Implementation:Farama Foundation Gymnasium FlattenObservation Wrapper

Knowledge Sources	Gymnasium Gymnasium Obs Wrappers
Domains	Reinforcement_Learning, Preprocessing
Last Updated	2026-02-15 03:00 GMT

Overview

Concrete tool for flattening structured observations into flat vectors provided by the Gymnasium library.

Description

The FlattenObservation wrapper flattens the environment's observation space and each observation from reset and step. It uses spaces.utils.flatten_space to compute the flattened space and spaces.utils.flatten to transform each observation. This converts Dict, Tuple, or multi-dimensional Box observations into a single flat Box vector suitable for standard neural network architectures.

Usage

Wrap environments with Dict or Tuple observation spaces when using agents that expect flat vector inputs (e.g., MLP-based policies). Not needed for environments that already produce flat observations.

Code Reference

Source Location

• Repository: Gymnasium
• File: gymnasium/wrappers/transform_observation.py
• Lines: L219-256

Signature

class FlattenObservation(TransformObservation):
    def __init__(self, env: gym.Env[ObsType, ActType]):
        """Flattens the environment's observation space and each observation.

        Args:
            env: The environment to wrap. Must have an observation space
                 that implements spaces.utils.flatten_space and spaces.utils.flatten.
        """

Import

from gymnasium.wrappers import FlattenObservation

I/O Contract

Inputs

Name	Type	Required	Description
env	gym.Env	Yes	Environment with any flattenable observation space

Outputs

Name	Type	Description
observation_space	Box	Flattened 1D Box space
observations	np.ndarray	Flat 1D arrays from reset() and step()

Usage Examples

Flatten Dict Observations

import gymnasium as gym
from gymnasium.wrappers import FlattenObservation

# Environment with Dict observation space
env = gym.make("CarRacing-v3")
print(env.observation_space.shape)  # (96, 96, 3)

env = FlattenObservation(env)
print(env.observation_space.shape)  # (27648,)

obs, _ = env.reset()
print(obs.shape)  # (27648,)
env.close()

Related Pages

Implements Principle

• Principle:Farama_Foundation_Gymnasium_Observation_Wrapping

Requires Environment

• Environment:Farama_Foundation_Gymnasium_Python_3_10_Runtime