Improved Clustered Optimized Particle Swarm Optimization for Global Optimization in Multimodal Problems for Systems with Limited Resources

This paper proposes an improved algorithm of Particle Swarm Optimization (PSO) based on stage-structured algorithm and added features to enhance the ability of finding the global optimum in multimodal multi-dimensional optimization problems, which may have local optima, using low computational effort. Search space clustering based on Euclidian distance, particles crossover and mutation are added features to enhance the ability of exploration and exploitation of the algorithm. The algorithm divides the optimization process into three stages, namely; global search, local search and final stage. The parameters of the PSO and the added features are modified by the algorithm in each stage to achieve the stage goal. Furthermore, if the performance is not satisfactory, the same algorithm is used to optimize the parameters of the proposed algorithm. Fifteen benchmark multimodal test functions, that can be expanded to multi-dimensions, are used to test the proposed algorithm and they demonstrated its superiority to find the global optimum over other algorithms with minimal computational effort which is vital for many modern/smart limited resources devices (such as smart phones, Internet of Things devices (IoT), self-driving cars, ... etc). Keywords— particle swarm optimization (PSO); crossover; mutation; optimization; limited resources.