Implementation:Haosulab ManiSkill RoboCasaKitchen

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

Overview

Concrete implementation of the RoboCasa kitchen environment in ManiSkill, providing realistic kitchen scenes for mobile manipulation.

Description

The RoboCasaKitchenEnv initializes a full kitchen environment using the RoboCasa scene builder. It supports placing a Fetch robot (or no robot) in a realistic kitchen scene with various fixtures, objects, and layout configurations.

Registered as RoboCasaKitchen-v1 with max_episode_steps=100 and asset_download_ids=["RoboCasa"]. Supported robots are fetch and none. The reward mode is "none" only, as this is primarily a scene environment for exploration and custom task development.

The environment supports configurable kitchen layouts through scene_builder_cls, placement samplers for kitchen objects, and various scene registrations from the RoboCasa dataset. It provides methods for sampling and placing kitchen objects within the scene.

Usage

Use this environment as a base for developing kitchen manipulation tasks with realistic scenes. It is particularly useful for sim-to-real transfer research and for building custom kitchen-based tasks.

Code Reference

Source Location

• Repository: Haosulab_ManiSkill
• File: mani_skill/envs/tasks/mobile_manipulation/robocasa/kitchen.py

Signature

@register_env(
    "RoboCasaKitchen-v1", max_episode_steps=100, asset_download_ids=["RoboCasa"]
)
class RoboCasaKitchenEnv(BaseEnv):
    SUPPORTED_ROBOTS = ["fetch", "none"]
    SUPPORTED_REWARD_MODES = ["none"]

Import

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

I/O Contract

Inputs

Name	Type	Required	Description
obs_mode	str	No	Observation mode
reward_mode	str	No	Reward mode: "none"
control_mode	str	No	Control mode for Fetch robot (if used)

Outputs

Name	Type	Description
obs	dict/array	Observation based on obs_mode
reward	float	Always 0 (reward mode "none" only)
terminated	bool	Whether episode ended
truncated	bool	Whether episode hit max steps (100)
info	dict	Environment information

Usage Examples

Basic Usage

import gymnasium as gym
import mani_skill.envs

env = gym.make("RoboCasaKitchen-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