This paper presents an AI-driven intelligent street lighting system that integrates Reinforcement Learning (RL) with cloud computing, edge AI processing, and multi-sensor fusion to optimize energy consumption while enhancing public safety in smart cities. Traditional street lighting operates at fixed brightness, leading to substantial energy waste. The proposed sys- tem dynamically adjusts illumination based on real-time inputs from radar sensor, LDR sensor, and optional camera module. An AI classifier distinguishes between humans, vehicles, and animals to provide context-aware lighting control. The system architecture consists of a cloud/AI model server for training and classification, an ESP32 microcontroller for edge AI processing, radar sensor for motion detection, LDR for ambient light sensing, optional camera for visual confirmation, an AI classifier for object type identification, and an LED driver with PWM control for actuation. Simulation results demonstrate energy savings of 35- 45%, response times of 0.8-1.5 seconds, and 96% classification accuracy, outperforming conventional static systems and super- vised ML models such as SVM, Decision Tree, and KNN. The scalable cloud-edge architecture enables deployment across large urban networks, contributing to sustainable and safe smart city infrastructure.

This paper presented an AI-driven intelligent street lighting system that combines Reinforcement Learning, cloud-edge architecture, and AI-based object classification. The system dynamically adjusts brightness based on real-time sensor data and object type (human/vehicle/animal), achieving 38% energy savings, 0.8-1.5 second response times, and 96% pattern detection accuracy. The AI classifier achieves 94% accuracy in distinguishing humans, vehicles, and animals, enabling class- specific brightness optimization. The cloud-edge architecture with ESP32 edge AI processing provides low-latency inference (50 ms) while maintaining scalability. The RCWL-0516 radar sensor provides superior detection range (7-10 m) and weather immunity. Future work will focus on federated learning, multi- sensor fusion, and predictive analytics.

[1]M. Mohan, V. V. S. S. S. Chakravarthy, and K. V. S. S. R. Murthy, ”IoT based intelligent street lighting system for smart cities,” in Proc. Int. Conf. Inventive Comput. Technol., 2018, pp. 1-5. [2]S. H. Yang and S. Y. Kung, ”A fuzzy logic based intelligent street light control system,” IEEE Trans. Ind. Electron., vol. 67, no. 2, pp. 1243- 1252, Feb. 2020. [3]A. Kumar, A. Sharma, and R. Singh, ”Machine learning approaches for traffic flow prediction and lighting control,” IEEE Trans. Intell. Transp. Syst., vol. 22, no. 8, pp. 4912-4922, Aug. 2021. [4]V. Mnih et al., ”Human-level control through deep reinforcement learn- ing,” Nature, vol. 518, no. 7540, pp. 529-533, Feb. 2015. [5]T. T. Zin, P. T. Htun, and Y. Y. Hlaing, ”Smart street lighting system using reinforcement learning,” in Proc. IEEE Conf. Comput. Commun. Eng., 2022, pp. 78-83. [6]S. Khemakhem and L. Krichen, ”A comprehensive survey on an IoT- based smart public street lighting system application for smart cities,” Franklin Open, vol. 7, p. 100142, 2024. [7]V. Krizanovic, A. Pejkovic, and P. Pracy, ”Smart street lighting for enhancing energy efficiency in urban environments,” in Int. Conf. Softw. Telecommun. Comput. Netw., 2024, pp. 1-6. [8]J. Lillaney and R. Jain, ”An automated smart street lighting model using integrated systems with deep learning and cloud deployment,” in World Conf. Commun. Comput., 2023, pp. 1-5. [9]B. A. Kindhi and I. S. Pratama, ”Fuzzy logic and IoT for smart city lighting maintenance management,” in 3rd East Indonesia Conf. Comput. Inf. Technol., 2021, pp. 369-373. [10]A. Martinovic et al., ”Embedded implementation of optimal city lighting lamp control,” in 31st Telecommun. Forum, 2023, pp. 1-4. [11]M. Singh, S. Saxena, and A. R. Prusty, ”Energy efficient intelligent lighting system for smart cities,” in IEEE Conf. Interdiscip. Approaches Technol. Manag. Soc. Innov., 2022, pp. 1-6. [12]S. K. Sahoo et al., ”Enhanced public safety and well-being through IoT-enabled light pollution mitigation with reinforcement learning,” in Second Int. Conf. Intell. Cyber Phys. Syst. IoT, 2024, pp. 406-411. [13]RCWL-0516 Microwave Radar Sensor Datasheet, ”Doppler effect mo- tion detector module,” 2023. [14]TensorFlow Lite Micro, ”Machine learning for microcontrollers,” Google, 2024.