An AI-Driven Affective Lighting System for Stress Reduction and Satisfaction in Office Environments

For years, dynamic lighting systems have been designed to supplement daylight and meet task-illumination requirements. However, the influence of light on humans extends beyond adequacy, as lighting attributes and their interactions have a profound impact on human psychology, including mood, and perception of space. This study presents a context-aware and user-centered lighting control framework for office environments that offers lighting modes focused on calmness, activation, and productivity designed to support office work through context-dependent presets and personalization. Unlike conventional dynamic systems that primarily respond to daylight levels, this framework links human affect to environmental conditions, such as blinds position and sky state, and adjusts correlated color temperature (CCT) and brightness accordingly. The system employs white light tones (2700K - 6500K) to both support task performance and influence users' affective state, ensuring suitability for everyday office environments. The system couples derived lighting presets and modes grounded in a user study with a reinforcement learning (RL) framework using a contextual bandit approach (LinUCB), which is designed to fine-tune lighting conditions to individual preferences through ongoing interaction. The implementation integrates a sensing system and a lighting control architecture that enables continuous real-time adaptation.