Abstract:
This paper introduces a comparatively new technique for Particle Swarm Optimization (P.S.O). The standard P.S.O technique is modified in a unique way to come up with B-positive Particle Swarm Optimization. The B.P.S.O which is simulated in Matlab lets the particles in a multi-dimensional space to move in an overall positive direction. In other words the particles are made to move from one side of the space to the other without negative (backward) displacement in search of the global best position. At the same time the displacement magnitude is slightly reduced randomly to discourage the particles from jumping out of the space boundary. The lost particles are randomly thrown around the then known best position, this in return saves a lot of time and effort resulting in improved overall simulation results. Five popular benchmark functions are used to evaluate the performance of B.P.S.O and the result in terms of mean and standard deviation values for global minimum and mean time per replica are compared with previous Standard P.S.O results. The B.P.S.O turns out to be more efficient in terms of optimum convergence and simulation speed.