Implementation:ARISE Initiative Robosuite LightingModder

Metadata:

Overview

Concrete class for randomizing lighting properties in MuJoCo simulations provided by the robosuite modding utilities.

Description

LightingModder perturbs light position, direction, specular/ambient/diffuse intensities, and active state. Default direction perturbation of 0.35 radians is ~20 degrees. Inherits from BaseModder.

Usage

Used internally by DomainRandomizationWrapper. Can be used standalone.

Code Reference

Source: robosuite

File: robosuite/utils/mjmod.py

Lines: L62-515

Signature:

class LightingModder(BaseModder):
    def __init__(self, sim, random_state=None, light_names=None,
                 randomize_position=True, randomize_direction=True,
                 randomize_specular=True, randomize_ambient=True,
                 randomize_diffuse=True, randomize_active=True,
                 position_perturbation_size=0.1, direction_perturbation_size=0.35,
                 specular_perturbation_size=0.1, ambient_perturbation_size=0.1,
                 diffuse_perturbation_size=0.1):

Import: from robosuite.utils.mjmod import LightingModder

I/O Contract

Inputs:

sim (required): MuJoCo simulation instance
random_state (optional): Random state for reproducibility
light_names (optional): List of light names to randomize
randomize_position (bool, optional, default=True): Whether to randomize light position
randomize_direction (bool, optional, default=True): Whether to randomize light direction
randomize_specular (bool, optional, default=True): Whether to randomize specular intensity
randomize_ambient (bool, optional, default=True): Whether to randomize ambient intensity
randomize_diffuse (bool, optional, default=True): Whether to randomize diffuse intensity
randomize_active (bool, optional, default=True): Whether to randomize active state
position_perturbation_size (float, optional, default=0.1): Size of position perturbation
direction_perturbation_size (float, optional, default=0.35): Size of direction perturbation in radians
specular_perturbation_size (float, optional, default=0.1): Size of specular intensity perturbation
ambient_perturbation_size (float, optional, default=0.1): Size of ambient intensity perturbation
diffuse_perturbation_size (float, optional, default=0.1): Size of diffuse intensity perturbation

Methods:

randomize(): Apply randomization to lighting properties
save_defaults(): Save current lighting properties as defaults
restore_defaults(): Restore lighting properties to saved defaults

Usage Examples

LightingModder is typically used via the DomainRandomizationWrapper by providing lighting_randomization_args:

from robosuite.wrappers import DomainRandomizationWrapper

env = DomainRandomizationWrapper(
    env,
    seed=0,
    randomize_lighting=True,
    lighting_randomization_args={
        'randomize_position': True,
        'randomize_direction': True,
        'randomize_specular': True,
        'randomize_ambient': True,
        'randomize_diffuse': True,
        'randomize_active': True,
        'position_perturbation_size': 0.1,
        'direction_perturbation_size': 0.35,
        'specular_perturbation_size': 0.1,
        'ambient_perturbation_size': 0.1,
        'diffuse_perturbation_size': 0.1
    }
)

Standalone usage:

from robosuite.utils.mjmod import LightingModder
import numpy as np

# Create modder
random_state = np.random.RandomState(0)
lighting_modder = LightingModder(
    sim=env.sim,
    random_state=random_state,
    randomize_direction=True,
    direction_perturbation_size=0.35
)

# Save default lighting
lighting_modder.save_defaults()

# Randomize lighting
lighting_modder.randomize()

# Restore defaults if needed
lighting_modder.restore_defaults()

Related Pages

Page Connections

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

Principle

Implementation

Heuristic

Environment