Multi-resolution Ant Colony - A New Approach to Use Swarm Intelligence in Continuous Problems

Since introduced bio-inspired algorithms have proven be the superior algorithm in many optimization problems. Near optimum performance of these algorithms along with their very simple calculation rules has resulted in widespread utilization of biological computing in many fields of science like communications, robotics, software engineering, networks, etc. Swarm intelligence observed in social insects such as ants, resulting in a behavior beyond the scope of individual members of society, has been an important source of inspiration for these algorithms. Ant Colony Optimization, a particularly successful research direction in swarm intelligence based algorithms dedicated to discrete optimization problems, has been effectively used in a variety of combinatorial problems such as quadratic assignments, traveling salesman problems, and routing in telecommunication networks. In this paper we propose a new algorithm named Multi Resolution Ant Colony, which uses swarm intelligence in a continuous (non-discrete) environment, for finding extremes, i. e. minimums and maximums, of a function. Genetic Algorithm (GA), up to now, has been a candidate for solving these kinds of problems when analytical methods failed. The dependency of analytical methods on initial condition is one of the most important reasons to utilize Genetic Algorithms instead. The method we present has advances over GA in that it can either, find the extremes itself (simulation results show the out-performance of our algorithm to GA in this area), or be fed to analytical methods (in low resolutions steps, not possible in GA).