Boundary Conditions for Particle Swarm Optimization Algorithm

In particle swarm optimization (PSO), some particles may fly outside the allowable solution space when searching the global best. In order to overcome the problem, a new group of restricted boundary conditions, which relocate the arrant particles randomly in the solution space, are proposed. Moreover a new hybrid unrestricted boundary condition named invisible/absorbing is developed by introducing the favorable characteristic of the absorbing boundary condition into the existing invisible boundary condition. The performances of the five new proposed boundary conditions and six existed boundary conditions are tested based on two benchmark functions. Simulation results are examined from both the global best and convergence rate of the algorithm. Comparisons show that the performances of the new proposed boundary conditions are better than these of the existing boundary conditions, especially the invisible/absorbing boundary condition.