Implementation:Haosulab ManiSkill DClaw

Knowledge Sources	Haosulab_ManiSkill
Domains	Robotics, Simulation, Dexterous_Manipulation
Last Updated	2026-02-15 08:00 GMT

Overview

Concrete tool for simulating the D'Claw 3-finger robot gripper in ManiSkill environments.

Description

The DClaw is a 9-DOF tri-finger robot with 3 fingers, each having 3 joints (base, middle, tip). Joint names follow the pattern joint_f{finger}_{segment} (e.g., joint_f1_0, joint_f1_1, joint_f1_2). The tip links (link_f1_head, link_f2_head, link_f3_head) use high-friction materials (static=2.0, dynamic=1.0). Joint stiffness is 1e2, damping 1e1, and force limit 2e1. The robot provides proprioceptive observations including tip poses for all three fingertips. Root joints identify the base rotation joints for each finger.

uid: dclaw

URDF path: {PACKAGE_ASSET_DIR}/robots/dclaw/dclaw_gripper_glb.urdf

Supported control modes: pd_joint_delta_pos, pd_joint_pos, pd_joint_target_delta_pos

Usage

Use DClaw for dexterous manipulation tasks involving object rotation and reorientation using three fingers. The tri-finger design is particularly suited for valve turning and object spinning tasks. Compact and efficient for in-hand manipulation research.

Code Reference

Source Location

• Repository: Haosulab_ManiSkill
• File: mani_skill/agents/robots/dclaw/dclaw.py

Signature

@register_agent()
class DClaw(BaseAgent):
    uid = "dclaw"
    urdf_path = f"{PACKAGE_ASSET_DIR}/robots/dclaw/dclaw_gripper_glb.urdf"
    joint_names = ["joint_f1_0", "joint_f2_0", "joint_f3_0",
                   "joint_f1_1", "joint_f2_1", "joint_f3_1",
                   "joint_f1_2", "joint_f2_2", "joint_f3_2"]
    joint_stiffness = 1e2
    joint_damping = 1e1
    joint_force_limit = 2e1
    tip_link_names = ["link_f1_head", "link_f2_head", "link_f3_head"]
    keyframes = dict(rest=Keyframe(pose=sapien.Pose(p=[0, 0, 0.5], q=[0, 0, -1, 0]), qpos=np.zeros(9)))

Import

from mani_skill.agents.robots.dclaw.dclaw import DClaw

I/O Contract

Inputs

Name	Type	Required	Description
(inherited from BaseAgent)

Outputs

Name	Type	Description
robot agent	DClaw	Configured 9-DOF tri-finger robot with PD joint controllers and tip pose feedback

Usage Examples

Creating Environment with Robot

import gymnasium as gym
import mani_skill.envs

env = gym.make("RotateValveLevel0-v1", robot_uids="dclaw")
obs, info = env.reset()

Related Pages

• Environment:Haosulab_ManiSkill_Python_SAPIEN_Core