Implementation:Haosulab ManiSkill AssemblingKits

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

Overview

Concrete implementation of the Assembling Kits shape insertion task environment in ManiSkill.

Description

The AssemblingKitsEnv requires a robot to pick up a misplaced shape and insert it into the correct empty slot on a board/kit. The kit geometry, shape types, and positions are randomized across 20 different shapes with rotational symmetry considerations.

Registered as AssemblingKits-v1 with max_episode_steps=200 and asset_download_ids=["assembling_kits"]. The supported robot is panda_wristcam. Reward modes include "sparse" and "none".

Success requires: correct position (within 2cm), correct rotation (within 4 degrees, accounting for symmetry), and the shape being fully inserted into the slot (height < 3mm). Kit configurations are loaded from JSON episode files.

Usage

Use this environment for research on shape-based insertion and precision placement tasks. The diverse shape geometries and rotational symmetry requirements make it a challenging manipulation benchmark.

Code Reference

Source Location

• Repository: Haosulab_ManiSkill
• File: mani_skill/envs/tasks/tabletop/assembling_kits.py

Signature

@register_env(
    "AssemblingKits-v1", asset_download_ids=["assembling_kits"], max_episode_steps=200
)
class AssemblingKitsEnv(BaseEnv):
    SUPPORTED_REWARD_MODES = ["sparse", "none"]
    SUPPORTED_ROBOTS = ["panda_wristcam"]
    agent: Union[PandaWristCam]

Import

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

I/O Contract

Inputs

Name	Type	Required	Description
obs_mode	str	No	Observation mode
reward_mode	str	No	Reward mode: "sparse", "none"
control_mode	str	No	Control mode for the Panda robot

Outputs

Name	Type	Description
obs	dict/array	Observation including TCP pose, object pose, goal position/rotation
reward	float	Sparse reward when shape is correctly inserted
terminated	bool	Whether episode ended by success
truncated	bool	Whether episode hit max steps (200)
info	dict	Contains success, pos_correct, rot_correct, in_slot flags

Usage Examples

Basic Usage

import gymnasium as gym
import mani_skill.envs

env = gym.make("AssemblingKits-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