Program > Plenary talks

Plenary Speakers 

Stefan Ropke

10 years of Adaptive Large Neighborhood Search (ALNS)
Dr. Stefan Ropke
Technical University of Denmark
Monday, June 6 - 9:45am
Room: Kastler

It is a decade ago since the first paper describing the Adaptive Large Neighborhood Search (ALNS) was published. Since then the heuristic has been applied to a multitude of vehicle routing problems and it has often been shown to produce competitive results. A potential reason for the popularity of the heuristic is that it is easy to incorporate new constraints and/or changes to the objective function and that it requires relatively little tuning in order for the heuristic to produce satisfactory results. These properties make the heuristic especially well-suited for handling real-life problems.

The ALNS is based on the metaheuristics Large Neighborhood Search and Ruin-and-Recreate. These heuristics move from solution to solution by repeatedly destroying part of the solution and afterwards repairing the solution again. The Adaptive Large Neighborhood Search (ALNS) heuristic extends the aforementioned heuristics by utilizing a portfolio of algorithms for both destroying and repairing a solution. The heuristic keeps tracks of the impact of the destroy/repair methods and favor methods that has been successful in the previous iterations. The reasoning behind this is to let the heuristic adapt to the instance at hand and to the current state of the search.

The talk will review the basic ideas behind the heuristic as well as the origins of it. We will look at the development that has taken place since the first papers describing the heuristic, including applications of the heuristic. From this we will derive information about the most important components of the heuristic and provide rules of thumb for key decisions encountered when implementing the heuristic for a new problem. Finally we will attempt to outline potential future research topics related to the adaptive large neighborhood search heuristic.



Recent advances in service network design
Dr. Michael Hewitt
Loyola University Chicago (USA)
Tuesday, June 7 - 9:15am
Room: Kastler


Consolidation carriers transport customer shipments that are small relative to container capacity and have enabled, amongst other things, the transformative effects of eCommerce. They typically participate in one of two industries: (1) less-than-truckload (LTL) freight, a roughly $30 billion industry in the United States, and, (2) small package/parcel, a much larger industry with one player alone (UPS) reporting $54 billion in revenue in 2012. Both LTL and small package carriers play a prominent role in the fulfillment of orders placed online, in brick-and-mortar stores, and through other channels. Fast shipping times and low costs are critical to the success of retailers that compete in a global marketplace; a survey by Pitney-Bowes reported that 49% of shoppers abandoned a purchase due to shipping costs. For a consolidation carrier to deliver goods in a cost-effective manner they must consolidate shipments, which in turn requires planning paths for different shipments that coordinate in both space and time. The processes that plan these paths have long been assisted by solving the Service Network Design problem, which prescribes the choice of paths for shipments and the services or resources necessary to execute them.

Advances in computational power have enabled researches to develop new, more complex, service network design models. Some of these new models seek to more accurately represent the operational landscape of a consolidation carrier. For example, while the earliest service network design models did not consider the time dimension at all, current research efforts are modeling time to the hour. Similarly, the initial service network design models assumed model parameter values were known with certainty a priori. Yet now much research is being done to solve models that recognize uncertainty in various parameter values (particularly demands and capacities).

Other advances seek to extend the scope of decisions prescribed by service network design models. For example, researchers are working on service network design models that can also inform strategic decisions such as fleet/resource acquisition and allocation. Other models also negotiate pick-up and drop-off time windows with customers. Finally, researchers are working on models that recognize new transportation infrastructures such as those prescribed by the Physical Internet Initiative. In this talk I will review these new models as well as the solution approaches developed to handle the added complexities. Finally, I will propose what I believe to be the next generation of models that the research community should develop and solve.



Online user: 1