Comparative analysis of hybrid techniques for an ant colony system algorithm applied to solve a real-world transportation problem

Abstract

This work presents a comparison of hybrid techniques used to improve the Ant Colony System algorithm (ACS), which is applied to solve the well-known Vehicle Routing Problem (VRP). The Ant Colony System algorithm uses several techniques to get feasible solutions as learning, clustering and search strategies. They were tested with the dataset of Solomon to prove the performance of the Ant Colony System, solving the Vehicle Routing Problem with Time Windows and reaching an efficiency of 97% in traveled distance and 92% in used vehicles. It is presented a new focus to improve the performance of the basic ACS: learning for levels, which permits the improvement of the application of ACS solving a Routing-Scheduling-Loading Problem (RoSLoP) in a company case study. ACS was applied to optimize the delivery process of bottled products, which production and sale is the main activity of the company. RoSLoP was formulated through the well-known Vehicle Routing Problem (VRP) as a rich VRP variant, which uses a reduction method for the solution space to obtain the optimal solution. It permits the use in efficient way of computational resources, which, applied in heuristic algorithms reach an efficiency of 100% in the measurement of traveled distance and 83% in vehicles used solving real-world instances with learning for levels. This demonstrates the advantages of heuristic methods and intelligent techniques for solving optimization problems. © 2010 Springer-Verlag Berlin Heidelberg.