Implementation:Farama Foundation Gymnasium MuJoCo Utils

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

Overview

A set of MujocoEnv-related utility functions provided by Gymnasium, primarily used for testing and state management purposes.

Description

The MuJoCo utilities module provides three functions for working with MuJoCo environment state: get_state retrieves the full physics state of an environment, set_state restores a previously captured state, and check_mujoco_reset_state verifies that environment reset is deterministic and unaffected by previous steps. These utilities require mujoco >= 2.3.6 and leverage the mjtState API for precise state serialization. The state type can be configured to capture either the full physics state (mjSTATE_FULLPHYSICS) or just the integration state (mjSTATE_INTEGRATION), which is useful for distinguishing between training and validation use cases.

Usage

Use these utilities for testing MuJoCo environments to ensure deterministic behavior, for implementing state save/restore functionality in planning algorithms, or for debugging environment state issues. The check_mujoco_reset_state function is particularly useful for validating that custom MuJoCo environments properly reset their state.

Code Reference

Source Location

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

Signature

def get_state(
    env: gymnasium.envs.mujoco.MujocoEnv,
    state_type: mujoco.mjtState = mujoco.mjtState.mjSTATE_FULLPHYSICS,
) -> np.ndarray

def set_state(
    env: gymnasium.envs.mujoco.MujocoEnv,
    state: np.ndarray,
    state_type: mujoco.mjtState = mujoco.mjtState.mjSTATE_FULLPHYSICS,
) -> np.ndarray

def check_mujoco_reset_state(
    env: gymnasium.envs.mujoco.MujocoEnv,
    seed=1234,
    state_type: mujoco.mjtState = mujoco.mjtState.mjSTATE_INTEGRATION,
) -> None

Import

from gymnasium.envs.mujoco.utils import get_state, set_state, check_mujoco_reset_state

I/O Contract

get_state

Name	Type	Required	Description
env	gymnasium.envs.mujoco.MujocoEnv	Yes	The MuJoCo environment whose state to retrieve
state_type	mujoco.mjtState	No	Type of state to capture (default: mjSTATE_FULLPHYSICS)

Returns: np.ndarray containing the serialized state.

set_state

Name	Type	Required	Description
env	gymnasium.envs.mujoco.MujocoEnv	Yes	The MuJoCo environment whose state to set
state	np.ndarray	Yes	State array previously generated from get_state
state_type	mujoco.mjtState	No	Type of state to restore (default: mjSTATE_FULLPHYSICS)

Returns: np.ndarray of the state that was set.

check_mujoco_reset_state

Name	Type	Required	Description
env	gymnasium.envs.mujoco.MujocoEnv	Yes	The MuJoCo environment to test
seed	int	No	Seed used for env.reset() (default: 1234)
state_type	mujoco.mjtState	No	Type of state to compare (default: mjSTATE_INTEGRATION)

Returns: None. Raises AssertionError if reset is not deterministic.

Usage Examples

import gymnasium as gym
from gymnasium.envs.mujoco.utils import get_state, set_state, check_mujoco_reset_state

# Save and restore state
env = gym.make("Ant-v5")
observation, info = env.reset(seed=42)

state = get_state(env)
observation, reward, terminated, truncated, info = env.step(env.action_space.sample())

# Restore the previous state
set_state(env, state)

# Verify deterministic reset
check_mujoco_reset_state(env)

env.close()

Related Pages

Page Connections

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

Principle

Implementation

Heuristic

Environment