Simulation-based multi-objective resource optimization in fog computing using Genetic Algorithms

Efficient task scheduling in heterogeneous fog computing environments remains a challenging problem because latency, energy consumption, and resource utilization are inherently conflicting optimization objectives. Conventional scheduling techniques often rely on static heuristics or single-objective optimization, limiting their ability to adapt to dynamic and resource-constrained fog infrastructures. This study presents a multi-objective resource optimization framework based on a constraint-aware Genetic Algorithm (GA) to achieve balanced task allocation while satisfying practical scheduling constraints. The proposed framework jointly minimizes task latency and energy consumption while maximizing resource utilization through an adaptive evolutionary search strategy. The framework was implemented and evaluated using a Python-based discrete-event simulation framework under diverse workload conditions representative of heterogeneous fog computing environments. Comprehensive experiments, including comparative performance evaluation, paired statistical significance testing, weight sensitivity analysis, scalability analysis, and convergence analysis, show that the proposed framework achieves latency statistically comparable to NSGA-II while requiring substantially lower optimization overhead. The evaluation also shows that the framework maintains feasible scheduling under the tested configurations, with cloud fallback explicitly represented as an execution destination and no task drops observed in the evaluated runs. The results further demonstrate stable convergence behavior and robust performance across varying workload sizes and optimization settings. Statistical testing further shows that the proposed GA is not significantly different from NSGA-II in scheduling quality, so its principal validated advantage is matching NSGA-II’s performance at markedly lower computational cost rather than outperforming it (Section “Results and discussion”). Overall, the proposed constraint-aware Genetic Algorithm provides a computationally efficient scheduling framework for the simulated heterogeneous fog environments evaluated in this study, with potential applicability to latency-sensitive domains subject to validation using real fog testbeds and workload traces.