인간중심조명을 위한 머신러닝 기반 저사양 임베디드 활용 밝기 및 색온도 최적화 시스템 개발
Development of a Machine Learning-based Brightness and Color Temperature Optimization System for Human-Centric Lighting using Low-spec Embedded Systems
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
his paper presents a brightness and color temperature optimization system for human-centric lighting (HCL) on low-spec embedded devices. Conventional HCL systems rely on rule-based methods, heavy computation, and cloud servers, limiting autonomy. We implemented a machine learning-based controller on the Raspberry Pi Zero 2 W, integrating sensor, HCL status, and weather data. Five algorithms—SVR, KNN, Ridge Regression, Random Forest, and MLP—were compared. Random Forest achieved the best accuracy (MAPE 0.96% dimming, 0.97% color temperature), while Ridge Regression had the fastest processing (0.067 s vs. 0.67 s for Random Forest). All algorithms ran stably with similar resource use (25% CPU, 70% memory). The system removes cloud dependency and supports autonomous small-scale HCL operation. A SQLite3-based logging function enables monitoring and optimization. This study shows the feasibility of a cost-effective, edge-computing HCL solution that adaptively controls lighting without high-performance servers.
Abstract
his paper presents a brightness and color temperature optimization system for human-centric lighting (HCL) on low-spec embedded devices. Conventional HCL systems rely on rule-based methods, heavy computation, and cloud servers, limiting autonomy. We implemented a machine learning-based controller on the Raspberry Pi Zero 2 W, integrating sensor, HCL status, and weather data. Five algorithms—SVR, KNN, Ridge Regression, Random Forest, and MLP—were compared. Random Forest achieved the best accuracy (MAPE 0.96% dimming, 0.97% color temperature), while Ridge Regression had the fastest processing (0.067 s vs. 0.67 s for Random Forest). All algorithms ran stably with similar resource use (25% CPU, 70% memory). The system removes cloud dependency and supports autonomous small-scale HCL operation. A SQLite3-based logging function enables monitoring and optimization. This study shows the feasibility of a cost-effective, edge-computing HCL solution that adaptively controls lighting without high-performance servers.