Many-objective firefly algorithm with two archives for computation offloading

Firefly algorithm (FA) has shown good performance over many optimization problems. However, most studies on FA focus on problems with no more than three objectives. For problems with larger number of objectives, i.e. many-objective optimization problems (MaOPs), FA encounters some difficulties in providing sufficient selection pressure and maintaining population diversity. So far, FA has rarely been used to solve MaOPs. In this paper, we try to propose a many-objective FA variant with two archives (called TaMaOFA) to challenge MaOPs. Inspired by the idea of an improved two-archive method (called Two_Arch2), TaMaOFA employs two archives including convergence archive (CA) and diversity archive (DA), which are used to strengthen convergence and diversity, respectively. By utilizing the search information of CA and DA, a convergence and a diversity biased search strategies are designed, respectively. To enhance the search efficiency, a random search model is constructed based on the two biased search strategies and random attraction. To validate the performance of TaMaOFA, the well-known MaF benchmark is utilized. Results show that TaMaOFA obtains competitive performance when compared with ten other state-of-the-art approaches. Finally, TaMaOFA is applied to computation offloading in edge computing environments. Simulation results demonstrate that TaMaOFA still achieves promising performance.