The rapid expansion of Internet of Things (IoT) ecosystems in smart cities generates massive heterogeneous data streams that demand energy-efficient, secure and low-latency processing. Traditional cloud-centric systems face communication overhead, workload imbalance and limited adaptability to dynamic traffic. This study aims to develop an intelligent, energy-aware and scalable load-balancing framework that ensures high performance, robust task allocation and secure data handling for next-generation smart city infrastructures.
The proposed model integrates deep learning and metaheuristic optimization to formulate an optimal load-balancing policy. Data integrity is ensured through transport layer security (TLS) and lightweight message queuing telemetry transport (MQTT) communication. Edge-level preprocessing incorporates Kalman filtering for distortion removal, Huffman compression to reduce traffic and principal component analysis (PCA) to eliminate correlated features. A dynamic resource-monitoring layer evaluates latency, bandwidth and central processing unit (CPU)/memory usage using a moving-average model. The decision-making core employs a deep convolutional pulse coupled neural network (DCPCNN) to learn adaptive load-balancing strategies, while Battle Royale optimization (BRO) schedules tasks based on energy, latency and throughput. The framework is evaluated using CityPulse and Google 2019 cluster datasets.
Experimental results demonstrate exceptional scalability, operational efficiency and balanced workload distribution. The proposed system achieves 99.12% and 98.54% success rates on CityPulse and the Google cluster dataset, respectively, outperforming existing models.
This work presents the first integrated DCPCNN-BRO-driven load-balancing framework combining secure communication, advanced preprocessing, dynamic resource monitoring and metaheuristic scheduling to enhance sustainability and performance in smart city IoT environments.
