Konferensartikel

In this paper, inspired by the herding behavior of rhinos, a new kind of swarm-based metaheuristic search method, namely Rhino Herd (RH), is proposed for solving global continuous optimization problems. In various studies of rhinos in nature, the synoptic model is used to describe rhino’s space use and estimate its probability of occurrence within a given domain. The number of rhinos increases year by year, and this increment can be forecasted by several population size updating models. Synoptic model and a population size updating model are formalized and generalized to a general-purpose metaheuristic optimization algorithm. In RH, null model without introducing any influences is generated as the initial herding. This is followed by rhino modification via synoptic model. After that, the population size is updated by a certain population size updating model, and newly-generated rhinos are randomly initialized within the given conditions. RH is benchmarked by fifteen test problems in comparison with biogeography-based optimization (BBO) and stud genetic algorithm (SGA). The results clearly show the superiority of RH in searching for the better function values on most benchmark problems over BBO and SGA.

rhino herd, synoptic model, population size updating model, benchmark functions, swarm intelligence