A decentralized quantum-inspired particle swarm optimization algorithm with cellular structured population

This paper proposes a decentralized form of quantum-inspired particle swarm optimization (QPSO) with cellular structured population (called cQPSO) for keeping the population diversity and balancing the global and local search. The cQPSO is further improved by re-designing the local attractor in the sub-population (called cQPSO-lbest) in order to accelerate the diffusion of the best solution and thus enhance the performance of cQPSO. The particles in cQPSO and cQPSO-lbest are distributed in a two-dimensional (2D) grid and only allowed to interact with their neighbors according to the specified neighborhood, which plays a role in exploiting the search space inside the neighborhood. The overlapping particles work for delivering the information among the nearest neighborhoods acting as exploring the search space with diffusion of solutions during the evolutionary process. Theoretical studies are made to analyze the global convergence of cPSO and cQPSO-lbest based on the theory of probabilistic metric space. We systematically investigate the performance of cQPSO-lbest on 42 benchmark functions with different properties (including unimodal, multimodal, separated, shifted, rotated, noisy, and mis-scaled) and compare with a set of PSO variants with different topologies and swarm-based evolutionary algorithms (EAs). The experimental results demonstrate the better performance of cQPSO-lbest. Moreover, two real-world problems, which are two-dimensional (2D) IIR digital filter design and economic dispatch (ED) problem from power systems area, are used to evaluate cQPSO-lbest and the experimental results verified the advantages of cQPSO-lbest.