Hierarchical Multi-Leader Multi-Follower Stackelberg Game for Dynamic Computation Offloading in MEC-Enabled IoT Systems

Abstract

Mobile Edge Computing (MEC) has become one of the enabling factors in the delivery of latency-sensitive, energy-constrained Internet of Things (IoT) applications. Nevertheless, offloading computation in heterogeneous and dynamic MEC-IoT settings is still a challenge due to resource contention, dynamic workload fluctuations, and economic issues between service providers and users. This paper proposes a solution to these problems by introducing an Enhanced Hierarchical Multi-Leader Multi-Follower (MLMF) Stackelberg game model for distributed computation offloading. According to the proposed model, MEC servers are leaders that use an Iterative Stackelberg Pricing Algorithm (ISPA). In contrast, IoT devices are followers that use an Iterative Proximal Offloading Algorithm (IPOA) to select the optimal task allocation. The interaction between the IPOA and ISPA will ensure that a capacity-constrained Stackelberg equilibrium is reached. Massive simulations in MATLAB 2023b demonstrate that the proposed framework achieves 60-97% reduction in execution delay compared to baseline schemes, 30-35% energy savings with effective offloading, a fairness index of 0.95-0.97 under Jain's fairness index, and utility optimization of the system through balanced economic incentives. The ablation experiment also confirms that either of the two ablation conditions can ruin performance when either IPOA or ISPA is removed, indicating that hierarchical joint optimization is required. The proposed MLMF system offers a large-scale, cost-effective, and justice-oriented solution to the MEC-enabled IoT systems.