TUM – TUM – Menü

 

 
SCoNDO IV

Students' Conference on Nonlinear and Discrete Optimization - July 13th, 2013

 

Acknowledgements

Organizers

The conference is organized by
Christian Böhm
Florian Lindemann
Andre Milzarek
Wolfgang Riedl
Michael Ritter
Paul Stursberg

(Preliminary) Schedule

Time Talk Team
9:00 a.m. Welcome The Organizers
9:20 a.m. TUM-specific Optimal University Timetabling University Timetabling
10:00 a.m. Smoothing Algorithms for Financial Model Calibration Volatility
Coffee Break
11:00 a.m. Mathematical synthesis of an optimal planetary gear train Gear Train Optimization
11:40 a.m. Optimization in Robotics Robotics
12:20 p.m. Robust Flight Scheduling Robust Flight Scheduling
Lunch Break
2:10 p.m. Optimal Open-Loop Control of Chemical Reactions Chemistry
2:50 p.m. Traffic Network Optimization Traffic Infrastructure Optimization
Coffee Break
4:00 p.m. Primal-Dual Methods for Matrix Completion Problems Recommender Systems
4:40 p.m. Operating Room Scheduling Operating Room Management
5:20 p.m. Evaluation The Organizers
6:00 p.m. Certificates The Organizers
7:00 p.m. Conference Dinner All Participants and Friends

Talks

TUM-specific Optimal University Timetabling

Team "University Timetabling"Johannes Hofbauer, Michael Lintl, Laura Velikonja, Jonas Wunderlich

The timetable is a crucial part of every student’s daily life. Do none of the courses overlap, is there enough time between classes to change the lecture room and maybe even a lunch break?

With more and more students crowding the TUM every year, the process of scheduling all courses in a satisfactory manner becomes a tough challenge. We will have a look at a mathematical model where we assign a certain time and room to every class by finding a shortest path through an enormous graph. Can this be implemented with a short running time and does this approach lead to a satisfying solution? What are the drawbacks of the first naive idea? Moreover, which alternative approaches are there to be examined in search of the best possible timetable? All results shall be presented in this well-scheduled talk.

Smoothing Algorithms for Financial Model Calibration

Team "Volatility"Vincent Bates, Katharina Juranek, Carina Ruckerbauer, Maria Veleva

The pricing accuracy of financial products has become increasingly important due to the significant growth in traded volumes. Most models depend on an arbitrage-free implied volatility surface. A major drawback of the well-known Black-Scholes formula, which can be used to price European-style options, is the assumption of a constant volatility of the stock price movements. More sophisticated models, like Dupire's local volatility model, incorporate changing volatilities. But in order to be properly calibrated, they rely on entirely arbitrage-free call-price data. We present an algorithm suggested by Fengler, which computes arbitrage-free prices from given potentially mispriced market data. Based on Fengler's work, we offer an improved version, which trades a bad error propagation for a higher-dimensional problem formulation. Comparing both, the gained advantages in pricing accuracy justify to neglect the increasing computational costs.

Mathematical synthesis of an optimal planetary gear train

Team "Gear Train Optimization"Ville Peri, Moritz Ratzesberger, Martin Schwenk, André Wagner

In a modern motorized vehicle, the gear train is a vital part which plays a major role in fuel economy, ride comfort and performance in all conditions. The planetary gear train in particular offers an excellent combination of power transmission and compactness, especially fulfilling the needs of modern passenger cars. In order to overcome its design complexity while maintaining the advantages, efficient methods for synthesis are needed. Our work presents a quadratic optimization model for the fast synthesis of planetary gear trains. We formulate the gear train components as a graph model and through mixed-integer programming obtain an optimal gear configuration for given transmission ratios. We analyze the applicability of the resulting gear with graph analysis and real world comparisons.

Optimization in Robotics

Team "Robotics"Agnes Ramle, Marius Ritter, Tina Stark, Nicole Zutavern

The movement of a robot arm between two prescribed points can be optimized in many different ways. For example, it could be reasonable to make the movement as fast as possible or to reduce the engine's effort by using gravitational forces. The tasks in the project were to build a mathematical model, search for good optimization criteria and implement the problem with AMPL. In the presentation the project's progress from the physical background, the mathematical modeling and the choice of suitable initial values and objective functions up to the implementation, analysis and visualization of the results will be demonstrated. The top priority is given to the presentation of the reached aims as well as to the difficulties in acquiring them for a robot arm with two joints in a two-dimensional space.

Flight Scheduling and Recovery

Team "Robust Flight Scheduling"Patricia Jozefowicz, Tanja Niels, Pia Puchner, Martin Siegmund

An important matter in daily airport activities is the reaction towards delays and cancellations. If the airport operates at the limit of its capacities a small delay can lead to a lot of further and bigger deviations. That is expensive and uncomfortable for passengers as well as the airport and the operating airline. But implementing a recovery concept can already begin in the scheduling process. Therefore we focused on generating a robust schedule that is naturally not optimal in terms of utilized capacity, but leads to simpler and more economic recovery strategies. We evaluated our results and found a reasonable compromise between robustness and optimality.

Optimal Open-Loop Control of Chemical Reactions

Team "Chemistry"Javier Fischer, Florian Ionescu, Philipp Krenz, Patrick Stocker

Chemical engineering is a significant economical field. The core of a chemical process is the reactor where the reaction takes place. Diverse models for reactor conditions in various reactors and with different reaction types have been developed. These models are based on idealized and simplified considerations and can be used as a basis for experimental optimization in practice. Control of reactor conditions for a transition between steady states is achieved by manipulating e.g. the flow. In this talk, the transition trajectories between steady states in a consecutive exothermic reaction in one continuous stirred-tank reactor (CSTR) as well as an exothermic reaction in a sequence of two CSTRs with recycle are analyzed using AMPL, an algebraic modeling language for solving large-scale optimization problems.

Traffic Network Optimization

Team "Traffic Infrastructure Optimization"Johanna Brandstetter, Lisa Gerstner, Saskia Schiele, Maria Theresia von Soden

Flow problems are a significant part of mathematical optimization and are often used to improve the traffic in a given network. In this project, we consider a different approach: How can one improve the traffic flow by changing the network (i.e. by building new streets or adding new lanes to already existing streets)? We used a genetic algorithm as an approximation in order to solve this problem. Using data from the Munich district "Perlach", we simulated a realistic traffic agent schedule with the simulation program MATsim, to benefit the traffic agents in that certain area.

Primal-Dual Methods for Matrix Completion Problems

Team "Recommender Systems"Fabian Stark, Felix Müller, Markus Sterflinger, Felix Xaver Krisch

Recommender systems seek to predict the rating that an user would give to an item (such as movies, music, or books) or social elements (e.g. people or groups) based on ratings that were given to other items. Recommender systems have become extremely common in recent years. A few examples of such systems are Moviepilot, Amazon (products you may be interested in), or Facebook (people you may know). Huge sparse datasets make it difficult to find a good recommendation for a user in a reasonable time. Under the assumption that only a few parameters influence the preference of the user we obtain a rank minimization problem which is NP-hard and cannot be solved efficiently. In this talk, several implementations of a convex relaxation of this problem are discussed and compared to each other.

Operating Room Scheduling

Team "Operating Room Management"Jaclyn Amberg, Julia Baukmann, Henrik Poppe, Alysa Reuter

Operations generate about 42% of a hospital’s revenues. Therefore the scheduling of operations plays an important role for every hospital.
In our project we considered a particular instance of the operating room (OR) scheduling problem with one surgeon and n operations that need to be allocated to r ORs. The goal is to find an allocation of the operations that minimizes the cost function, which arises through idle time, vacant OR time and overtime.
In the first part of our project, we set up a mathematical model and devised a heuristical algorithm for the problem. Afterwards, we computed the exact solution to the problem by "Xpress Mosel". To improve the running time we used methods from combinatorial optimization.

Organizational Remarks

Abstract Submission

  • For the conference program each team needs to submit a title and an abstract for their talk (in English). The abstract is limited to a maximum of 800 characters.
  • Please submit title and abstract via email to m.ritterma.tum.de until July 1st, 2013.
  • Submissions are accepted both in plain text format and in LaTeX format. Please do not send PDF files.
  • Optionally, a small (roughly 1cm x 1cm) image may be placed next to your abstract. If you wish to submit an image, please include it with the abstract in either PDF, PNG or JPG format.

Talks

  • Each team will have 25 minutes for their talk, plus an additional 10 minutes for questions and discussions.
  • The room is equipped with a projector and whiteboards. If you need anything beyond that, please let the organizers know well in advance.
  • A common notebook for hosting the presentations will be available to save time. Please send your presentations to m.ritterma.tum.de or deliver them on a USB pendrive to Michael Ritter before the conference starts, i.e. until 8:50 a.m. on the day of the conference. We will have both a Mac and a Windows Notebook available with PDF viewer, Powerpoint and Keynote installed. A presenter with built-in laser pointer is also available for the talks.
  • If you need additional software or absolutely must use your own notebook, please let us know as soon as possible.

Conference Venue and Schedule

  • The conference will take place in room 0606 on the TUM main campus (Arcisstr.), see below for directions.
  • If you wish to rehearse your talk in that room, please consult TUMOnline to see when the room is free. You might also want to get a key for switching on the lights which is available from the organizers on request. Room 0602 right next to the conference room looks just the same, so you can also use that room for your rehearsals.
  • A detailed conference schedule will be posted on this website shortly. The conference will approximately start at 9:00 a.m. and end around 7:00 p.m.

Conference Dinner

  • To celebrate a successful conference, we will meet for a conference dinner after SCoNDO IV at a restaurant close to the conference venue. Details will be announced later.
  • If you cannot join us for the conference dinner, please let us know until Monday, June 24th, 2013 so that we can reserve enough tables.
  • If you have any special dietary requirements that might restrict our choice of restaurants please let us know immediately so we can make the proper arrangements. We will choose a restaurant that serves a choice of vegetarian food.

Conference Venue

The conference takes place at room 0606 at
Technische Universität München, Stammgelände
main building
Arcisstraße 21

Please remember to enter through the main entrance using your student ID if necessary (or stating that you want to visit the conference or the university library) as the other entrances will be closed on Saturday.


Stammgelände der Technischen Universität München, weitere Pläne und Anfahrtsbeschreibung im TUM-Roomfinder.

Parking is very limited in the area, thus public transport is recommended for getting to the conference venue. The closest subway stops are Theresienstraße (U2), Königsplatz (U2) and Odeonsplatz (U3/4/5/6), the closest bus stop is "Technische Universität" (bus 100).


Größere Karte anzeigen

Research Unit M9


Department of Mathematics
Boltzmannstraße 3
85748 Garching b. München
Germany
phone:+49 89 289-16858
fax:+49 089 289-16859
sekretariat-m9ma.tum.de

Professors

Prof. Dr. Peter Gritzmann
Applied Geometry and Discrete Mathematics

Prof. Dr. Andreas S. Schulz
Mathematics of Operations Research
(affiliated member of M9)

Prof. Dr. Stefan Weltge
Discrete Mathematics

News

Jan 25th, 2019
Case Studies 2019: Preliminary Meeting on Wed, Feb 6th, at 16:00 in room MI 03.06.011.