|M.Sc Student||Yona Dror|
|Subject||Robust Solutions of the Military Logistical Network Problem|
|Department||Department of Industrial Engineering and Management||Supervisor||Professor Emeritus Aharon Ben-Tal|
During wartime, the military logistical network supplies the fighting units with resources of ammunition, fuel, water and food in the right place and at the right time to exploit fire capabilities as efficient as possible. A linear optimization program can compute the optimal initial stock levels and the optimal supply flows through a discrete time network, to provide the units consumptions during war, which are assumed to be known in advance. But as war is a series of unpredictable events, we can not assume to know the exact consumptions realization prior to war.
The purpose of the thesis is to suggest an efficient method to find an optimal logistical plan under uncertain consumptions. First we have assumed that uncertain consumptions are nested inside a defined uncertainty region that includes every possible realization without having any information considering the consumptions distribution. Then, we have used a simulation to find an optimal plan under uncertainty and adaptive and nonadaptive robust counterpart optimization, which inherently include uncertainty and can be solved in reasonable time.
Simulation has provided unreliable solutions. Conversely, the robust counterpart optimization has provided solutions with tunable reliability. The adaptive solutions were less expensive than the nonadaptive and using the folding horizon technique has improved the results even more. Additionally, an underestimation of uncertainty level has been successfully absorbed by using adaptive methods with an ellipsoid as the uncertainty region. Finally, it has become apparent that units must report their consumptions at the end of each period as soon as possible to reduce costs and risks.