Dedicated Hardware for Ant Colony Optimization Using Distributed Memory

Ant colony optimization (ACO) is based on behavior of food gathering of ants. ACO is a powerful search tool when applied to combinatorial optimization problems. However, ACO requires a lot of calculation time, because the search mechanism of ACO is based on repetitive calculations. Reducing calculation time is the most important priority in case of applying ACO to combinatorial optimization problems. In this paper we propose novel dedicated hardware for ACO in order to reduce the calculation time. The proposed hardware introduces a new memory access technique and new parallel processing, and achieves real-time processing while keeping the quality of solution in comparison with software processing. Experiments using benchmark data prove the effectiveness of the proposed hardware.