Adaptive AI-Driven Dynamic Lighting Control System for Smart Streets and Highways for Balancing Safety and Energy Efficiency

Abstract

Efficient and adaptive lighting systems are crucial for ensuring safety, energy conservation, and sustainability in smart cities. The rising demand for energy-efficient and adaptive street and highway lighting systems has prompted the investigation of intelligent control mechanisms. Conventional and semi-automated systems frequently exhibit elevated energy consumption and restricted adaptability to variable conditions, including traffic density, weather, and pedestrian activity. This research introduces an innovative Advanced Hybrid LSTM-Dense Neural Network (AHL-DNN), which employs a dual-branch architecture to overcome existing limitations. The model attains a classification accuracy of 99.72%, exceeding baseline models such as the Unified Dense Neural Network (UDNN) at 99.26% and conventional architectures like CNNs at 97.31%. Dynamically optimizing lighting levels enhances energy efficiency, resulting in estimated energy savings of up to 38-42% compared to static systems. Scalability is achieved via compatibility with edge AI and federated learning, facilitating real-time adaptability and maintaining data privacy. The AHL-DNN model exhibits high precision (99.72%), recall (100%), and F1-score (99.7%), establishing it as an effective solution for intelligent lighting control in smart cities.