DENİZ AKSEN'S APPLICATION FOR PROMOTION

August 2012

 Main

 Résumé (opens in a new window)

 Statistics in a nutshell

 Research statement

  Ongoing research 

 Teaching

 List of journal publications

 Books, book chapters and proceedings

 Analysis of citations

ONGOING RESEARCH

1.  PARTIAL FACILITY INTERDICTION MODELED AS BILEVEL AND TRILEVEL PROGRAMMING PROBLEMS

In this research stream which produced four journal papers and one refereed book chapter to date, my next step is to look into the triplet problem of facility location, capacity planning, and partial facility interdiction. My progress and agenda in this area can be summarized as follows:

1.1.  ELECTROMAGNETISM-LIKE ALGORITHM IN LIEU OF MULTI-START REVISED SIMPLEX SEARCH

One master thesis student under my co-supervision has completed her research on the implementation of an electromagnetism-like algorithm (EM) for the bilevel partial facility interdiction problem (BPFIP) with capacitated facilities and demand outsourcing. We tackled this problem in our most recent paper [COR2014] with a multi-start revised simplex search method (MSS). Computer experiments with EM showed that it works substantially faster than MSS while achieving a slightly better average objective value. In the test problems where it falls behind MSS, EM does not suffer from serious deterioration in solution quality.

This branch of our research which revisits the BPFIP has been completed. Reporting the implementation details of EM and its results in a new journal paper will follow before the end of 2012.

Copyright © 2007 Elsevier Ltd. All rights reserved.

1.2.  JOINING TWO METAHEURISTICS INTO A HYPER-MATHEURISTIC : ELECTROMAGNETISM-LIKE ALGORITHM AND TABU SEARCH WITH HASHING

We embedded EM in a Tabu Search with Hashing method (TSH). Recently, we successfully applied TSH to a triplet facility location, protection, and full interdiction problem in [COR2012]. The triplet problem was modeled as a mixed-integer bilevel programming problem with binary variables in the lower level. The role of the leader was assigned to the system planner (the defender) and the role of the follower was attributed to the attacker. The aim of the attacker is to inflict as much disruption as possible on a median-type service network where service is delivered at facility sites chosen by the customers on the basis of their proximity.

This time, we capture in a trilevel programming problem the game theoretic relationship between the system planner and the attacker. The top level which involves the BPFIP in its lower level captures the system planner's facility location and capacity installation decisions. Not only does the system planner decide the number and locations of the facility sites but also determines their initial capacities which can be installed only as integer multiples of a base capacity. Given the location and capacity configuration of the facilities before interdiction, it is the attacker's turn to figure out how to react to inflict the maximum degree of disruption. He solves the associated BPFIP where the leader of the comprehensive problem--namely the system planner--becomes this time the follower, and reacts to the attacker's partial interdiction by reassigning customers to facilities with sufficient residual capacities, or alternatively by outsourcing their demands to an external supplier.

The goal in formulating this problem as a trilevel program is to pursue "best preparedness against the possibly worst man-made disaster scenario." This time we model the relationship between the leader and the follower from the perspective of the system planner. She wants to build as much as an efficient median type service network against the imminent threat from an intelligent attacker. To this end, she has to take proactive measures in the construction of the median type service network. Trilevel programming is attempted very rarely in the interdiction literature. With this study we will contribute to the literature a hyper-matheuristic tailored for the solution of such problems.

1.3.  SURVEY OF FACILITY INTERDICTION MODELS AND SOLUTION METHODS PROPOSED IN THE LITERATURE

Facility interdiction models in median and coverage type networks constitute a relatively young research venue. They were first introduced and formulated in the seminal paper of Church, Scaparra and Middleton (Ann Assoc Am Geogr, 2004). A great deal of research has been published since then. Bilevel and trilevel programming models have been proposed to capture the leader-follower dilemma between the defender and the attacker either from the defender's or the attacker's perspective. Models in which the attacker acts first and assumes the role of the leader seek to determine the most vulnerable (critical) facilities, while models in which the defender is the leader aim at finding out how to locate and/or protect those physical assets.

As we approach the 10th anniversary of the pioneering study by Church et al. (2004), the need for an annotated literature survey is conspicuous. Facility interdiction models have been of interest to me since 2007. I was introduced to this subject after reading the "online first" version of Scaparra and Church's paper (Comput Oper Res, 2008) about protection of critical infrastructure. In the past five years I became familiar with the overwhelming majority of the related studies. I have put an in-depth survey of these studies on my agenda for the next year. Without doubt, once completed, It will  benefit practitioners who need hands-on solution techniques to deal with mixed-integer bilevel programming problems as well as the OR community at large.

 Top  Top of the page

2. COLLECTION OF MEDICAL WASTE FROM HOSPITALS: VEHICLE ROUTING WITH INTERMEDIATE FACILITIES

One of my MS thesis students (see my current and past research supervisees on my résumé page) is nearing completion of her thesis work. Her research is about the application of a reactive tabu search heuristic to a real-world vehicle routing problem with an intermediate facilities (VRPIF). More precisely, she has been investigating the collection of medical waste by the Metropolitan Municipality of Istanbul from hospitals and health centers on the European and Asian sides of the city. The vehicles allocated to this operation are dispatched from a municipal parking lot, visit hospitals on their route, and eventually bring the collected hazardous waste to a landfill facility where it is burnt to ashes and buried. All vehicles have to stop by at the landfill facility before returning to their parking lot. Vehicle routes are constrained by a common maximum tour duration and cargo capacity. The problem is treated separately for the Asian and European sides of Istanbul each having a dedicated landfill facility.

The VRPIF is a well-studied combinatorial optimization problem. In our research we use real-world distance data that belongs to the complete road network connecting the pickup points, the parking lot, and the landfill facility. This data is obtained from the application programming interface (API) of Google Maps with JavaScript and PHP code which is compiled on a virtual PHP server. I am planning to convert this study into an applied reverse logistics / waste management paper to be submitted to an SCI journal in the first half of 2013.

3. APPLIED VEHICLE ROUTING WITH SIMULTANEOUS PICKUPS AND DELIVERIES

The purpose of this research is to optimize the distribution and collection of cash from or to the branch offices and ATM booths of an international banking company in Turkey. The transfer of cash and other valuables such as gold and silver is outsourced by the bank to a foreign company well-known in city logistics. The company operates its own armored vehicle fleet to accomplish this task according to a weekly schedule determined a priori. Armored vehicles are dispatched daily from the central garage of the company, operate for a period of four to eight hours at most, and return to the garage at the end of their tours. Vehicles which arrive at the garage before or at noon get prepared for the next tour after a certain break period.

In addition to banks and ATM machines, the set of nodes to be visited includes also some coffee shops which deliver their cash revenue at the end of the day to the headquarters. These shops as well as the bank offices constitute backhaul customer nodes where only cash collection takes place. ATM machines, on the other hand, may require simultaneous pickup of empty and delivery of replenished cash cassettes. Thus, ATM  booths are either linehaul, or backhaul, or pickup-and-delivery nodes.

Copyright © 2012 Texas Armoring Corporation

The comprehensive problem needs to be split into two separate problems since an armored vehicle tour must not cover coffee shops and ATM booths or bank offices at the same time. The former nodes are visited on dedicated tours the planning of which can be modeled as a vehicle routing problem (VRP) with maximum tour duration. The second problem can be formulated as a pickup-and-delivery problem (PDP) with the same type of temporal constraint. The VRP literature is abundant with algorithms tailored for the PDP. However, before adapting any solution method to the company's armored vehicle routing problem, we have to obtain the real-life distance matrix containing the garage and all other relevant nodes. To this end, we will benefit again from the application programming interface (API) of Google Maps.

A master thesis student in the department of Industrial Engineering of Boğaziçi University has been recruited and assigned to this project. We are planning to complete this project within six months starting in the last quarter of 2012.

 Top Top of the page


Last updated on 22-10-2014 by Deniz Aksen  ©2003-2012

















Private Matters

Privacy