Levy flight incorporated hybrid learning model for gravitational search algorithm

Abstract

Gravitational search algorithm (GSA) is a widely used meta-heuristic algorithm for global optimization. Its strong social interaction abilities and easy to implement nature make it more applicable than its contemporaries. However, multi-modality always remains a challenging task for GSA search mechanism due to its incapabilities towards premature convergence. This paper proposes a novel GSA variant called ‘Levy flight incorporated gravitational search algorithm with an adaptive spiral strategy (LevyGSA)’ to address the shortcomings of GSA with the following developments: First, a levy flight associated position update strategy for elite agents of the swarm is proposed for a better interior search. Secondly, an adaptive spiral update strategy is introduced for the rest swarm to balance the trade-off between exploration and exploitation for a robust search. Finally, a dimensional reduction based strategy for enhancing the local search around the known global optimal region is introduced. The proposed algorithm is tested over 23 classical test problems and 30 CEC 2014 test problems. The numerical results demonstrate the outstanding performance of the proposed algorithm through which it outperforms the well-known existing meta-heuristics along with recent GSA variants. Furthermore, finding more accurate solutions for five engineering design problems validates its applicability in real-world scenarios.