Lighting

May 20, 2026 · View on GitHub

RoboLab uses IsaacLab's light spawners to configure scene lighting. Lighting configs are @configclass objects that add light sources to the simulation and are passed as lighting_cfg during environment registration. These configs can live in your own repository.

Built-in Lighting Configurations

RoboLab ships several lighting presets in robolab/variations/lighting.py:

Sphere Lights

Config	Intensity	Color	Position	Notes
`SphereLightCfg`	5,000	White	(0.0, -0.6, 0.7)	Default for most registrations
`ExtremelyDimSphereLightCfg`	50	White	(0.0, -0.6, 0.7)	For testing low-light conditions
`RedSphereLightCfg`	100,000	Red	(0.0, 0.0, 1.0)	Colored light variation
`GreenSphereLightCfg`	100,000	Green	(0.0, 0.0, 1.0)	Colored light variation
`BlueSphereLightCfg`	100,000	Blue	(0.0, 0.0, 1.0)	Colored light variation

Directional Lights

Config	Direction	Intensity	Notes
`FrontDirectionalLightCfg`	Front	3,000	Facing the robot from the front
`BehindDirectionalLightCfg`	Behind	3,000	Facing the robot from behind
`TopDownDirectionalLightCfg`	Top-down	3,000	Overhead lighting
`LeftDirectionalLightCfg`	Left	3,000	Side lighting from the left
`RightDirectionalLightCfg`	Right	3,000	Side lighting from the right

Using a Lighting Config

Import the config and pass it as lighting_cfg in your registration function (see Environment Registration for the full example):

from robolab.variations.lighting import SphereLightCfg

# Inside your register_envs() function:
auto_discover_and_create_cfgs(
    lighting_cfg=SphereLightCfg,
    # ... other registration kwargs
)

Defining Custom Lighting

A lighting config is a @configclass with one or more AssetBaseCfg fields, each spawning a light. It can live in your own repository.

Sphere Light

import isaaclab.sim as sim_utils
from isaaclab.assets import AssetBaseCfg
from isaaclab.utils import configclass


@configclass
class MySphereLightCfg:
    my_light = AssetBaseCfg(
        prim_path="/World/my_sphere_light",
        spawn=sim_utils.SphereLightCfg(
            intensity=8000,
            color=(1.0, 0.95, 0.9),   # Warm white
            radius=0.1,
        ),
        init_state=AssetBaseCfg.InitialStateCfg(pos=(0.0, -0.5, 1.0)),
    )

Directional (Distant) Light

@configclass
class MyDirectionalLightCfg:
    directional_light = AssetBaseCfg(
        prim_path="/World/directional_light",
        spawn=sim_utils.DistantLightCfg(
            intensity=3000,
            angle=0.53,
            exposure=3.0,
        ),
        init_state=AssetBaseCfg.InitialStateCfg(
            pos=(0.0, 0.0, 5.0),
            rot=(0.7071, 0.0, 0.7071, 0.0),  # Front-facing
        ),
    )

Multiple Lights

You can define multiple light fields in one config:

@configclass
class MyStudioLightingCfg:
    key_light = AssetBaseCfg(
        prim_path="/World/key_light",
        spawn=sim_utils.SphereLightCfg(intensity=10000),
        init_state=AssetBaseCfg.InitialStateCfg(pos=(0.5, -0.8, 1.2)),
    )
    fill_light = AssetBaseCfg(
        prim_path="/World/fill_light",
        spawn=sim_utils.SphereLightCfg(intensity=3000),
        init_state=AssetBaseCfg.InitialStateCfg(pos=(-0.5, -0.3, 0.8)),
    )

Available Light Types

IsaacLab provides the following light spawners through isaaclab.sim:

Spawner	Description
`sim_utils.SphereLightCfg`	Point light from a sphere; parameters: `intensity`, `color`, `radius`
`sim_utils.DistantLightCfg`	Directional light (sun-like); parameters: `intensity`, `angle`, `exposure`
`sim_utils.DomeLightCfg`	Environment map light (HDR/EXR); used for Backgrounds
`sim_utils.DiskLightCfg`	Area light from a disk
`sim_utils.CylinderLightCfg`	Area light from a cylinder
`sim_utils.RectLightCfg`	Area light from a rectangle

Lighting Variation for Robustness Testing

To evaluate policy robustness under different lighting conditions, register multiple environment variants with different lighting_cfg values. See robolab/registrations/droid/auto_env_registrations_lighting_variations.py for a complete example.

The built-in evaluation script policies/pi0_family/run_lighting.py runs benchmarks across lighting conditions automatically.