Integrating forecasting in metaheuristic methods to solve dynamic routing problems: evidence from the logistic processes of tuna vessels

Abstract

The multiple Traveling Salesman Problem (mTSP) is a widespread phenomenon in real-life scenarios, and in fact it has been addressed from multiple perspectives in recent decades. However, mTSP in dynamic circumstances entails a greater complexity that recent approaches are still trying to grasp. Beyond time windows, capacity and other parameters that characterize the dynamics of each scenario, moving targets is one of the underdeveloped issues in the field of mTSP. The approach of this paper harnesses a simple prediction method to prove that integrating forecasting within a metaheuristic evolutionary-based method, such as genetic algorithms, can yield better results in a dynamic scenario than their simple non-predictive version. Real data is used from the retrieval of Fish Aggregating Devices (FADs) by tuna vessels in the Indian Ocean. Based on historical data registered by the GPS system of the buoys attached to the devices, their trajectory is firstly forecast to feed subsequently the functioning of a genetic algorithm that searches for the optimal route of tuna vessels in terms of total distance traveled. Thus, although valid for static cases and for the Vehicle Routing Problem (VRP), the main contribution of this method over existing literature lies in its application as a global search method to solve the multiple TSP with moving targets in many dynamic real-life optimization problems.