A Clustering Approach Based on Artificial Neural Networks to Solve Routing Problems

Artificial neural networks (ANN) are rising as a promising heuristic approach to solve combinatorial optimization problems. Those based on self-organizing maps have presented themselves as the most promising ones because their results are very competitive, despite their apparent drawback of being totally unsupervised. This paper presents a two-phase algorithm based on artificial neural networks to solve three routing problems: traveling salesman, multiple traveling salesmen and capacitated vehicle routing. Several tests for the three problems are conducted on standard data and the results are compared with other proposals based on ANN from the literature, suggesting that the proposed approach is competitive.