Implementation:Haosulab ManiSkill QuadrupedSpin

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

Overview

Concrete implementation of the quadruped spinning locomotion task in ManiSkill.

Description

The QuadrupedSpinEnv requires a quadruped robot to spin in place as fast as possible, rewarded by its z-axis angular velocity. One robot-specific variant is registered:

• AnymalC-Spin-v1 (max_episode_steps=200): AnymalC robot spins in place.

The robot is initialized in a standing position on a ground plane. Failure occurs when the robot falls (main body hits the ground). The reward is 2 * angular_velocity_z plus penalties for z-axis linear velocity, xy angular velocity, undesired contacts, and joint deviation from standing pose. Falling results in a reward of -100. Supported robots: anymal_c, unitree_go2_simplified_locomotion. Reward modes include "dense", "normalized_dense", "sparse", and "none".

Usage

Use this environment for agile quadruped locomotion research focused on rotational movement. The spinning task tests whole-body coordination for in-place rotation.

Code Reference

Source Location

• Repository: Haosulab_ManiSkill
• File: mani_skill/envs/tasks/quadruped/quadruped_spin.py

Signature

class QuadrupedSpinEnv(BaseEnv):
    SUPPORTED_ROBOTS = ["anymal_c", "unitree_go2_simplified_locomotion"]
    ...

@register_env("AnymalC-Spin-v1", max_episode_steps=200)
class AnymalCSpinEnv(QuadrupedSpinEnv): ...

Import

import gymnasium as gym
import mani_skill.envs
env = gym.make("AnymalC-Spin-v1")

I/O Contract

Inputs

Name	Type	Required	Description
obs_mode	str	No	Observation mode
reward_mode	str	No	Reward mode: "dense", "normalized_dense", "sparse", "none"
control_mode	str	No	Control mode for the quadruped robot

Outputs

Name	Type	Description
obs	dict/array	Observation including root linear/angular velocities
reward	float	Dense reward from angular velocity + stability penalties
terminated	bool	Whether robot has fallen
truncated	bool	Whether episode hit max steps (200)
info	dict	Contains fail, is_fallen flags

Usage Examples

Basic Usage

import gymnasium as gym
import mani_skill.envs

env = gym.make("AnymalC-Spin-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