Implementation:Farama Foundation Gymnasium AntEnv 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 Ant v5 MuJoCo locomotion environment provided by Gymnasium.

Description

The Ant environment is based on Schulman, Moritz, Levine, Jordan, and Abbeel's work in "High-Dimensional Continuous Control Using Generalized Advantage Estimation". The ant is a 3D quadruped robot consisting of a torso (free rotational body) with four legs attached to it, where each leg has two body parts. The goal is to coordinate the four legs to move in the forward (right) direction by applying torque to the eight hinges connecting the body parts. The reward function combines a healthy reward, a forward reward based on x-velocity, a control cost penalty, and a contact cost penalty. The episode terminates if the ant becomes unhealthy (z-coordinate outside the healthy range or non-finite state values).

Usage

Use this environment for benchmarking continuous control and locomotion reinforcement learning algorithms. It is commonly used in research papers for evaluating policy gradient methods, model-based RL, and hierarchical RL approaches. The v5 version adds support for custom MuJoCo XML models, configurable frame skip, and improved reward/observation handling over v4.

Code Reference

Source Location

Repository: Farama_Foundation_Gymnasium
File: gymnasium/envs/mujoco/ant_v5.py

Signature

class AntEnv(MujocoEnv, utils.EzPickle):
    def __init__(
        self,
        xml_file: str = "ant.xml",
        frame_skip: int = 5,
        default_camera_config: dict[str, float | int] = DEFAULT_CAMERA_CONFIG,
        forward_reward_weight: float = 1,
        ctrl_cost_weight: float = 0.5,
        contact_cost_weight: float = 5e-4,
        healthy_reward: float = 1.0,
        main_body: int | str = 1,
        terminate_when_unhealthy: bool = True,
        healthy_z_range: tuple[float, float] = (0.2, 1.0),
        contact_force_range: tuple[float, float] = (-1.0, 1.0),
        reset_noise_scale: float = 0.1,
        exclude_current_positions_from_observation: bool = True,
        include_cfrc_ext_in_observation: bool = True,
        **kwargs,
    )

Import

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

I/O Contract

Inputs

Name	Type	Required	Description
action	np.ndarray (8,)	Yes	Torques applied to the 8 hinge joints connecting leg segments, range [-1, 1]

Outputs

Name	Type	Description
observation	np.ndarray (105,)	State vector: qpos (13 elements), qvel (14 elements), cfrc_ext (78 elements); x,y excluded by default
reward	float	healthy_reward + forward_reward - ctrl_cost - contact_cost
terminated	bool	True if ant is unhealthy (z outside [0.2, 1.0] or non-finite state)
truncated	bool	Episode truncation (handled by TimeLimit wrapper)
info	dict	Contains x_position, y_position, distance_from_origin, x_velocity, y_velocity, reward_forward, reward_ctrl, reward_contact, reward_survive

Usage Examples

import gymnasium as gym

env = gym.make("Ant-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