Development and Experimental Validation of an Embedded Intelligent Street Lighting System Based on Traffic Flow Detection and Cloud Cover Monitoring

ABSTRACT

Street lighting accounts for up to 30%–50% of municipal electricity consumption, yet conventional systems operate on fixed schedules with no adaptation to real‐time environmental or traffic conditions, resulting in significant energy waste. To address this challenge, this paper presents the design and experimental validation of an intelligent and energy‐efficient street lighting system based on artificial intelligence and embedded Internet of Things technologies. The proposed system dynamically adjusts light‐emitting diode lighting intensity according to real‐time road traffic flow and ambient cloud cover to improve energy efficiency while ensuring user safety and visual comfort. Road traffic data are extracted using the You Only Look Once version 8 (YOLOv8) object detection model applied to video streams, whereas cloud cover conditions are indirectly estimated through a light dependent resistor sensor. The system is implemented on a Raspberry Pi platform with message queuing telemetry transport communication for real‐time data exchange. Three intelligent control approaches—fuzzy logic, artificial neural networks (ANN), and adaptive neuro‐fuzzy inference system (ANFIS)—are developed and experimentally compared on a physical lighting prototype under four representative operating scenarios combining different traffic‐flow and cloud‐cover conditions. Results demonstrate that all models effectively adapt lighting power to environmental conditions, with fuzzy logic providing stable responses, ANN ensuring a good compromise between accuracy and smooth control and ANFIS offering higher responsiveness and precision. The proposed architecture demonstrates the feasibility of low‐cost intelligent street lighting solutions for smart city energy optimization.