Machine Learning Guided Optimization for Demand Responsive Transport Systems

Abstract

Most of the time, objective functions used for solving static combinatorial optimization problems cannot deal efficiently with their real-time counterparts. It is notably the case of Shared Mobility Systems where the dispatching framework must adapt itself dynamically to the demand. More precisely, in the context of Demand Responsive Transport (DRT) services, various objective functions have been proposed in the literature to optimize the vehicles routes. However, these objective functions are limited in practice because they discard the dynamic evolution of the demand. To overcome such a limitation, we propose a Machine Learning Guided Optimization methodology to build a new objective function based on simulations and historical data. This way, we are able to take the demand’s dynamic evolution into account. We also present how to design the main components of the proposed framework to fit a DRT application: data generation and evaluation, training process and model optimization. We show the efficiency of our proposed methodology on real-world instances, obtained in a collaboration with Padam Mobility, an international company developing Shared Mobility Systems.