Performance Improvement of Smell Agent Optimization Algorithm Using Chaotic Map

Metaheuristic algorithms have become dominant in solving different kinds of optimization problems due to their simplicity, adaptability and derivative-free approach. The Smell Agent Optimization (SAO) algorithm is a recent metaheuristic algorithm that is inspired by the concept of smell perception. The algorithm operates in three modes known as sniffing, trailing and random mode. The sniffing mode was modelled based on how an agent perceives the smell molecules. The trailing mode was modelled based on how an agent trails the smell molecule to identify its source. The random mode is a strategy employed by the algorithm to escape the state of confusion known as the local minimum. The SAO just like other metaheuristic algorithms has the problem of local minima, imbalance between exploration and exploitation and slow convergence as a result of the different modes involved. Chaotic maps have been shown to improve the performance of metaheuristic algorithms. The sinusoidal, logistic and singer maps were introduced in each of the modes of SAO to form a new algorithm known as chaotic smell agent optimization (cSAO). This modification was to improve its general performance and convergence of the original SAO. The cSAO was tested on seventeen benchmark functions and the results obtained were compared with SAO and PSO. The statistical result showed that cSAO and SAO obtained the best solution in 12 functions and PSO in 10 functions but cSAO is ranked higher than SAO and PSO with final rank values of 1.33, 1.66 and 1.86 respectively. The cSAO also converges faster than SAO by 25% but fails with PSO due to the number of function evaluations and high exploitation rate of PSO.