Ant-Based System Analysis on the Traveling Salesman Problem Under Real-World Settings

Many optimization problems have huge solution spaces, deep restrictions, highly correlated parameters, and operate with uncertain or inconsistent data. Such problems sometimes elude the usual solving methods we are familiar with, forcing us to continuously improve these methods or to even completely reconsider the solving methodologies. When decision makers need faster and better results to more difficult problems, the quality of a decision support system is crucial. To estimate the quality of a decision support system when approaching difficult problems is not easy, but is very important when designing and conducting vital industrial processes or logistic operations. This paper studies the resilience of a solving method, initially designed for the static and deterministic TSP (Traveling Salesman Problem) variant, when applied to an uncertain and dynamic TSP version. This investigation shows how a supplementary level of complexity can be successfully handled. The traditional ant-based system under investigation is infused with a technique which allows the evaluation of its performances when uncertain input data are present. Like the real ant colonies do, the system rapidly adapts to repeated environmental changes. A comparison with the performance of another heuristic optimization method is also done.