Implementation:Haosulab ManiSkill KitchenObjectUtils

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

Overview

Concrete tool providing utility classes and functions for managing kitchen object categories, model discovery, and object instantiation in RoboCasa environments.

Description

This module defines the ObjCat class and utility functions for working with kitchen objects in RoboCasa. It bridges between the object category definitions in kitchen_objects.py and the actual MJCF model files on disk.

ObjCat encapsulates metadata for an object category, including:

Category name and types (e.g., "fruit", "meat", "packaged_food")
Boolean properties: graspable, washable, microwavable, cookable, freezable
Model folders pointing to MJCF model directories
Physics parameters: scale, solimp, solref, density, friction, priority
Support for both AI-generated (aigen) and Objaverse object models

The module provides model discovery by scanning the BASE_ASSET_ZOO_PATH (which points to the RoboCasa objects asset directory) for available MJCF models. It supports model exclusion lists and separate configurations for AI-generated vs. Objaverse object variants.

Key utility functions include:

Object category registration and lookup from the OBJ_CATEGORIES dictionary
Model path resolution for individual object instances
XML metadata extraction for object dimensions and properties

Usage

Used internally by the RoboCasa scene builder and placement samplers to discover available kitchen object models and instantiate them with appropriate physics properties.

Code Reference

Source Location

Repository: Haosulab_ManiSkill
File: mani_skill/utils/scene_builder/robocasa/objects/kitchen_object_utils.py

Signature

class ObjCat:
    def __init__(self, name, types, model_folders=None, exclude=None,
                 graspable=False, washable=False, microwavable=False,
                 cookable=False, freezable=False, scale=1.0,
                 solimp=(0.998, 0.998, 0.001), solref=(0.001, 2),
                 density=100, friction=(0.95, 0.3, 0.1),
                 priority=None, aigen_cat=False): ...

Import

from mani_skill.utils.scene_builder.robocasa.objects.kitchen_object_utils import ObjCat

I/O Contract

Parameter	Type	Description
`name`	`str`	Object category name (e.g., "apple", "mug")
`types`	`tuple` or `str`	Category types (e.g., "fruit", "drink")
`model_folders`	`list`	Directories containing MJCF models
`exclude`	`list`	Model names to exclude from the category
`graspable`	`bool`	Whether objects in this category can be grasped
`scale`	`float`	Default scale for objects in this category
`density`	`float`	Default density for physics simulation (default 100)
`aigen_cat`	`bool`	True if using AI-generated models, False for Objaverse

Usage Examples

from mani_skill.utils.scene_builder.robocasa.objects.kitchen_object_utils import ObjCat

# Create an object category
apple_cat = ObjCat(
    name="apple",
    types=("fruit",),
    graspable=True,
    washable=True,
    cookable=True,
    scale=1.0,
)

# Access category properties
print(apple_cat.graspable)  # True
print(apple_cat.types)      # ("fruit",)

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment