Implementation:Farama Foundation Gymnasium HumanoidStandupEnv V5

Knowledge Sources	Farama_Foundation_Gymnasium Gymnasium Docs
Domains	Reinforcement_Learning, MuJoCo_Environments
Last Updated	2026-02-15 03:00 GMT

Overview

Concrete implementation of the HumanoidStandup v5 MuJoCo environment provided by Gymnasium.

Description

The HumanoidStandup environment is based on Tassa, Erez, and Todorov's work in "Synthesis and stabilization of complex behaviors through online trajectory optimization". It uses the same 3D bipedal humanoid robot as the Humanoid environment, but the environment starts with the humanoid laying on the ground. The goal is to make the humanoid stand up and then keep it standing by applying torques to the various hinges. The reward function is: uph_cost + 1 - quad_ctrl_cost - quad_impact_cost, where uph_cost encourages upward movement (absolute z-coordinate divided by timestep), quad_ctrl_cost penalizes large actions, and quad_impact_cost penalizes large contact forces. Unlike other locomotion environments, the humanoid standup never terminates; episodes only end through truncation.

Usage

Use this environment for benchmarking RL algorithms on the challenging task of learning to stand up from a prone position. It tests an agent's ability to discover complex motor skills. The v5 version adds configurable reward weights, excludes worldbody data from observations, and provides tendon information.

Code Reference

Source Location

Repository: Farama_Foundation_Gymnasium
File: gymnasium/envs/mujoco/humanoidstandup_v5.py

Signature

class HumanoidStandupEnv(MujocoEnv, utils.EzPickle):
    def __init__(
        self,
        xml_file: str = "humanoidstandup.xml",
        frame_skip: int = 5,
        default_camera_config: dict[str, float | int] = DEFAULT_CAMERA_CONFIG,
        uph_cost_weight: float = 1,
        ctrl_cost_weight: float = 0.1,
        impact_cost_weight: float = 0.5e-6,
        impact_cost_range: tuple[float, float] = (-np.inf, 10.0),
        reset_noise_scale: float = 1e-2,
        exclude_current_positions_from_observation: bool = True,
        include_cinert_in_observation: bool = True,
        include_cvel_in_observation: bool = True,
        include_qfrc_actuator_in_observation: bool = True,
        include_cfrc_ext_in_observation: bool = True,
        **kwargs,
    )

Import

import gymnasium as gym
env = gym.make("HumanoidStandup-v5")

I/O Contract

Inputs

Name	Type	Required	Description
action	np.ndarray (17,)	Yes	Torques applied to the 17 hinge joints (abdomen, hips, knees, shoulders, elbows), range [-0.4, 0.4]

Outputs

Name	Type	Description
observation	np.ndarray (348,)	State vector: qpos (22), qvel (23), cinert (130), cvel (78), qfrc_actuator (17), cfrc_ext (78); x,y excluded by default
reward	float	uph_cost - quad_ctrl_cost - quad_impact_cost + 1
terminated	bool	Always False (HumanoidStandup never terminates)
truncated	bool	Episode truncation (handled by TimeLimit wrapper, default 1000 timesteps)
info	dict	Contains x_position, y_position, z_distance_from_origin, tendon_length, tendon_velocity, reward_linup, reward_quadctrl, reward_impact

Usage Examples

import gymnasium as gym

env = gym.make("HumanoidStandup-v5")
observation, info = env.reset(seed=42)

for _ in range(1000):
    action = env.action_space.sample()
    observation, reward, terminated, truncated, info = env.step(action)
    if terminated or truncated:
        observation, info = env.reset()

env.close()

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment