A prognostic decision-making approach under uncertainty for an electric vehicle fleet routing problem

Abstract

Electric Vehicles (EVs) have gained popularity over the past few years, given their potential to reduce the emission of Greenhouse Gases, which contribute to Climate Change. Researchers have put substantial effort into handling the limitations EVs have, for instance, battery capacity, battery management and charging infrastructure in urban areas. The Electric Vehicle Routing Problem (EVRP) represents a relevant challenge since it directly affects the supply chain of many businesses and could help transitioning to electromobility. Several methods have been used to solve the EVRP incorporating realistic elements, for example, energy consumption models, dynamic and stochastic conditions, charging stations, and others.\\ In this thesis, we designed, implemented and tested a Monte Carlo Tree Search-based algorithm to solve the EVRP on-line. This is an incipient methodology to solve the EVRP and has great potential given its Tree structure, real-time functioning, evaluation of future scenarios and search procedure. Our work shows how the proposed methodology successfully solves the problem online and can update the route given new information. Future efforts aim to add new elements for the EVRP, study the scope of this methodology in other areas where uncertain prognostic information is valuable, explore different strategies in the EVs fleet case, and others.