Enabling adaptive road lighting through lighting class prediction with real time and historical data

Abstract

Road lighting plays a crucial role in transportation systems as it significantly contributes to ensuring safety for road users during hours when there is not enough daylight. A road lighting system that meets prescribed standards aims to prevent traffic accidents that may occur as a result of insufficient lighting while providing visual comfort conditions for road users. In these days when it is necessary strategize and execute energy consumption reduction plans in various areas in order to use limited energy resources more efficiently. As a substantial contributor to urban energy consumption, road lighting is currently undergoing the development of sustainable and energy-efficient systems through the formulation of standards and implementation of projects. The first method that comes to mind in order to reduce energy consumption is to replace existing luminaries with LED luminaries. It has been proven in many studies that road lighting made with LED luminaries is much more energy efficient compared to other types of light sources. One of the benefits of LEDs compared to other conventional lamp varieties is the ability to regulate the amount of luminous flux emitted by the luminaries. By utilizing the dimming feature, it is possible to provide illumination at different levels for different needs from the same light source. Adaptive road lighting (ARL) adjusts luminous flux and luminance levels based on environmental conditions and traffic flow, enhancing energy efficiency and driving safety by ensuring optimal lighting conditions. Through the utilization of environmental data and traffic analysis, these systems are capable of adapting lighting levels to ensure optimal illumination while simultaneously lowering energy usage. The effectiveness of such systems largely depends on accurate predictions of traffic data and weather conditions. Therefore, to ensure the effectiveness of adaptive road lighting systems, it is crucial to have a comprehensive understanding of the road's characteristics. By integrating the real-time traffic flow data with the historical data gathered from the road sensors, more precise predictions can be made. In addition to ARL systems in the form of simulation or prototypes, which have become increasingly popular in recent years, it is essential to provide guidance for sample applications, reports, and specification studies. This study explores different strategies for analyzing traffic data, ranging from traditional statistical models to advanced machine learning algorithms. It highlights the significance of various factors, including historical traffic data, weather conditions, and road topology, in the development of precise prediction models. The effectiveness of these prediction techniques in diverse urban environments is presented with case studies. In this study, the focus is on exploring various strategies that can be employed for real-time data processing, model updating, and decision-making algorithms. The primary aim of these strategies is to optimize lighting adjustments by utilizing predicted lighting class. Additionally, the study investigates the potential synergies that can be derived from integrating these strategies with smart city initiatives and traffic management systems. By combining predictiveanalytics with intelligent lighting infrastructure, this research highlights the wider impact that can be achieved, underscoring the improvement in overall efficiency and effectiveness. The road chosen as the case study is located within the city and provides services in two different directions. It consists of a total of six lanes, each is 3 meters wide. The lighting class of the road had been determined as M2. In order to provide the specified lighting class, the luminaires were placed on 13-meter-high lighting poles with a distance of 40 meters between them. Traffic flow rate and vehicle speed data taken with 2-minute intervals for each lane on the road has been recorded for a year and new data continues to be collected. In order to identify appropriate methodologies for this objective, existing traffic data of the pilot road were visualized, and a model was developed that can estimate the lighting class for a desired period from a selected time of the day with minimum error using machine learning algorithms. In the most successful models created, the accuracy of hourly lighting class predictions exceeds 97%. In addition to predicting lighting classes, the calculation of energy consumption based on the established lighting scenarios was conducted, and the energy saving rates for the suggested adaptive road lighting system were explicitly outlined. Given the potential decrease in driving visibility and comfort during rainy weather, it has been determined that the road lighting setup will adhere to M2 lighting class requirements at all times, with the suspension of ARL scenarios. This research results that the transformation of road lighting systems into adaptive road lighting through automation strategies aligned with traffic flow rates, vehicle speeds, and meteorological conditions could result in elevated rates of energy savings. Overall, this study provides a comprehensive analysis of the role of lighting class prediction in enhancing adaptive road lighting systems and, contributes to advancing knowledge in optimizing adaptive road lighting systems for energy efficiency and driving safety, with potential practical applications in smart city infrastructure. A detailed field study needs to be continued to ensure that the applied scenarios and strategies do not adversely affect traffic safety and visual comfort conditions of road users. The need for standardized data collection and sampling protocols, storage of big data, and privacy issues will be some of the challenges and topics of future discussion in this field.