Smart cities today are not limited to smart traffic lights or data analysis platforms. Technology is increasingly being used to address everyday urban maintenance tasks as well. One such solution is being developed in the final thesis of Laurynas Zalaga, a master’s student at the Faculty of Fundamental Sciences at VILNIUS TECH. Studying in the „Management of Artificial Intelligence Solutions“ programme, the student decided to explore how drones and artificial intelligence (AI) could be used to automate the monitoring of urban lighting systems and identify faults more quickly.
Laurynas Zalaga
„The idea came from looking for a practical solution to one of the everyday problems associated with maintaining urban infrastructure. Approximately half of the streetlights within the territory of the Vilnius City Municipality do not have controllers that would allow faults to be detected automatically. As a result, the responsible services often learn about a malfunctioning streetlight only after receiving reports from residents. This created a need for a reliable yet cost-effective solution allowing the condition of the city’s lighting systems to be monitored more efficiently,“ explains L. Zalaga.
Both drones and AI play an important role in the solution proposed by the student. First, a flight route is planned and transmitted to the drone. At a predetermined time, the drone automatically takes off from a docking station and follows the route, photographing locations where lighting poles are recorded according to their coordinates. After the flight, the collected data is transferred to a cloud platform, where it is analysed by a pre-trained AI model. By evaluating the images, the model can accurately determine which streetlights are not working and which ones require repairs.
Image collected by drones
„Such a solution would provide tangible benefits to the city. Faults would be detected more quickly and therefore repaired more promptly, while the quality of urban lighting would improve. At the same time, infrastructure maintenance work could be organised more efficiently and dependence on random reports of faults could be reduced,“ the VILNIUS TECH student explains, describing the benefits of the solution.
However, according to L. Zalaga, urban lighting is only one of the potential areas where the technology could be applied. Images collected by drones could also be used to assess the condition of other parts of urban infrastructure. Even images captured during nighttime could be analysed to assess road surface conditions, traffic control infrastructure, bicycle paths, road markings, and other urban elements. This would make it possible in one flight to collect data that could be useful for different areas of urban maintenance.
„The biggest challenge in developing the solution was preparing the dataset required to train the AI model. This required the shutdown of specific streetlights during nighttime to coincide with the drone’s flight and photography schedule. The process had to be carried out efficiently and promptly, as prolonged lack of lighting could cause inconvenience or even pose a danger to residents.
Image collected by drones
Additional challenges also arose when annotating the images. Sometimes, even when we knew which streetlights had been switched off, they were difficult to locate because of insufficient lighting.
Nevertheless, every stage of the project provided valuable practical experience and helped me better understand both AI and the possibilities for managing urban infrastructure,“ L. Zalaga shares.
In the opinion of the VILNIUS TECH student, in the future the management and maintenance of cities will increasingly become data-driven and based on real-time information. More and more processes will operate autonomously, while AI and other advanced technologies will help identify problems more quickly as well as make it possible to predict them before they become serious challenges. This would open up opportunities to create cities that operate more efficiently, are safer and smarter.
You can find this article in the September 2026 issue of Sapere Aude.
