Light is a visual medium and a major environmental cue for human circadian regulation. Under visual-quality constraints, daylight-related spectral approximation and control of non-visual biological effects form a challenging multi-objective trade-off. This study proposes a hardware-explicit computational framework for circadian-oriented spectral optimization using five-channel LED basis spectra characterized from discrete SPD data. The framework integrates static scenario-constrained spectral optimization, time-series daylight-emulation analysis, and exploratory sleep-outcome characterization. The static model generates implementable spectra for fixed daytime and nighttime scenarios, while the time-series model evaluates the ability of the same LED basis to approximate daylight-related temporal characteristics. Sleep outcomes are analyzed from available repeated-measures data collected under static nighttime conditions: conventional lighting, darkness, and a predefined reference optimized nighttime condition derived from the static scenario. In a small-sample exploratory analysis (n = 11), the optimized condition showed a significant N3% difference relative to darkness (p = 0.03, |d| = 0.97), and the comparison with conventional lighting and most other sleep endpoints did not reach statistical significance. These findings suggest condition-associated variation in sleep architecture within the present dataset and provide exploratory evidence for hypothesis generation. Overall, the proposed framework offers a reproducible route for multi-channel LED spectral trade-off design, daylight-emulation assessment, and preliminary sleep-outcome characterization in circadian-oriented healthy-lighting applications.

TOPSIS-entropy weight method, differential evolution, SLSQP, time series constrained optimization, spectral optimization