Implementation:Farama Foundation Gymnasium SwimmerEnv 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 Swimmer v5 MuJoCo locomotion environment provided by Gymnasium.

Description

The Swimmer v5 environment corresponds to the Swimmer described in Remi Coulom's PhD thesis "Reinforcement Learning Using Neural Networks, with Applications to Motor Control". The default swimmer consists of three segments (links) connected by two rotor joints. The swimmer is suspended in a 2D pool and the goal is to move as fast as possible to the right by applying torque to the rotors and exploiting fluid friction. The reward is forward_reward - ctrl_cost. The Swimmer never terminates; episodes end through truncation. The v5 version adds support for custom MuJoCo XML models, configurable frame_skip, observation_structure, non-empty reset info, and consistent reward naming (reward_forward instead of reward_fwd).

Usage

Use this environment for benchmarking RL algorithms on a swimming locomotion task. The Swimmer is notable for testing algorithms in environments with simple dynamics but challenging reward landscapes. The v5 version is recommended for new research.

Code Reference

Source Location

Repository: Farama_Foundation_Gymnasium
File: gymnasium/envs/mujoco/swimmer_v5.py

Signature

class SwimmerEnv(MujocoEnv, utils.EzPickle):
    def __init__(
        self,
        xml_file: str = "swimmer.xml",
        frame_skip: int = 4,
        default_camera_config: dict[str, float | int] = None,
        forward_reward_weight: float = 1.0,
        ctrl_cost_weight: float = 1e-4,
        reset_noise_scale: float = 0.1,
        exclude_current_positions_from_observation: bool = True,
        **kwargs,
    )

Import

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

I/O Contract

Inputs

Name	Type	Required	Description
action	np.ndarray (2,)	Yes	Torques applied to the two rotor joints, range [-1, 1]

Outputs

Name	Type	Description
observation	np.ndarray (8,)	State vector: qpos (3 angles, x/y excluded by default), qvel (5 velocities)
reward	float	forward_reward - ctrl_cost
terminated	bool	Always False (Swimmer never terminates)
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

Usage Examples

import gymnasium as gym

env = gym.make("Swimmer-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