Memetic Particle Swarm for Continuous Unconstrained and Constrained Optimization Problems

Particle Swarm Optimization (PSO) is known for its effective and efficient global search and is one of the most effective Swarm Intelligence (SI) methods. PSO however fails to guarantee convergence to even locally optimal solution and so the method of switching to an effective local search at a safe point in the search is investigated with in-house General-Purpose PSO (GP-PSO). Combining the two algorithms results in guaranteed locally optimal convergence. Relations between various convergence criteria are investigated and methods derived to successfully switch, to the local search. Furthermore, user control is given with the derived method of switching, utilising the choice between accuracy and computational expense. With the added local search, this offers to extend the capabilities of the GP-PSO to competitive results with those in comparison in the literature.