Software-Defined Lighting

For much of the past century, indoor lighting has been based on incandescent or gas-discharge technology. But, with LED lighting experiencing a 20x/decade increase in flux density, 10x/decade decrease in cost, and linear improvements in luminous efficiency, solid-state lighting is finally cost-competitive with the status quo. As a result, LED lighting is projected to reach over 70% market penetration by 2030. In this dissertation, we claim that solid-state lighting's real potential has been barely explored, that now is the time to explore it, and that new lighting platforms and applications can drive lighting far beyond its roots as an illumination technology. Scaling laws make solid-state lighting competitive with conventional lighting, but two key features make solid-state lighting an enabler for many new applications: the high switching speeds possible using LEDs, and the color palettes realizable with Red-Green-Blue-White (RGBW) multi-chip assemblies.

For this dissertation, we have explored the post-illumination potential of LED lighting in applications as diverse as health, entertainment, visible light communications, indoor positioning, smart dust time synchronization, and embedded device configuration, with an eventual eye toward supporting all of them using a shared lighting infrastructure and under a unified system architecture that provides software-control over lighting. To explore the space of software-defined lighting (SDL), we design a compact, flexible, and networked SDL platform to allow researchers to rapidly test new ideas. Using this platform, we demonstrate the viability of several applications, including multi-luminaire synchronized communication to a photodiode receiver, communications to mobile phone camera, and indoor positioning using unmodified mobile phones. Furthermore, we present a unified, modularized SDL architecture, and design a SDL that provides flexibility to swap subsystems. The SDL supports many applications concurrently on a single, shared lighting testbed.