Implementation:Haosulab ManiSkill HumanoidStand

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

Overview

Concrete implementation of the humanoid standing balance task environment in ManiSkill.

Description

The HumanoidStand module defines standing balance tasks for Unitree humanoid robots. The base class HumanoidStandEnv provides common logic, and two robot-specific variants are registered:

• UnitreeH1Stand-v1 (max_episode_steps=1000): Unitree H1 humanoid must maintain standing posture.
• UnitreeG1Stand-v1 (max_episode_steps=1000): Unitree G1 humanoid must maintain standing posture.

The scene contains only a ground plane. The robot is initialized from its "standing" keyframe with small random perturbations (noise of 0.05). Evaluation checks agent.is_standing() and agent.is_fallen(). Failure occurs when the robot falls. Reward modes include "sparse" and "none".

Usage

Use this environment as a baseline balance task for humanoid robots. It tests the ability to maintain upright posture from slightly perturbed initial conditions.

Code Reference

Source Location

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

Signature

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

@register_env("UnitreeH1Stand-v1", max_episode_steps=1000)
class UnitreeH1StandEnv(HumanoidStandEnv):
    SUPPORTED_ROBOTS = ["unitree_h1_simplified"]
    ...

@register_env("UnitreeG1Stand-v1", max_episode_steps=1000)
class UnitreeG1StandEnv(HumanoidStandEnv):
    SUPPORTED_ROBOTS = ["unitree_g1_simplified_legs"]
    ...

Import

import gymnasium as gym
import mani_skill.envs
env = gym.make("UnitreeH1Stand-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 humanoid robot

Outputs

Name	Type	Description
obs	dict/array	Observation based on obs_mode
reward	float	Sparse reward: 1 if standing, 0 otherwise
terminated	bool	Whether the robot has fallen
truncated	bool	Whether episode hit max steps (1000)
info	dict	Contains is_standing, fail flags

Usage Examples

Basic Usage

import gymnasium as gym
import mani_skill.envs

env = gym.make("UnitreeH1Stand-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