Implementation:Haosulab ManiSkill TriFingerPro

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

Overview

Concrete tool for simulating the TriFinger Pro robot in ManiSkill environments.

Description

The TriFingerPro is a 9-DOF tri-finger robot platform modified from the NVIDIA IsaacGymEnvs implementation. It has 3 fingers, each with 3 joints (base_to_upper, upper_to_middle, middle_to_lower), oriented at 0, 120, and 240 degrees. The tip links (finger_tip_link_0, finger_tip_link_120, finger_tip_link_240) use high-friction materials (static=2.0, dynamic=1.0). Joint stiffness is 1e2, damping 1e1, force limit 2e1. The robot provides proprioceptive observations including tip poses (7 values per tip) and tip linear velocities (3 per tip). Also provides root joint tracking for finger base rotations.

uid: trifingerpro

URDF path: {PACKAGE_ASSET_DIR}/robots/trifinger/trifingerpro.urdf

Supported control modes: pd_joint_delta_pos, pd_joint_pos, pd_joint_target_delta_pos, pd_joint_vel, pd_joint_pos_vel, pd_joint_delta_pos_vel

Usage

Use TriFingerPro for dexterous manipulation benchmarks involving object reorientation, precise positioning, and in-hand manipulation. The symmetrical tri-finger design enables stable three-point contact manipulation. Commonly used in the Real Robot Challenge and related benchmarks.

Code Reference

Source Location

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

Signature

@register_agent()
class TriFingerPro(BaseAgent):
    uid = "trifingerpro"
    urdf_path = f"{PACKAGE_ASSET_DIR}/robots/trifinger/trifingerpro.urdf"
    joint_names = [
        "finger_base_to_upper_joint_0", "finger_upper_to_middle_joint_0", "finger_middle_to_lower_joint_0",
        "finger_base_to_upper_joint_120", ..., "finger_base_to_upper_joint_240", ...
    ]  # 9 joints
    joint_stiffness = 1e2
    joint_damping = 1e1
    joint_force_limit = 2e1
    tip_link_names = ["finger_tip_link_0", "finger_tip_link_120", "finger_tip_link_240"]

Import

from mani_skill.agents.robots.trifingerpro.trifingerpro import TriFingerPro

I/O Contract

Inputs

Name	Type	Required	Description
(inherited from BaseAgent)

Outputs

Name	Type	Description
robot agent	TriFingerPro	Configured 9-DOF tri-finger robot with tip pose and velocity feedback

Usage Examples

Creating Environment with Robot

import gymnasium as gym
import mani_skill.envs

env = gym.make("TriFingerRotateCube-v1", robot_uids="trifingerpro")
obs, info = env.reset()

Related Pages

• Environment:Haosulab_ManiSkill_Python_SAPIEN_Core