Tél. +33 5 64 27 04 59 | Mon. to Fri. 9am-5pm (local time)

Team Mapotempo: Gwénaël’s portrait, Operations Research PhD student at Mapotempo

Team Mapotempo: Gwénaël’s portrait, Operations Research PhD student at Mapotempo

Gwénaël-rault-mapotempo

The portrait of Gwénaël, Operations Research PhD student at Mapotempo in Bordeaux (R&D department):

“The experiences and variety of profiles bring a great richness to the exchanges. This allows us to compare our ideas as a developer with the field reality as well as the business process. In my opinion, the real added value of Mapotempo is to use field knowledge as an engine of innovation.”

Can you introduce yourself in a few words?

“I arrived in Mapotempo in November 2015 after graduating from my Master’s degree in Computer Science obtained in Nantes, where I studied Optimization in Operational Research. This specialization aims to formalize problems and define decision support methods to reach the best result. A job ad had been made in April on the master’s mailing list. Wishing to go to Bordeaux for personal reasons, I had carefully kept this job ad. Indeed, back then, I was doing my final internship in Cologne at Transalliance, mainly in the management of input and output flows for Smurfit Kappa, on a storage platform; before redistribution to production sites in the Benelux-North Rhine-Westphalia sector. This experience allowed me to compare university theory with field exploitation. As my internship ended at the end of June, I logically contacted Frédéric and Mehdi to find out if the position was still vacant.”

Why did you decide to join the Mapotempo adventure?

“In addition to the location in Bordeaux, I am interested in the fields of logistics and transport. Sending a product to a customer is not impressive as such, it offers little alternative and is simply a matter of moving from point A to point B. However, by adding thousands of customers, products and platforms, things get more difficult, and that’s where we can question the sharing and efficiency of travel.
On the other hand, joining a human-sized structure is stimulating in many ways. Everything is to be built, when I arrived I was the only representative of my specialty, although a few tools have been implemented, we have been able to increase our skills over the past three years and attack markets that were not in the initial spectrum of the company’s targets.”

Can you tell us about the Operations Research team?

“We are currently 3 in the Operational Research team, Adeline Fonseca, Mickaël Gaury and myself. Our mission is to provide an algorithmic response to the transport problems encountered by our customers. “How do I visit 100 customers with my fleet of vehicles? “, is the type of questions we answer with various constraints: working hours, visiting hours, quantities to be transported, driver skills, specific vehicle equipment… For this, we have some tools at our disposal to deal with the problems submitted to our optimization API, first, the time and distance matrices are obtained through our OSRM instances based on OpenStreetMap data. The data are enriched with information such as times and distances between each of the points. Then we call, depending on the case, different solvers that allow us to model the problems and “optimize” them.
We currently have several projects underway. Mickaël upgrades our test architecture and sets up our algorithms in particular to meet the specific needs of urban routing. Adeline brings our API up to speed on scheduling issues related to vehicle route optimization, for example to meet the expectations of sales representatives who must make regular visits to their customers.
For my part, I started in April a PhD on the combination of clustering and partitioning methods with resolution methods for vehicle rouring problems.”

Can you tell us about your thesis as part of your experience at Mapotempo?

“Optimization methods make it possible to respond to complex problems with many constraints but a limited volume of data. Clustering methods allow large volumes of data to be addressed and partitioned, but the control you can have over them is limited. Urban logistics is organized by dividing it into distinct sectors that are assigned to one or more field agents. It is therefore possible to use this same method automatically. The idea is to carry out a first high-level breakdown and then to carry out an optimization on each of the sectors obtained. A second step is to assemble these different subroutes in order to obtain geographically coherent routes, all the points of the same area will be carried out by the same agent, and correct in terms of compliance with constraints. Once this first base has been established, an evolution of the clusters according to the results obtained can be envisaged as well as improvements in the allocation of points to the vehicles. The current state, allows me to perform a complete resolution, from the reading of the problem, to the cutting up to the optimization of the subroutes and the global optimization that results from it; With then all the mechanics of cluster exchange between the tours thus built. Moreover, the use of clustering makes it possible to avoid calculating the entire time or distance matrix of the road network and to use the distance as the crow flies, which has the advantage of reducing calculation times. However, some topographical constraints prevent the association of points (highways, rivers, etc.), obstacles that the distance as the crow flies cannot identify. So for the moment I have set up a procedure to extract these obstacles from OpenStreetMap data and define areas where it is consistent and interesting to cluster. This makes it possible to obtain a first topological breakdown, then a second by clustering and partitioning methods.
The ultimate objective is to increase the limit of points managed by optimization from about 1000 points to 30,000, which was the request of one of our clients when this thesis project was founded. In addition, the idea is to carry out this project in order to manage multi-modal routes in urban areas, carry out a main route with a light vehicle, then deliver end customers from an intermediate stop point either on foot or by bicycle for example.”

Did you have a significant experience during your adventure at Mapotempo?

“The Mapotempo adventure as a whole is very precious to me. Already through the history of each of the profiles that make up the company, in particular Mehdi, the logistician, and Frédéric, the enthusiast of free mapping. The experiences and variety of profiles bring a great richness to the exchanges. This allows us to compare our ideas as a developer with the field reality as well as the business process. In my opinion, the real added value of Mapotempo is to use field knowledge as an engine of innovation.”

After three years in Mapotempo, how do you see the technologies and the progress you have made today?

“In 3 years, we have been able to move from a single module, Mapotempo Web, to several independent projects for Mapotempo Web, the optimization brick was initially part of it. So we extracted what allowed us to call the solvers and made it a full-fledged API. Initially we only managed the optimization of one route at a time, without time constraints. Today we manage optimizations with multiple vehicles, multiple strict or informal time constraints, capacities (weight, volume, number of visits… etc.) as well as restrictions on the use of certain vehicles for certain deliveries. We can also manage the links between the visits, a particular order, the obligation to have two crews simultaneously at the same point, a difference of several days between two visits. We broaden the spectrum of possibilities that we can manage in accordance with our customers’ needs, while continuing to gradually improve those we already manage. Using Ruby in the API gives us a flexibility and responsiveness that is very pleasant before moving from problems to more specific languages like C++ for calling solvers. The modular architecture of the model defined for vehicle rouring problems allows us to extend it simply to cover a very wide range of constraints.”

0 Comments

Leave a reply

Your email address will not be published. Required fields are marked *

*

shares