Researchers Use Machine Learning to Accurately Model Street Lighting Systems
A recent study by a group of three researchers from West Bengal has underscored the effectiveness of the application of artificial neural networks (ANNs), a type of machine learning now commonly used to model complex engineering problems, to easily calculate lighting parameters and energy efficiency of street lighting systems. Their study, entitled ‘Designing metal halide-based road illumination systems in developing countries using regression and neural networks’ has been published in the Results in Engineering journal in its 26th volume, a distinguished journal published by Elsevier, and has already been cited twice in other research papers within months of publication.
Their study is centred around answering three research questions:
- Do ANNs perform better than the traditional multiple regression models in the prediction of lighting parameters and energy demand of street lighting systems?
- Will ANN models be suitable for municipal lighting departments for rapid planning and good estimation of lighting parameters?
- What design variables are important to consider for good street lighting?
The researchers, Sourin Bhattacharya, Khondekar Lutful Hassan, and Pallav Dutta, made street lighting simulations with metal halide-based gas discharge lamps of high power demand in a standard lighting software platform and generated a dataset with 80,000 samples, considering 10 inputs and 4 outputs. The generated dataset considered important design parameters – height of lamp poles, distance between consecutive lamp poles, lamp orientation, power supply, and road surface reflection properties for dry road surfaces. They formulated several models using conventional statistics and ANN, and made a comparative evaluation of the models for some standard designs. Their ANN model, trained with the ‘Levenberg-Marquardt’ algorithm, was shown to make good predictions of light level, uniformity, and energy efficiency. They compared the performance of their ANN model with a traditional regression model and observed that their ANN model made better predictions. Also, their study found that installed wattage per kilometre of road length and specular (mirror-like) reflection properties of road surface are the two important design variables that influence the effectiveness of street lighting.
The research article discussed the effectiveness of ANNs as a powerful predictive tool for street lighting planning, as resources for extensive lighting simulation are argued to be often quite limited at the municipal level. The findings reported in the research also indicate that lighting simulation with machine learning can act as a potent tool for faster and cost-effective planning of street lighting projects.
About the Researchers
Sourin Bhattacharya, M.Tech., is a Public Official who is acting as a Visiting Faculty at the School of Illumination Science, Engineering and Design, Jadavpur University, Kolkata, West Bengal, where he is also pursuing a Ph.D. programme. He has published fourteen research articles on lighting in peer-reviewed journals and conference proceedings and performs research on daylighting, road lighting, lighting ergonomics, and indoor lighting.
Khondekar Lutful Hassan, Ph.D., is an Assistant Professor at the Department of Computer Science and Engineering, Aliah University, Kolkata, West Bengal. He is a Life Fellow of the Institution of Electronics and Telecommunications Engineers (IETE) and a Life Member of the Computer Society of India (CSI). He has served as a Program Co-Chair of various international conferences. His research interests include cybersecurity, ML, and wireless networks.
Pallav Dutta, M.Tech., is an Assistant Professor at the Department of Electrical Engineering, Aliah University, Kolkata, West Bengal, where he was the Head (Officiating) in 2019 – 2021. He is a Chartered Engineer of the Institution of Engineers (India). His research interests include road lighting, human-centric lighting, renewable energy, IoT, AI, and ML.
The paper can be accessed at: https://www.sciencedirect.com/science/article/pii/S2590123025010928
Citation: Bhattacharya, Sourin, Khondekar Lutful Hassan, and Pallav Dutta. "Designing Metal Halide-Based Road Illumination Systems in Developing Countries Using Regression and Neural Networks." Results in Engineering (2025): 105017.