Efficient Computation Offloading of IoT-Based Workflows Using Discrete Teaching Learning-Based Optimization

Abstract

As the Internet of Things (IoT) and mobile devices have rapidly proliferated, their computationally intensive applications have developed into complex, concurrent IoT-based workflows involving multiple interdependent tasks. By exploiting its low latency and high bandwidth, mobile edge computing (MEC) has emerged to achieve the high-performance computation offloading of these applications to satisfy the quality-of-service requirements of workflows and devices. In this study, we propose an offloading strategy for IoT-based workflows in a high-performance MEC environment. The proposed task-based offloading strategy consists of an optimization problem that includes task dependency, communication costs, workflow constraints, device energy consumption, and the heterogeneous characteristics of the edge environment. In addition, the optimal placement of workflow tasks is optimized using a discrete teaching learning-based optimization (DTLBO) metaheuristic. Extensive experimental evaluations demonstrate that the proposed offloading strategy is effective at minimizing the energy consumption of mobile devices and reducing the execution times of workflows compared to offloading strategies using different metaheuristics, including particle swarm optimization and ant colony optimization.