A study of particle swarm optimization particle trajectories
Information Sciences: an International Journal
Comprehensive learning particle swarm optimizer for global optimization of multimodal functions
IEEE Transactions on Evolutionary Computation
| Hi-index | 0.00 |
Since a particle swarm optimization (PSO) algorithm uses a coordinated search to find the optimum solution, it has a better chance of finding the global solution. Despite this advantage, it is also observed that some parameters used in PSO may affect the solution significantly. Following this observation, this research tries to tune some of the parameters and to add mechanisms to the PSO algorithm in order to improve its robustness in finding the global solution. The main approaches include using uniform design to ensure uniform distribution of the initial particles in the design space, adding a mutation operation to increase the diversity of particles, decreasing the maximum velocity limitation and the velocity inertia automatically to balance the local and the global search efforts, reducing velocity when constraints are violated, and using Gaussian distribution based local searches to escape local minima. Besides these efforts, an algorithm is also developed to find multiple solutions in a single run. The results show that the overall effect of these approaches can yield better results for most test problems.