A Discrete Grey Wolf Optimizer Metaheuristic for Task Offloading in Multi-Server MEC with Batteryless Devices

The maturation of energy harvesting technologies enables the integration of batteryless devices in advanced computing paradigms. For example, in mobile edge computing (MEC), batteryless mobile devices can charge when they have insufficient energy to perform task offloading, thereby relaxing the energy constraints in developing offloading strategies. This paper studies the problem of minimizing the latency of task execution in an MEC system with multiple resource-limited servers and multiple batteryless devices under intermittent operation conditions. We formulate this problem as an integer program-based optimization model and propose a discrete grey wolf optimizer (DGWO) algorithm to solve the formulated problem. DGWO uses a task sequence, which is a permutation of all tasks to be offloaded, to represent an offloading solution and introduces a discrete representation of grey wolves to link each grey wolf with a solution. For each discrete grey wolf, we design an effective task allocation strategy to designate the computing resources of MEC servers for each offloaded task. We further define a set of discrete operations upon the discrete representation to update the positions of grey wolves, for the purpose of enhancing DGWO’s global search capability. Experimental results demonstrate that DGWO outperforms other baseline metaheuristics with reduced task execution latency and improved computational efficiency.