FuzzyNET: A Fuzzy Logic-Based Framework for Enhancing Urban Efficiency in Smart Cities

The smart city paradigm emphasizes the strategic use of IoT, Big Data, and digital technologies to optimize urban operations and improve service delivery for citizens. Embedding digital technologies within operational workflows is driven by the objective of improving efficiency and long-term sustainability. The core transformation is Artificial Intelligence (AI) that revolutionizes management of cities through automation and instantaneous decisions based on data. Artificial intelligence can significantly reshape essential areas such as power infrastructure, waste services, and public safety. For urban mobility, deep learning has been used for advanced image classification tasks. AI-based traffic signal recognition also has the potential to facilitate optimal vehicle flow, resulting in less fuel consumption and less carbon emissions. Flexible traffic signals that respond to real-time traffic density can help cities significantly decrease idling and air pollution. To address these challenges, this paper proposes a novel two-phase framework: First, a hybrid CNN–Fuzzy model for robust traffic sign classification, and second, a dynamic traffic signal optimization system utilizing YOLOv8 for real-time vehicle density estimation. In the first phase, a hybrid CNN–Fuzzy is proposed in order to improve classification performance using the feature extraction capability of Convolution Neural Networks (CNNs) and fuzzy decision support. Strong model generalization was ensured. The overfitting was mitigated by rigorous preprocessing with data augmentation, normalization and some moderate dropout regularization applied at all levels. Quantitatively, the proposed hybrid method demonstrates clear superiority over recent state-of-the-art models; it achieved an impressive validation accuracy of 99.72% and a minimal validation loss of 0.0126, significantly outperforming the 96.12% validation accuracy and 10.95 validation loss reported by the baseline model. These performance criteria further support the effectiveness of fuzzy-aided deep learning in identifying such a crucial decision-making process. Lastly, the work offers an approach that is scalable to energy-efficient traffic control as a component of smart urban mobility in conjunction with environmental sustainability to lead towards CO2 reduction.