ILLUMINATIO

Bob Tianqi Wei, Shm Almeda, Zhendong Xiao, Lintao Tang
UC Berkeley, Tsinghua University
2022-2024

-->ILLUMINATIO: Towards Biologically-Informed, AI-Driven Adaptive Environment Control with Situated Pervasive Illumination Devices (Full Text)
-->Intelligent Illuminating Product Design Based on Machine Learning (original version of ILLUMINATIO)

While working on smart home projects, I noticed a gap between what current smart lighting offers and what people actually need. Most smart lights just follow schedules or basic sensor triggers. I wanted to create something more intuitive - a light that truly understands and adapts to its user.

I imagined a desk lamp that could:
Learn from how people naturally adjust their lighting
Support healthy biological rhythms
Adapt to different activities automatically
Give users both convenience and control

Fig.1. Transitioning between four scenes while using ILLUMINATIO

Let me walk you through a typical day with ILLUMINATIO:

Morning Work Session
When you sit down at your computer, ILLUMINATIO provides bright, cool light that helps you feel alert and focused. It's smart enough to position itself to avoid screen glare.

Reading Time
Switch from computer to book? ILLUMINATIO notices and adjusts automatically. It focuses light directly on your reading material with perfect brightness for paper.

Evening Wind-Down
As night approaches, ILLUMINATIO gradually shifts to warmer, dimmer light. This helps your body prepare for sleep naturally - no more harsh light keeping you awake!

Understanding Human Needs

I started by researching how lighting affects us:
Office workers prefer different lighting for different tasks
Light influences our hormones and biological clock
Poor lighting can cause fatigue and reduce productivity

Prototyping

Fig. 2. Drawings in AutoCAD and the Functional Model

Functional Model:
Built around a Raspberry Pi 4 for processing
Integrated 2 color LED systems for task and ambient lighting
Added a camera for scene recognition
Designed a 4-servo motor system for flexible positioning
‍

Fig. 3. Sketches and Rendering
‍

Fig. 4. Modeling, Printing and Installing ILLUMINATIO

Concept Model:
Used Rhino to create a form that balances function and aesthetics
Designed a seamless blend of square and circular elements
Kept the structure minimal but expressive

The AI System

Fig. 5. "Focused Mode" uses computer vision to identify the book's location and responsively adjust how the light is projected.

Vision System
Used OpenCV for real-time scene analysis
Can detect objects like books and screens
Adjusts lighting position and intensity accordingly
‍

Fig. 6. User behavior data exhibits clustered pattern.
Fig. 7. The decreasing trend of prediction error

Fig. 8. Calculation of error

Learning System
Implemented a K-Nearest Neighbor model
Learns from each user adjustment
Gets noticeably better at predicting user preferences after just 16 interactions

Key Features in Action

When you're reading a book, ILLUMINATIO:
Detects the book's position
Adjusts light angle for optimal coverage
‍
Throughout the day, it automatically:
Provides energizing light in the morning
Offers balanced light during work hours
Transitions to warm, calm light in the evening

Testing showed that ILLUMINATIO successfully:
Learned user preferences over time
Reduced the need for manual adjustments
Maintained appropriate lighting for different activities

What's Next?

This project opens up exciting questions about adaptive environments:
How will people interact with AI-driven furniture?
Can we make technology that's both intelligent and engaging?
What's the right balance between automation and user control?

ILLUMINATIO

ALL WORKS