PhD Position F/M Integration of forecasting methods into the optimization models: an application to city logistics - Villeneuve-d'Ascq, France - INRIA

Description

Contexte et atouts du poste

A central issue in city logistics is to design systems that move goods to, from, and within

urban areas while meeting sustainability goals. Such city logistics systems are generally based

on new business models, cooperation among stakeholders, resource sharing, consolidation,

synchronization of operations, multi and intermodality. Here, we consider an orchestrator

that manages a system involving freight transporters which can be carriers or logistics service

providers. One of the critical activities of the orchestrator in coordinating and managing

the resources offered by the freight transporters is to distribute the transportation demand

among them. However, this task is complicated by the fact that transportation demand is

uncertain.

To help the orchestrator in his decision process, we address the associated planning problem:

the Allocation Resource Problem in city logistics with Demand Uncertainty (ARPDU). The

ARPDU is an operational problem that typically has to be solved the day before the resources

are deployed. Given the solution of the ARPDU, the orchestrator can inform the freight

providers. They can then plan their activities for the next day by including the requests of

the orchestrator in the set of the other logistics tasks they have to perform.

The ARPDU aims to determine what logistics facilities should be used and when and where

the vehicles of the carriers should be assigned to cover the demand over the planning period

in the most efficient way. More precisely, the ARPDU aims to select the facilities (crossdock

platform, storage area, parking spaces...) to be used, the types and the numbers of

vehicles (vans, cargo-bicycles,...), to determine what are their starting points and during

which periods they are used. Additional operational constraints can be considered according

to the products delivered and to the logistics means.

A key feature of the ARPDU is that demand and its characteristics (quantity, origin, destination,

time slots when the freight becomes available or may be delivered) is uncertain, i.e.,

unknown or partially known. In practice, the orchestrator must estimate the demand on the

urban area for the considered time horizon using historical data. This demand forecast is

based on a model of the urban territory that must be built first.

This thesis is performed in the context of the ANR project Adele.

Principales activités

This thesis is part of the general field of decisions-focused predictions. It will address the

scientific challenge related to the integration of forecasting methods into the optimization

models and solution methods. This implies determining useful information in available

data, developing some ad-hoc forecasting methods, managing their integration into decision

models, and developing innovative optimization algorithms in which the selection of the best

demand estimators is part of the decision process.

The main steps of the thesis will be:

. Demand modeling and design of forecasting methods;

. Mathematical modeling;

. Development (design and implementation) of innovative ad-hoc optimization algorithms

based on mathematical modeling;

Compétences

Technical skills and level required :

Good knowledge in combinatorial opimization, Stochastic optimization, machine learning

Coding: C++, Java

Avantages