Data-driven robust optimization for the itinerary planning via large-scale GPS data

Abstract

Listen

Translate article

In this paper, we propose a data-driven robust optimization for establishing reliable itineraries through the use of GPS trajectories. The goal of the study is to provide a robust solution that is able to maximize the probability of achieving the expected travel time and minimize the delay. The designed framework can be viewed as an incremental approach, where data-driven robust optimization cooperates with a learning procedure such that both the uncertainty set and the objective function are incrementally adjusted according to the current data analysis results. In fact, two types of training models are designed in order to adapt the robust optimization model through analyzing GPS-data. The first training model tries to generate the uncertainty set for establishing the model, and the second one establishes the best parameter-settings allowing to converge towards a robust solution. Finally, a data-based simulation framework is designed for analyzing the robustness of the proposed method, where achieved solutions are tested on a simulated traffic network by using real-world orders as the comparison targets.