Implementation:Haosulab ManiSkill CartPole

Knowledge Sources	Haosulab_ManiSkill
Domains	Robotics, Simulation, Control
Last Updated	2026-02-15 08:00 GMT

Overview

Concrete implementation of the CartPole control task environment in ManiSkill, adapted from the DeepMind Control Suite.

Description

The CartPole environment implements the classic cart-pole balancing and swingup tasks. A cart slides along a rail and a pole is attached via a hinge joint. The robot is defined inline via MJCF as CartPoleRobot.

Registered variants:

• MS-CartPoleBalance-v1 (max_episode_steps=500): Balance the pole upright from near-upright initial conditions.
• MS-CartPoleSwingUp-v1 (max_episode_steps=500): Swing the pole from a hanging position to upright.

The robot uses PD joint position control on the slider joint, with a passive controller on the hinge. Rewards are computed from the pole's upright angle, cart centering, and angular velocity penalties. Reward modes include "dense", "normalized_dense", and "none".

Usage

Use this environment for benchmarking basic control algorithms. CartPole Balance provides a simpler task while CartPole SwingUp requires more sophisticated control strategies.

Code Reference

Source Location

• Repository: Haosulab_ManiSkill
• File: mani_skill/envs/tasks/control/cartpole.py

Signature

@register_env("MS-CartPoleBalance-v1", max_episode_steps=500)
class CartPoleBalanceEnv(CartPoleEnvBase): ...

@register_env("MS-CartPoleSwingUp-v1", max_episode_steps=500)
class CartPoleSwingUpEnv(CartPoleEnvBase): ...

Import

import gymnasium as gym
import mani_skill.envs
env = gym.make("MS-CartPoleBalance-v1")

I/O Contract

Inputs

Name	Type	Required	Description
obs_mode	str	No	Observation mode (default: "state")
reward_mode	str	No	Reward mode: "dense", "normalized_dense", "none"
control_mode	str	No	Control mode for the cart

Outputs

Name	Type	Description
obs	dict/array	Observation including cart position, pole angle, velocities
reward	float	Reward based on upright angle, cart position, angular velocity
terminated	bool	Whether episode ended
truncated	bool	Whether episode hit max steps (500)
info	dict	Contains evaluation metrics

Usage Examples

Basic Usage

import gymnasium as gym
import mani_skill.envs

env = gym.make("MS-CartPoleBalance-v1", obs_mode="state", render_mode="rgb_array")
obs, info = env.reset()
for _ in range(100):
    action = env.action_space.sample()
    obs, reward, terminated, truncated, info = env.step(action)
    if terminated or truncated:
        obs, info = env.reset()
env.close()

Related Pages

• Environment:Haosulab_ManiSkill_Python_SAPIEN_Core