Seminar: Restoring Accessibility for Effective Disaster Response

Loading Map....

Date(s) - 03/11/2017
13:30 - 15:30

Bilkent-Unv EA409

Categories No Categories

Doç. Dr. Sibel Salman

School of Engineering, Koc University

Nov 3, Friday 13:40



A common consequence of natural disasters is the destruction of infrastructure, including roadways that are critical for disaster response operations. For instance, after an earthquake, collapsed buildings, bridges and viaducts, as well as damage in the roads caused by sliding, etc., lead to blockage of road segments.  Loss of functionality in the road network impedes accessibility within the disaster area and may even render some regions isolated. In the immediate disaster response phase, roads should be recovered by clearing the debris or reconstructing damaged segments to restore accessibility in shortest time. The main objective of this research is to provide an efficient solution method to generate a synchronized work schedule for the work troops responsible for recovering the roads. We define an arc routing problem that seeks for the optimized routes of multiple work troops. In this problem, we maximize the total prize (benefit) gained by reconnecting isolated road network components within a specified time limit. Closed roads are traversable only after they are unblocked/cleared by one of the teams. The solution should determine the synchronized routes of each work troop such that none of the closed roads are traversed unless their unblocking/clearing procedure is finished. In this study, we develop an exact Mixed Integer Programming (MIP) formulation to solve this problem. Since this formulation falls short of solving realistic sized instances, we develop a math-heuristic method to obtain near optimal solutions. The math-heuristic works by solving single vehicle problems sequentially with updated prizes. To obtain an upper bound, we first relax the timing elements in the exact formulation and then solve its relaxed MIP, which decomposes into single vehicle problems, by Lagrangian Relaxation. We show the effectiveness of the proposed methods computationally on both random Euclidean and Istanbul road network data generated with respect to predicted earthquake scenarios.



Brief bio of the speaker

Sibel Salman is an Associate Professor of Industrial Engineering at the College of Engineering, Koç University.  Prior to joining Koç University, she held a faculty position at Purdue University. She got her Ph.D. in Operations Research from Carnegie Mellon University, and M.Sc. and B.Sc. degrees from Bilkent University.  Her research focuses on network optimization problems arising in disaster management, distribution logistics and production planning.