A hierarchical fog computing framework for efficient and scalable task offloading in IoT systems

The rapid evolution and growing complexity of the modern automotive sector along with increasing interconnectivity among vehicular systems have sparked significant academic and industrial interest. Emerging intelligent transportation systems demand computational architectures capable of achieving ultra-low latency and high processing efficiency. However, these performance requirements are constrained by high energy consumption and system wide power limitations. To mitigate these challenges, intelligent workload distribution specifically adaptive task allocation and optimal utilization of decentralized computing resources offers a scalable and energy efficient solution. This study proposes an advanced framework for resource provisioning and task offloading embedded within a hierarchical tri-tier vehicular fog-cloud infrastructure. The actual contribution of this work is the integration of a mobility-aware hierarchical offloading policy with a constrained resource-aware MPSO specifically designed for vehicular fog environments. The proposed system supports reliable, real-time task processing and dynamic resource allocation tailored for latency-critical applications in fast changing vehicular environments. To enhance the efficiency of the fog computing layer, a meta-heuristic optimization algorithm is introduced to maximize resource usage and reduce end-to-end communication delays. After several experimentation, the obtained results confirm that the proposed MPSO reduces average latency upto 23.8%, energy consumption, execution time and convergence iterations upto 22.1%, 32.4% and 34.9% compared with conventional PSO and other meta-heuristic baselines. It also improves resource utilization and task success rate upto 9.7% and 4.5%. Overall, these results confirm that MPSO provides faster convergence, lower delay, reduced energy usage and more efficient fog computing resource utilization for large-scale vehicular fog-cloud computing.