Multi-objective Sparrow Search Algorithm Based on Knee Points for Task Offloading in Multi-Cloudlet

Computation offloading is a promising technique to improve the quality of service for mobile users in mobile edge computing. However, efficient computation offloading is an NP- hard problem when the mobile user is in a multi-cloudlet scenario. To solve this problem, we construct a multi-objective optimization model which minimizes the delay and cost of all tasks under the constraints of user's requirements. For this purpose, a multi-objective intelligent algorithm, referred as Multi-objective Sparrow Search Algorithm (MSSA) is proposed, which derived from single-objective Sparrow Search Algorithm (SSA). First, we design an adaptive strategy to identify keen points and a ranking rule based on keen points to facilitate the evolution of the population. Second, to address the problem that SSA is prone to fall into local solutions, we design a scaling factor to adjust the population adaptively and propose a formula with better searching capability. Meanwhile, Gaussian mutation is added to explore more precisely in later stage. Then, an elite strategy is introduced to retain high-quality solutions. Finally, experimental results show that our proposed algorithm outperforms NSGA2 and SPEA2 in diversity, convergence and stability. In addition, our algorithm achieves better tradeoff between delay and cost compared to the benchmark experiments.