Opposition-based initialization and a modified pattern for Inertia Weight (IW) in PSO

Particle Swarm Optimization (PSO) is an evolutionary computing algorithm and is successfully used to solve complex real world optimization problems. Due to the complex nature of optimization problems, PSO endures the problems like premature convergence or being trapped in local minima, to avoid such situation the role of swarm initialization is very important. In this research we propose a new method to initialize the swarm particles on the basis of Generalized Opposition-based Learning (GOBL). The aim for GOBL strategy is to have an initial swarm with already fittest particles to set a solid ground for the rest of PSO algorithm to execute. Moreover, a strategy for linearly decreasing Inertia Weight has been proposed to equalize the proportions of exploration as well as exploitation capabilities of particles during the search process. The motivation behind incorporating the changes in standard PSO is to evade the earlier convergence and to help the algorithm in escaping from being trapped in local minimum. To assess the performance of proposed PSO variant, we practiced this algorithm on 8 different benchmark functions and results were compared with 4 other PSO versions found in literature. From the results analysis it is apparent that projected changes in the PSO increases its overall performance and efficiency especially when dealing with the noisy optimization problems. Also the proposed algorithm performs better and is more robust as compared to other algorithms for achieving desired results.