Weighted-leader search: A new choice in metaheuristic and its application in real-world large-scale optimization

Abstract

Aiming at the robustness and real-time requirements of engineering applications, this paper develops a novel metaheuristic algorithm with a simple mathematical framework for numerical optimization, namely the weighted-leader search (WLS). With the WLS, a weighted mean point of the top solutions is developed to evolve the population. Additionally, several simple local search mechanisms are introduced to prevent the algorithm from premature convergence. Moreover, the WLS possesses fewer constant coefficients to reduce the parameter sensitivity. The performance of the WLS is evaluated with the CEC 2017 test suite with 10D, 30D, 50D and 100D problems, the CEC 2008 test suite with 100D, 500D and 1000D problems, large-scale benchmarks with 100D, 500D and 1000D (or beyond) problems and constrained engineering design problems. The statistical results of the experiments indicate that the WLS exhibits an outstanding ability among the state-of-the art algorithms in different competitions, which verifies the accuracy and efficiency of our algorithm in solving complex problems, as well as its robustness in large-scale evaluation. Moreover, the WLS is employed to address the cooperative path planning problem of an unmanned aircraft vehicle (UAV) swarm with 100 members. Through problem model construction, the path planning issue is transformed into a constrained optimization problem and solved by the WLS. The experimental results show that the proposed WLS is able to solve the path planning problem in real time within the constraints and significantly outperforms the other competitors, demonstrating the accuracy and time savings of the algorithm in solving real-world optimization problems.