)
23	Angular position of the PIP joint of the little finger	-Inf	Inf	LFJ1	-	hinge	angle (rad)
24	Angular position of the DIP joint of the little finger	-Inf	Inf	LFJ0	-	hinge	angle (rad)
25	Horizontal angular position of the CMC joint of the thumb finger	-Inf	Inf	THJ4	-	hinge	angle (rad)
26	Vertical Angular position of the CMC joint of the thumb finger	-Inf	Inf	THJ3	-	hinge	angle (rad)
27	Horizontal angular position of the MCP joint of the thumb finger	-Inf	Inf	THJ2	-	hinge	angle (rad)
28	Vertical angular position of the MCP joint of the thumb finger	-Inf	Inf	THJ1	-	hinge	angle (rad)
29	Angular position of the IP joint of the thumb finger	-Inf	Inf	THJ0	-	hinge	angle (rad)
30	x positional difference from the palm of the hand to the ball	-Inf	Inf	-	Object,S_grasp	-	position (m)
31	y positional difference from the palm of the hand to the ball	-Inf	Inf	-	Object,S_grasp	-	position (m)
32	z positional difference from the palm of the hand to the ball	-Inf	Inf	-	Object,S_grasp	-	position (m)
33	x positional difference from the palm of the hand to the target	-Inf	Inf	-	Object,target	-	position (m)
34	y positional difference from the palm of the hand to the target	-Inf	Inf	-	Object,target	-	position (m)
35	z positional difference from the palm of the hand to the target	-Inf	Inf	-	Object,target	-	position (m)
36	x positional difference from the ball to the target	-Inf	Inf	-	Object,target	-	position (m)
37	y positional difference from the ball to the target	-Inf	Inf	-	Object,target	-	position (m)
38	z positional difference from the ball to the target	-Inf	Inf	-	Object,target	-	position (m)

Rewards

The environment can be initialized in either a dense or sparse reward variant.

In the dense reward setting, the environment returns a dense reward function that consists of the following parts:

• get_to_ball: increasing negative reward the further away the palm of the hand is from the ball. This is computed as the 3 dimensional Euclidean distance between both body frames. This penalty is scaled by a factor of 0.1 in the final reward.

• ball_off_table: add a positive reward of 1 if the ball is lifted from the table (z greater than 0.04 meters). If this condition is met two additional rewards are added:

  • make_hand_go_to_target: negative reward equal to the 3 dimensional Euclidean distance from the palm to the target ball position. This reward is scaled by a factor of 0.5. - make_ball_go_to_target: negative reward equal to the 3 dimensional Euclidean distance from the ball to its target position. This reward is also scaled by a factor of 0.5.

• ball_close_to_target: bonus of 10 if the ball’s Euclidean distance to its target is less than 0.1 meters. Bonus of 20 if the distance is less than 0.05 meters.

The sparse reward variant of the environment can be initialized by calling gym.make('AdroitHandReloateSparse-v1'). In this variant, the environment returns a reward of 10 for environment success and -0.1 otherwise.

Starting State

The ball is set randomly over the table at reset. The ranges of the uniform distribution from which the position is samples are [-0.15,0.15] for the x coordinate, and [-0.15,0.3] got the y coordinate. The target position is also sampled from uniform distributions with ranges [-0.2,0.2] for the x coordinate, [-0.2,0.2] for the y coordinate, and [0.15,0.35] for the z coordinate.

The joint values of the environment are deterministically initialized to a zero.

For reproducibility, the starting state of the environment can also be set when calling env.reset() by passing the options dictionary argument (https://gymnasium.farama.org/api/env/#gymnasium.Env.reset) with the initial_state_dict key. The initial_state_dict key must be a dictionary with the following items:

• qpos: np.ndarray with shape (36,), MuJoCo simulation joint positions

• qvel: np.ndarray with shape (36,), MuJoCo simulation joint velocities

• obj_pos: np.ndarray with shape (3,), cartesian coordinates of the ball object

• target_pos: np.ndarray with shape (3,), cartesian coordinates of the goal ball location

The state of the simulation can also be set at any step with the env.set_env_state(initial_state_dict) method.

Episode End

The episode will be truncated when the duration reaches a total of max_episode_steps which by default is set to 200 timesteps. The episode is never terminated since the task is continuing with infinite horizon.

Arguments

To increase/decrease the maximum number of timesteps before the episode is truncated the max_episode_steps argument can be set at initialization. The default value is 50. For example, to increase the total number of timesteps to 400 make the environment as follows:

import gymnasium as gym
import gymnasium_robotics

gym.register_envs(gymnasium_robotics)

env = gym.make('AdroitHandRelocate-v1', max_episode_steps=400)

Version History

• v1: refactor version of the D4RL environment, also create dependency on newest mujoco python bindings maintained by the MuJoCo team in Deepmind.

• v0: legacy versions in the D4RL.