Implementation:Haosulab ManiSkill InsertFlower

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

Overview

Concrete implementation of the flower insertion dexterous manipulation task in ManiSkill using a floating ability hand.

Description

The InsertFlowerEnv requires the FloatingAbilityHandRight robot to grasp a flower and insert it into a target area. The task uses assets from the OakInk-v2 dataset.

Registered as InsertFlower-v1 with max_episode_steps=300 and asset_download_ids=["oakink-v2"]. The supported robot is floating_ability_hand_right.

The flower spawn position is randomized within a half-size of 0.05m on the table surface. The target insertion area is defined by min=[-0.3, -0.25, 0.25] and max=[-0.2, -0.15, 0.35]. The hand is initialized at a height of 0.25m above the table.

Usage

Use this environment for dexterous manipulation research involving grasping and precise insertion with a floating hand. It tests fine-grained control with the ability hand robot.

Code Reference

Source Location

• Repository: Haosulab_ManiSkill
• File: mani_skill/envs/tasks/dexterity/insert_flower.py

Signature

@register_env(
    "InsertFlower-v1", max_episode_steps=300, asset_download_ids=["oakink-v2"]
)
class InsertFlowerEnv(BaseEnv):
    agent: Union[FloatingAbilityHandRight]
    hand_init_height = 0.25
    flower_spawn_half_size = 0.05

Import

import gymnasium as gym
import mani_skill.envs
env = gym.make("InsertFlower-v1")

I/O Contract

Inputs

Name	Type	Required	Description
obs_mode	str	No	Observation mode
reward_mode	str	No	Reward mode
control_mode	str	No	Control mode for the ability hand

Outputs

Name	Type	Description
obs	dict/array	Observation including hand state, flower pose, target area
reward	float	Task reward
terminated	bool	Whether episode ended by success/failure
truncated	bool	Whether episode hit max steps (300)
info	dict	Contains success flag

Usage Examples

Basic Usage

import gymnasium as gym
import mani_skill.envs

env = gym.make("InsertFlower-v1", obs_mode="state", render_mode="rgb_array")
obs, info = env.reset()
for _ in range(100):
    action = env.action_space.sample()
    obs, reward, terminated, truncated, info = env.step(action)
    if terminated or truncated:
        obs, info = env.reset()
env.close()

Related Pages

• Environment:Haosulab_ManiSkill_Python_SAPIEN_Core