Object-Guided Ant Colony Optimization Algorithm for Quadratic Assignment Problem

Ant colony algorithm is a swarm intelligence optimization algorithm. Recently, it has been used in many optimization problems. Classical ant colony optimization algorithms emphasize on the study of swarm's intelligence through the use of pheromone, but seldom have been done on the improvement of ant's individual intelligence. This paper presents an object-guided ant colony optimization algorithm. In the process of solution construction, each ant always keeps a complete solution. Each time ant probabilistically selects a solution component, it will calculate the difference between current solution and the new solution after accepting the selected component, and only those, which can improve the solution, will be accepted. Otherwise, the original corresponding component is retained. In order to keep algorithm from premature stagnation, classical construction strategy is used to keep the swarm diversity. The simulation results on benchmark instances from QAPLIB show that the object-guided strategy can improve the performance of ant colony optimization algorithm significantly.