Divided opposition strategy in particle swarm framework for constrained optimization problem

In nature inspired algorithms population initialization techniques play an important role to find an optimal solution. In this study we proposed a novel population initialization technique Divided opposition-based learning Particle Swarm Optimization (D-PSO). This technique is inspired by Opposition Based Learning (OBL). D-PSO is a technique in which elements of initial population are uniformly cover the search space so the possibility of obtaining the optimal solution is highest. To validate the results D-PSO is tested on 16 benchmark functions for dimensions 10 and 30 and 12 CEC22 functions along with standard PSO, OBL-PSO, I-PSO. In standard PSO elements of initial population is randomly generated and in OBL-PSO elements of initial population are generated using OBL technique. I-PSO generate initial population elements using improved OBL technique. D-PSO gives better outcomes for all benchmark functions for dimension 10, 30 and 10 CEC22 function out of 12 as compared to other initialization techniques. To measure the significance of results a statistical analysis is also done in this study. Complexity analysis and convergence analysis is also measured for both set of benchmark functions. The convergence behavior of D-PSO for all benchmark function for dimension 10, 30 and 10 CEC22 function is best as compared to other initialization technique.