A comparative review of machine learning and deep learning architectures for correlated colour temperature control in smart road lighting

Abstract

The review provides a focused analysis of machine learning (ML) and deep learning (DL) models for managing correlated colour temperature (CCT) in adaptive road lighting systems. As urban infrastructure evolves towards smart, responsive technologies, the ability to dynamically control lighting parameters is critical for enhancing safety and energy efficiency. We systematically evaluate the motivation for moving beyond traditional control logic to data-driven ML methodologies. We conduct a detailed survey of applicable ML paradigms, including supervised, unsupervised and reinforcement learning. Specific techniques such as regression models, support vector machines and fuzzy logic systems are critically assessed for their utility in CCT selection. Furthermore, we explore the advanced capabilities of DL architectures, examining how convolutional neural networks can leverage visual data for rich contextual awareness and how recurrent neural networks can process time-series data for predictive control. A comprehensive comparative analysis evaluates these techniques against key criteria, including accuracy, computational complexity, data requirements and real-time applicability. This review synthesizes the key advantages and trade-offs of each approach, offering guidance for researchers and practitioners in selecting and implementing the most suitable computational strategies for developing next generation, intelligent road lighting systems.