DMMP

Assignments (theses)

The researchers of DMMP work in several research areas with applications in many fields, like health care, traffic, energy, ICT, games and auctions, and timetabling, see the overview of previously completed theses. Therefore, the list below is not exhaustive; only more concrete assignments are listed. If you are interested in assignments for an internship or master's thesis in the field of DMMP, please contact any member of the group.

List of ongoing and completed master theses.

Algorithm Improvement For Tracking Complex Targets

Master's thesis at Thales (Hengelo)
Tracking can be described as the processing of measurements obtained from objects in order to estimate the states of these objects. In case of a radar system, the object of interest can be an aircraft with the position and the velocity as the states.

Tracking applications in current radar systems are often based on Multiple Hypotheses Tracking (MHT) techniques. The MHT algorithm is based on two important assumptions. First of all, each measurement can belong to at most one target. Secondly, each target can only produce at most one measurement in each scan. The second assumption of the MHT algorithm is for some complex targets not always true, which can lead to degraded performance of the algorithm.

In this study we would like to explore the possibilities to include these measurement issues on these complex targets into our current solutions.

For more information, please contact Georg Still.

Graafopdeling Voor Netwerken Van Wegen

Master thesis
Gegeven een digitale wegenkaart van Europa (zo'n 40 miljoen nodes), bepaal een partitie/opdeling in verschillende cellen met maximaal x nodes per cel en zo min mogelijk edges tussen de cellen. Met behulp van een dergelijke partitie kunnen we razendsnel route berekeningen doen: hoe beter de partitie des te sneller de berekening.

Deze opdracht zal uitgevoerd worden in samenwerking met CQM.
Voor meer informatie, kun je contact opnemen met Marc Uetz.

Hoeveel Fietsen Op Het Station?

Master's thesis
Onlangs heeft Goudappel Coffeng een nieuwe methode ontwikkeld (WINST) om het aantal in- en uitstappers van een nieuw station te voorspellen (zie: http://www.goudappel.nl/actueel/2012/03/12/winst-methode/). Deze methode maakt gebruik van een zogenaamde Reference Class Forecasting. We denken dat deze methode ook toe te passen is op andere vraagstukken zoals het fietsparkeren bij stations. Met name de vraag hoeveel parkeerplaatsen zijn noodzakelijk en welke variabelen/gegevens zijn nodig om dit goed te voorspellen. En is de methodiek achter WINST daarvoor geschikt of zijn er betere methoden.

De exacte opdracht zullen we met de student, UT-begeleider en begeleider vanuit Goudappel Coffeng verder vorm geven. Het is de bedoeling dat de student bij Goudappel op kantoor werkzaam is. De locatie (Eindhoven of Deventer) is afhankelijk van de interne begeleiding en kunnen we nog wel verder afstemmen.

Neem voor meer informatie contact op met Marc Uetz.

Smoothed Analysis Of High-Dimensional Optimization Problems

Master's thesis
For many optimization problems, finding optimal solutions is prohibitive because the problems are NP-hard. This often holds even in the natural case, where the instances of the optimization problem consists of points in the Euclidean plane. In order to still be able to solve these problems, heuristics have been developed in order to find close-to-optimal in reasonable time. While many such heuristics show a remarkable performance in practice, their theoretical performance is poor and fails to explain practical observations.

Smoothed analysis is a relatively new paradigm in the analysis of algorithms that aims at explaining the performance of such heuristics for which there is a gap between theoretical analysis and practical observation.

Recently, Bläser et al. (Smoothed Analysis of Partitioning Algorithms for Euclidean Functionals, Algorithmica, to appear) have developed a framework to analyze so-called partitioning heuristics for optimization problems embedded in the Euclidean plane. The goal of this thesis is to generalize this framework to higher dimensions and to apply it to analyze further heuristics for Euclidean optimization problems.

For more information, please contact Bodo Manthey.