A Novel Levy Walk-based Framework for Scheduling Power-intensive Mobile Edge Computing Tasks

Abstract Mobile edge computing (MEC) enables computationally intensive tasks to be processed at the network edge to provide low-latency services. However, inefficient task scheduling can negatively impact performance metrics like completion time and energy consumption. This paper proposes CAPL-MEC, an adaptive task scheduling framework that utilizes Levy walk modeling to address mobility patterns in MEC. The system model generates random edge nodes within defined bounds to simulate heterogeneous environments. A power consumption model is also presented to optimize dynamic and static power. Device mobility follows an adaptive Levy walk distribution where the power law exponent is time-varying. Latency and reliability (task replication) models are also defined. The CAPL-MEC algorithm utilizes an adaptive Levy walk approach to predict device locations and schedule tasks accordingly. A hybrid task allocation strategy combines proximity awareness, mobile-centric execution, and handovers between mobile and edge devices. Simulations evaluate CAPL-MEC across met-rics like completion time, energy consumption, CPU and memory utilization, and wait times under various configurations. Results demonstrate that CAPL-MEC outperforms other algorithms by minimizing completion time through efficient resource allocation based on predicted mobility patterns. Energy consumption is also reduced through power-conscious scheduling. Overall, the framework presents an effective and adaptable solution for task scheduling in dynamic MEC environments.