Skip to Main Content
Velocity threshold vmax is an important parameter of Artificial physics optimization. Different from other parameters, it affects the algorithm performance by restricting the moving size and direction of each particle. Because of the complex optimisation problems, the proper vmax setting may provide a reasonable solution within an allowed generation. However, up to now, there are only few scholars who are concerned in this problem. Therefore, in this paper, the authors investigate two selection principles of vmax (a canstant vmax and an adaptive vmax) with high dimension on numerical optimisation problems. To make a deep insight, the test suit consists of three different type benchmarks: unimodel, multi-modal functions with a few local optima and multi-modal functions with many local optima. Simulation results show an adaptive vmax may generally obtain the satisfied solution within the allowed iterations.