PhD Defence: Vehicle Routing with Uncertain Demand


Fixed schedules for the delivery of goods make planning easier for distribution centres, but can lead to delays in deliveries and erratic delivery times. In his dissertation entitled, <link doctoral-programme phd-in-management phd-projects detail>Vehicle Routing with Uncertain Demand, Spliet explores various methods of incorporating uncertainty and demand fluctuation variables in vehicle routing models.

Having experienced the inconvenience of irregular delays at his own part-time job, Remy decided to apply his PhD research to the solution of the uncertainty issues related to delivery scheduling. His research tackles several scheduling problems including the cost of rerouting vehicles, the difficulty of planning time windows for uncertain demand quantities and the difficulty of assigning drivers to customers before demand is known.

By applying state-of-the-art vehicle routing mathematical programming, Spliet demonstrates how organisations can substantially decrease transportation costs. He argues that not only direct transport costs should be taken into account, though. Indirect transportation effects, such as the cost to customers of late deliveries should be factored in.

<link people remy-spliet>Remy Spliet defended his dissertation on 18 October 2013. His supervisor was <link people rommert-dekker>Professor Rommert Dekker and his co-supervisor was <link people adriana-gabor>Dr. Adriana Gabor. Other members of the Doctoral Committee were Professor Guy Desaulniers (Polythechnique Montréal), and <link people dennis-huisman>Dr. Dennis Huisman and <link people albert-wagelmans>Professor Albert Wagelmans (both of Erasmus University Rotterdam).

About Remy Spliet

Remy Spliet (Netherlands, 1986) received his bachelor's degree in Econometrie & Besliskunde (Econometrics and Management Science) in 2007 from the Erasmus University in Rotterdam. In 2008, he received his master's degree in Econometrics & Management Science, the Operations Research and Quantitative Logistics programme, equally from the Erasmus University in Rotterdam with distinction. Following this, he started his PhD research at the Econometric Institute at the Rotterdam University in Rotterdam.

As a student, Remy worked as a non-traveling salesman at a local liquor store in his hometown IJsselstein. There, he underwent the hardships caused by deliveries arriving too late or too early. Quite often, he had to wait for a truck that arrived late, followed by working overtime to handle the late delivery. Also customers had to be disappointed when luxurious bottles of wine or whiskey were not delivered in time. Missed sales frustrated Remy, but not as much as being called in the morning, while still sleeping, to come to work because the truck had arrived early. He took it upon himself to do something about it. He devoted his studies for both his bachelor's as well as his master's thesis to planning methods to make trucks be on time. Eventually, he became a PhD student to take the design of delivery schedules to a higher level.

As a PhD student, Remy became passionate about combinatorial optimization, and in particular mathematical programming. His interests are in analysing different formulations, finding valid inequalities and accompanying separation algorithms, experimenting with different types of relaxations and designing algorithms to solve these, and of course implementing branch-and-bound procedures.

Abstract of Vehicle Routing with Uncertain Demand

In distribution networks a supplier transports goods from a distribution centre to customers by means of vehicles with limited capacity. Drivers will drive routes on which they visit multiple customers to make deliveries. Typically, deliveries are made regularly and a fixed schedule is maintained. A fixed schedule is beneficial for many operational purposes, as it for instance allows for easy planning of the packing of the vehicles at the distribution centre, or it allows the customer to roster the delivery handling personnel. A fixed schedule is often reused to make weekly deliveries for a period of a year or longer.

However, at the moment of designing a schedule, the demand of the customers is usually unknown. Moreover, in most cases, demand of a customer will be different for each delivery. Therefore, it will be necessary to construct or adapt vehicle routes for each day of delivery, without deviating too much from the fixed schedule.

In this thesis several different views on a fixed schedule are explored. It addresses the need from practice to incorporate the uncertainty of demand in transportation models to increase the efficiency of transport. Innovative vehicle routing models are presented taking uncertain or varying demand into account. New algorithms using state-of-the-art methods are presented based on these models, to construct fixed schedules and vehicle routes. The algorithms make use of recent scientific advances in mathematical programming, specifically in the domain of vehicle routing.

Photos: Chris Gorzeman / Capital Images