Implementation:Haosulab ManiSkill HumanoidPickPlace

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

Overview

Concrete implementation of the humanoid pick-and-place task environment in ManiSkill using the Unitree G1 robot.

Description

The HumanoidPickPlace module defines a base class HumanoidPickPlaceEnv and a concrete task UnitreeG1PlaceAppleInBowlEnv. The task requires the Unitree G1 humanoid robot to grab an apple with its right arm and place it in a bowl on a kitchen counter.

Registered as UnitreeG1PlaceAppleInBowl-v1 with max_episode_steps=100. The supported robot is unitree_g1_simplified_upper_body_with_head_camera.

Randomizations include the bowl's xy position (+/-0.025m), the apple's xy position (+/-0.025m), and the apple's z-axis rotation. Success requires the apple to be within 0.05m euclidean distance of the bowl position, and the robot's right hand to be above the bowl by at least 0.125m. Reward modes include "normalized_dense", "dense", "sparse", and "none".

Usage

Use this environment for humanoid manipulation research involving upper-body coordination, grasping, and precise placement tasks.

Code Reference

Source Location

• Repository: Haosulab_ManiSkill
• File: mani_skill/envs/tasks/humanoid/humanoid_pick_place.py

Signature

class HumanoidPickPlaceEnv(BaseEnv):
    SUPPORTED_REWARD_MODES = ["sparse", "none"]
    ...

@register_env("UnitreeG1PlaceAppleInBowl-v1", max_episode_steps=100)
class UnitreeG1PlaceAppleInBowlEnv(HumanoidPlaceAppleInBowl):
    SUPPORTED_ROBOTS = ["unitree_g1_simplified_upper_body_with_head_camera"]
    agent: UnitreeG1UpperBodyWithHeadCamera

Import

import gymnasium as gym
import mani_skill.envs
env = gym.make("UnitreeG1PlaceAppleInBowl-v1")

I/O Contract

Inputs

Name	Type	Required	Description
obs_mode	str	No	Observation mode
reward_mode	str	No	Reward mode: "normalized_dense", "dense", "sparse", "none"
control_mode	str	No	Control mode for the G1 robot

Outputs

Name	Type	Description
obs	dict/array	Observation including grasping state, TCP pose, bowl/apple positions
reward	float	Dense reward based on reaching, grasping, placing, and releasing
terminated	bool	Whether episode ended by success/failure
truncated	bool	Whether episode hit max steps (100)
info	dict	Contains success, hand_outside_bowl, is_grasped flags

Usage Examples

Basic Usage

import gymnasium as gym
import mani_skill.envs

env = gym.make("UnitreeG1PlaceAppleInBowl-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