History-driven firefly algorithm for optimisation in dynamic and uncertain environments

Due to dynamic and uncertain nature of many optimisation problems in real-world, the applied algorithm in this environment must be able to continuously track the changing optima over time. In this paper, we report a novel speciation-based firefly algorithm for dynamic optimisation, which improved its performance by employing prior landscape historical information. The proposed algorithm, namely history-driven speciation-based firefly algorithm (HdSFA), uses a binary space partitioning (BSP) tree to capture the important information about the landscape during the optimisation process. By utilising this tree, the algorithm can approximate the fitness landscape and avoid wasting the fitness evaluation for some unsuitable solutions. The proposed algorithm is evaluated on the most well-known dynamic benchmark problem, moving peaks benchmark (MPB), and also on a modified version of it, called MPB with pendulum-like motion among the environments (PMPB), and its performance is compared with that of several state-of-the-art algorithms in the literature. The experimental results and statistical test prove that HdSFA outperforms most of the algorithms in different scenarios.