HomeAbout UsBrowse This IssueBack IssuesNews DispatchesSubscribing to Army LogisticianWriting for Army LogisticianContact UsLinks































Optimizing Logistics Through Operations Research

Army logisticians often have to make quick decisions about how to support the warfighter based on incomplete data and ill-informed assumptions. Making such decisions is a necessary skill for logisticians, because an 80-percent solution that arrives on time is better than a 100-percent solution that is late. The ability to approximate a solution to a logistics problem and then accomplish the mission is a valuable skill for any logistician. However, when they have time to plan and analyze courses of action, logisticians need to look for optimal or improved solutions to logistics problems that consider all functional constraints. This is important because the best possible solutions save resources and time while maximizing a unit’s ability to support.

Operations research (OR) is a branch of applied mathematics that uses algorithms, simulation, modeling, queuing, and stochastic methods to optimize or improve a real-world situation. OR was developed by a group of British and American mathematicians who were studying strategic logistics problems during World War II. Since that war, this branch of mathematics has been used in a variety of industrial and military applications.

For example, at New Mexico State University, an ongoing OR project for the New Mexico Chile Task Force is optimizing a supply chain network involved in the harvesting, transporting, and processing of red chile peppers. In the project, data and statistics fitting is used to assign probability distribution functions to crop maturation, harvesting rates, transportation rates, crop damage, and processing rates. After assigning valid queuing rules to the system, the research team was able to write a program using ProModel, a discrete-event simulation modeler that accurately mirrored the actual conditions. Using this program as a base, the team adapted the model to quantify the value of lengthening the growing season, optimizing transportation, and constructing storage based on the amount of crop loss.

The project’s value was in creating a model of a supply chain that used accurate stochastic data to accurately describe and then improve on a complicated logistics system. Although writing a program is not a valid method of solving problems in an Army unit, logisticians can use OR methods to deduce and compare feasible solutions to common problems.

This article will describe a few available OR techniques. Using the references listed below, those interested can examine methods for improving systems within their work areas.

Analyzing a Network

A useful OR technique is finding optimal solutions to problems involving start nodes, arcs, and destination nodes. A basic problem in OR is the “transportation problem,” in which there are known supply bases, known customer demands, and known costs to each route from supply base to customer. (The costs could be in time, risk, shipping costs, or something else that is considered important.) The objective is to minimize total cost while meeting all demands. The chart below shows an example of a transportation network and a sample transportation network spreadsheet with costs.

This problem can be solved using either a program or a number of algorithms performed by hand. In this example, all demands equal the sum of all supplies in order to simplify the spreadsheet. In more realistic examples, a dummy variable is added to balance the equation in order to perform either an algorithm or a linear program. After using an algorithm and checking with linear algebra, the optimal solution is shown in the chart below.

Networks, such as maximum flow networks (which are useful in describing port activities), decision trees, lattices, and other deterministic flows, can be used in many ways to solve logistics problems. For those interested in learning more, see the references listed below.


In OR, programming is used to quantify a problem involving an objective function that is subject to one or more constraints in the system. An objective function attempts to perform actions that affect the output of a system, such as minimizing shipping cost, maximizing throughput, or maximizing material shipped to an area. Constraints are functions that place limits on the range of the objective function. These can include limitations on infrastructure capacity, warehouse space, cost, trucks available, and integer, or non-negativity, limits.

The chart (below) outlines a scheduling problem that illustrates the usefulness of programming. In this example, a maintenance shop must complete four jobs in a week using three mechanics who are given different times to complete the jobs.

For this example, a number of constraints must be listed. First, each worker can only log a maximum of 40 hours during the week. Second, each job must be completed at the end of the week. Third, only one worker can be assigned to a job. For this problem, the decision variables are labeled as “Xij” (assigning “i” worker to “j” task) and will have a value of either “1” or “0” (that is, to either assign or not assign a worker to a task). The cost attribute is labeled as “Bij” (the time for “i worker to complete “j” task).

The resulting problem can be written as shown in the chart below. It is solved by using an optimizing algorithm, called the simplex method, that uses linear algebra in order to solve series of linear equations. [In nonlinear programs (those involving objective functions and constraints with higher-ordered variables), other methods, such as interior point methods, linear approximation, and Hessian/Duality matrices, must be used.] In real-world problems with multiple variables and constraints, it usually is not feasible to do calculations by hand. For this type of work, many computer programs like MATLAB and LINDO are available. The key to solving these problems is to set up the problem correctly in order to describe the situation and the desired type of solution accurately.


Simulation involves using a combination of deterministic and probabilistic functions to model the problem and then predict actual system improvements after changes. Because of the large number of calculations involved and the need for multiple runs, simulations are almost always run on a computer. A number of excellent simulation programs are available, such as ProModel or Arena, that use graphical interfaces to help model an actual system. Simple models, however, can be run from Microsoft Excel.

The basis for the simulation is generating random numbers and assigning ranges of numbers to values that fit a probability distribution function (otherwise known as Monte Carlo simulation). Although this type of simulation has been used in many scientific disciplines, it was first used to calculate the properties of the neutron by Enrico Fermi and Stanislaw Ulam in 1930.

To use Monte Carlo simulation, you could generate a uniform series of random numbers from zero to one (U(0,1)) and then assign an equal range to each probability in the following way—

  • 0.001-0.166: die roll equals 1.
  • 0.167-0.332: die roll equals 2.
  • 0.333-0.498: die roll equals 3.
  • 0.499-0.665: die roll equals 4.
  • 0.666-0.832: die roll equals 5.
  • 0.833-0.999: die roll equals 6.

Using a computer-generated pseudo-random number generator (PRNG), you then could model the real situation of a die roll. More complicated simulations, for example, could model fuel usage in a division based on past data and assumptions about the current mission.

In a real-world logistics model, the simulation would look simultaneously at a variety of distributions where the means, variances, and distribution types are fit by examination of past data while adding in deterministic data based on known parameters of the future mission. By running the simulation in multiple iterations, you would be able to get a stasis, or expected solution, along with data such as network utilization, efficiency, and expected system variance.

OR is directly related to logistics because it focuses on optimizing real-world tasks, such as designing an efficient supply chain. We highly recommend that anyone who is contemplating an advanced civil schooling assignment look into industrial/OR engineering as an option. Most programs require an engineering or science background and coursework in multivariable calculus, differential equation, and statistics. We also recommend that anyone pursuing a master’s degree review discrete/vector mathematics and linear algebra before starting.

For those either not interested in advanced civil schooling or not in that period of their careers, we recommend any of the reference books in the list on page 49. Operations research is a continuously evolving discipline that is directly related to the main objective of logisticians—efficiently supporting the warfighter.

Dr. Delia J. Valles-Rosales is an assistant professor of industrial engineering at New Mexico State University. She received her Ph.D. degree in industrial engineering from New Mexico State University. Her research specialties are microelectromechanical manufacturing and packaging and discrete-event simulation. She can be contacted at dvalles@ad.nmsu.edu.

Major Donovan O. Fuqua is assigned to the 595th Transportation Group-Provisional (SDDC) in Kuwait. He completed an M.S. degree in industrial engineering/operations research at New Mexico State University while on an advanced civil schooling assignment and currently is working on his doctorate in engineering with a specialty in operations research and supply chain management and a minor in mathematics. He can be contacted at donovan.fuqua@us.army.mil.