Journal Paper

Metaheuristic algorithms, including evolutionary approaches, are vital for solving non-trivial and non-convex optimization problems. However, real-world engineering often involves high-dimensional, expensive problems that deteriorate performance due to the substantial amount of required fitness evaluations. To address this, a growing trend utilizes evolutionary algorithms assisted by surrogate models, which limit the computational burden by providing alternatives to expensive evaluations. Leveraging the exploration capabilities of the recently developed Slime Mould Algorithm—a metaheuristic with only one tuning parameter that ignores personal best information—this work develops its surrogate-assisted counterpart: the Surrogate-Assisted Slime Mould Algorithm (SASMA). This new approach features an original database management strategy and surrogate building mechanism. To confirm its effectiveness and versatility, SASMA is tested on benchmark mathematical functions for 30 and 100 dimensions, as well as a classical truss design problem, against several surrogate-assisted and metaheuristic algorithms. The proposed SASMA achieved statistically significant improvements in both case studies, outperforming the selected benchmark algorithms on most test functions.