Proceedings of the 2008 Winter Simulation Conference S. J. Mason, R. R. Hill, L. Mönch, O. Rose, T. Jefferson, J. W. Fowler eds. A DESIGN OF EXPERIMENTS APPROACH TO MILITARY DEPLOYMENT PLANNING PROBLEM Uğur Z. Yıldırım Ahmet Balcıoğlu İhsan Sabuncuoğlu Barbaros Tansel Department of Industrial Engineering Turkish Army Headquarters Bilkent University Turkish Army Bilkent, Ankara 06800, TURKEY Yücetepe, Ankara,06100, TURKEY ABSTRACT Even a small-scale military deployment scenario can have a large number of input variables (factors) that may We develop a logistics and transportation simulation that impact the outcome of the plan. In classical design of ex- can be used to provide insights into potential outcomes of periments (DOE), one can explore only a handful of cases. proposed military deployment plans. More specifically, With the use of computers and in a simulation setting, the we model the large-scale real-world military Deployment user can vary a large number of input variables. Yet, even Planning Problem. It involves planning the movement of with today’s computing power, complete enumerations of military units from their home bases to their final destina- all possible scenarios is an exhaustive task. To intelli- tions using different transportation assets on a multimodal gently sample the state space of possible alternatives in a transportation network. We use an intelligent design of simulation setting, we use an approach developed for experiments approach to evaluate logistics factors with the large-scale simulation experiments with many factors greatest impact on the overall achievement of a typical (Kleijnen et al. 2005), (Cioppa and Lucas 2007). real-world military deployment plan. In Section 2, the DPP is explained in detail. The simu- lation model of deployment problem is briefly presented in 1 INTRODUCTION Section 3. In Section 4, relevant information on DOE ap- proach used is explained, and details of our case study are Regional and asymmetric threats and the increase in presented. Our results are presented in Section 5. Final worldwide terrorist activity have made logistics and mobil- remarks are made in Section 6. ity increasingly important in our rapidly changing world. This paper deals with logistics and transportation simula- 2 PROBLEM AND SYSTEM DEFINITION tions that are used to provide insight into the potential out- comes of proposed logistical courses of actions prior to and The DPP involves the simultaneous and coordinated relo- after committing members of the military into harm’s way. cation of many military units from their home bases to More specifically, we model the Deployment Planning their designated destinations according to a given military Problem (DPP), defined and thoroughly described first by mission. This mission may be a preplanned and rehearsed (Akgün and Tansel 2007). DPP involves movement and one or may be due to a contingency that arose because of positioning of many military units from their home bases an escalating military confrontation or a natural disaster. In to their designated destinations to carry out a mission. the latter case, timely deployment of military units to their This movement mostly occurs on a multimodal (land, rail, destinations in the crisis area is of utmost importance. In sea, air, and inland waterways) transportation network us- the former case, the more important concern is cost. Usu- ing different transportation assets. During peace time, ally, a least cost deployment plan is preferred during peace plans are made to deploy the required number of troops time when time is not of essence. and equipment to potential threat or disaster areas. During If there is ample time to plan for a particular deploy- a time of crisis or natural disaster, it will be necessary to ment, this is called deliberate planning. When the time use these plans as they are, to modify them as necessary or available for planning and execution of an actual deploy- to create new deployment plans in a short time. For these ment is short, this is called crisis action planning or time- reasons, we have developed a simulation model of military sensitive planning. During crisis action planning, quick deployment with accurate transportation network infra- response and flexibility to adapt to rapidly changing situa- structure data and a medium-resolution allowing planners tions are crucial. Each deployment plan, deliberate or not, to develop and analyze plans in a relatively short time has as a minimum a list of cargo and pax (troops) to be (Yıldırım, Sabuncuoğlu, and Tansel 2007) . transported by type and quantity, and movement data by 978-1-4244-2708-6/08/$25.00 ©2008 IEEE 1234 Yıldırım, Sabuncuoğlu, Tansel, and Balcıoğlu mode of transportation, earliest times of departures from Real-world deployments are characterized by unpre- home bases, and earliest and latest times of arrivals at des- dictable stochastic events (breakdowns, accidents, delays tinations. etc.), load/unload/idle times at home bases, destinations, The movement of military units from their home bases and transfer points. Thus, a stochastic simulation is suit- to their final destinations are conducted in groups. A unit able to determine potential outcomes of deployment plans. is mostly divided into three components (advance party, pax party, and cargo party) during deployment. Ground 3 THE SIMULATION MODEL movement is conducted in convoys. The speed and com- position of convoys are decided by operational and tactical We created our discrete-event simulation of DPP by utiliz- objectives and constraints. Coordinated movements of ing Schruben’s (1983) Event Graph (EG) methodology. In components of several units are dictated mainly by avail- an EG, events are vertices (nodes) on a directed graph and ability and capacity of lift assets, capacities of transfer they represent state changes of the system. Directed edges points, weather, operational requirements, and intelligence (arcs) between vertices indicate how occurrence of one on enemy’s probable courses of action. event triggers another event. Solid arcs can be referred to Large-scale deployments over long distances may re- as scheduling arcs, whereas dashed arcs are called cancel- quire the outsourcing of heavy lift assets of civilian com- ing arcs. A tilde on an arc represents the conditional panies or other nations’. Most of the time, a unit’s own scheduling of the event at the head of the arc whenever the (organic) transportation assets will suffice to conduct a de- stated Boolean condition is true. EGs are not flow charts. ployment. Deployments over long distances (usually out- They represent the conceptual and logical models of our side a country’s borders) are classified as strategic de- simulation. More detailed information on EGs are avail- ployments. Strategic deployments are mostly conducted able in Sargent (1988) and Buss (2001). using sea and air lift assets. Once a strategically deploying The modular structure of our simulation is constructed unit reaches its destination country, other available modes via the Listener Event Graph Object (LEGO) framework of transportation may be utilized. Deployments inside a (Buss and Sanchez 2002). The listener pattern of object country’s borders are referred to as intra-theater deploy- oriented programming allows LEGOs to work. EGs and ments. This type of deployment may utilize any of avail- LEGOs can be programmed using Simkit, a Java Applica- able (land, sea, air, and rail) modes of transportation. tion Programming Interface (API) developed at the Naval Transportation mode changes may be required at Postgraduate School and freely available via transfer points during both strategic and intra-theater de- <http://diana.nps.edu/Simkit/> . ployments. Furthermore, successive mode changes may be We have integrated a geographical information system necessary at different transfer points during a deployment. (GIS), named GeoKIT, into our simulation. Benefits of However, the fewer the mode changes are at transfer combining simulation and GIS include usage of real and points, the easier is the deployment. Main transfer points current transportation network data, an accurate animation, are harbors, train stations and airports. At these locations, and extended analysis possibilities in route selection. the pax (troops) and cargo (weapon systems, material, GeoKIT is a Java API, which enables easier integration equipment, and supplies) a unit has, collectively referred to with Simkit. Furthermore, selection of GeoKIT as the GIS here as items, are transferred from one set of transportation system of choice is due to its superb capabilities and com- assets to another set that operate on a different network. prehensive set of components to manipulate and use geo- This location is also called a Port of Embarkation (POE). spatial information. More information on GeoKIT can be The next mode change location, where the items are off- found at <http://geokit.bilgigis.com>. loaded and loaded onto another set of transportation assets The details of our simulation model, developed using is called a Port of Debarkation (POD). These may be sea, the tools briefly described above, are in (Yıldırım, Sabun- rail and air POEs or PODs (namely, SPOE/SPOD, cuoğlu, and Tansel 2007). RPOE/RPOD etc.). At transfer points, units usually queue up before being 3.1 Verification and Validation loaded on vessels. This location is called a staging area where units wait and prepare for shipment. A staging area Verification and validation were conducted by using ap- can be regarded as a service point, i.e. one with a certain propriate methods explained in Sargent (2001). To men- capacity of material handling equipment and load/unload tion a few more specifically, for face validity, we have docks. When there is not enough capacity at a staging area discussed inputs and outputs of the model and its EGs with to hold large number of deploying units, a marshaling area potential users of the model. We used assertion checking is operated. A marshalling area can be regarded as a wait- to verify that the model functioned within its acceptable ing/parking place.