Models address the complex problem of assigning global intelligence, surveillance and reconnaissance resources.
A George Mason University (GMU) model could help the intelligence, surveillance and reconnaissance (ISR) community gauge which assets are more valuable than others.
Two seemingly unrelated events occurred in May.
Renowned mathematician George B. Dantzig died on May 13 at age 90, and the U.S. Strategic Command stood up a new component to focus on global intelligence, surveillance and reconnaissance efforts on May 31. A mathematical model used to optimize resource allocation could tie Dantzig’s work to the new group.
As a U.S. Air Force combat analysis branch chief during World War II, Dantzig applied his prowess as a statistician to analyzing bombing missions over
ISR expertise is coalescing under the U.S. Strategic Command’s (STRATCOM’s) Joint Functional Component Command–Intelligence, Surveillance and Reconnaissance (JFCC-ISR) component. In March, Secretary of Defense Donald Rumsfeld appointed Vice Adm. Lowell Jacoby, USN, director, Defense Intelligence Agency (DIA), as the JFCC–ISR’s component commander. The new group will reside at the DIA’s facility but will remain a STRATCOM asset. Scholars and defense industry officials are pushing the new group to adopt Dantzig’s solution to speed up and optimize the process of assigning global ISR assets. The nudge comes as the nation’s ISR community sharpens its focus on the issue.
Officials from both organizations declined to characterize current ISR asset allocation processes specifically, citing the discussion as premature until the new group solidifies. But current processes are “likely the same processes that have been in place,” Donald L. Black, a DIA spokesman, says.
If this is the case, some say those processes could use a serious computer upgrade. Today, assigning ISR assets is largely a manual effort, says Dr. Andrew G. Loerch, associate professor, George Mason University (GMU),
Because of his background and credentials, Loerch says he has access to high-level military analysts who are working on the global ISR asset allocation effort. He characterizes the current process in this way: “Smart guys sit around and say, ‘This is what we’re going to do.’” Computers assist ISR analysts, but “there’s not a prescriptive method where you put in the parameters, punch a button and out comes a good solution of what to do,” Loerch adds.
Linear programming could provide such a solution. In Dantzig’s 1940s-era U.S. Army Air Corps, computers lacked enough power to handle linear programming. “Now you can solve huge problems using the method because computers are better and there are much more efficient algorithms,” he explains.
Linear programming solutions addressing ISR asset allocation are surfacing in the academic and private sectors. A GMU graduate-level paper, presented on the day of Dantzig’s death, takes on the problem of assigning global manned and unmanned ISR assets. Titled “Effective Allocation of Global (EAGLE) Airborne ISR Assets using a Mixed Integer Optimization Model,” the recently submitted GMU effort is “a step in the right direction,” Loerch says. He reviewed the model, nicknamed EAGLE ISR, as a graduate adviser for the operations research department.
Heath A. Hammett, a GMU graduate student in operations research and systems engineering, says the effort grew out of a realization that STRATCOM “wasn’t using a very robust or optimal method for allocating ISR assets. Linear programming seemed like a perfect fit.” Hammett and three other graduate students created the model to satisfy a GMU applications seminar requirement.
Hammett’s idea for the paper emanated from observations made while working with STRATCOM officials as an associate for McLean, Virginia-based Booz Allen Hamilton, a strategy and technology consulting firm. He describes his take on the command’s current process: Lower level commands send requests up the chain to subject matter experts who sit around a table and generate an allocation plan based on different criteria. They send those plans to combatant commands, which relocate assets when or where needed. The combatant commands then task the actual asset and fly the mission.
“From everything I’ve heard, it takes several days to generate these plans,” Hammett says. STRATCOM officials were not available to verify Hammett’s assessment of the command’s current process. A DIA official says that fulfilling a normal request takes one to two weeks depending on the urgency. Urgent requests can be done within hours or sooner, the official says.
Linear programming would reduce those weeks and hours to minutes and seconds for any request, urgent or not, Hammett says. The optimization model’s objective is to “generate ISR allocation plans that address the greatest number of intelligence requirements possible given the resource constraints on the ISR aircraft assets,” states the GMU paper. “We wanted to provide a mechanism for the people at the combatant commands to enter in requirements that would feed directly into a linear program,” Hammett says.
The GMU group created a spreadsheet-based tool with a graphical user interface to capture asset and priority data and requirements. Combatant commanders can use the tool to input requirements, which allocation planners can then access. Subject matter experts can easily create and update the model’s ISR asset and mission priority definition pages to keep data relevant and effective, Hammett says. The group’s optimization model erases several steps in a traditionally manual effort and minimizes inconsistency, subjectivity and time spent conferring with higher level officials, Hammett says.
Subject matter experts remain engaged in the process by reviewing and validating the program’s solution and making necessary changes. Their involvement is significantly minimized but still necessary. “A model can’t have intuition,” Hammett says of the human contribution. But, while human brains cannot process the voluminous quantities of data associated with the complex ISR asset allocation problem and arrive at an optimal solution, linear programming can, Hammett says. “Even a simple linear model will consistently outperform so-called ‘expert judgment,’” the GMU study states. Loerch concurs: “It can be shown mathematically that you could not do better.”
The GMU effort also tackled an ongoing issue within the ISR community regarding asset-based versus capability- or intelligence-based requests. Traditionally, industry analysts agree, commanders request ISR resources based on the asset rather than the intelligence they need or the capability the asset provides. A commander might request a Predator unmanned aerial vehicle (UAV), for example, rather than an overhead visual of a target or site. “If you ask for intelligence rather than assets, you can get a lot more efficient dissemination of the existing assets,” Hammett says.
The ISR community continues to struggle with the issue, says Dr. John Regner, director of modeling and analysis, Teledyne Brown Engineering,
The culture is changing. The military is orientating itself to a more effects-based model. “Don’t ask us for an airframe, ask us for a specific result,” DIA’s Black says. In some cases, he says, the intelligence-based requests may have been already fulfilled by another asset. The new centralized JFCC-ISR group will help de-conflict those allocation requests. “Let the ISR guys figure out the best solution,” Black adds.
The best solution could come from a tool originally developed to optimize weapon types. The Sensor-Platform Allocation Analysis Tool, a mixed integer optimization model developed by Teledyne Brown Engineering, came from an equivalent capability that was built in the late 1980s to look at weapon and force allocation. “The idea was that if I could do that for weapons-to-target or -target types, I ought to be able to do that for sensor-to-sensor items of interest,” Regner says.
The tool allows an analyst to “determine what mix of sensors, platforms and groundstations will deliver the required capabilities/performance at the minimum cost,” according to a Teledyne Brown paper titled “On Determining An Optimum C4ISR Architecture: the Sensor-Platform Allocation Analysis Tool (SPAAT).” It assists in determining how ISR planners might “mix space-based radar with JointSTARS [Joint Surveillance Target Attack Radar System] and UAVs to do a job,” Regner says. Analysts can set the model to meet the minimum or maximum coverage requirements. A feasibility function alerts analysts when the request cannot be satisfied, allowing them to diagnose the source of the infeasibility, such as an inadequate number of sensors or other operational or budgetary constraints.
Hammett says the SPAAT tool seems to offer a less comprehensive, combatant-command-oriented approach than the GMU model’s global ISR asset allocation solution. “It seemed to do a good job, but it’s very focused,” he says about the SPAAT effort. He called the tool “more developed” than the EAGLE ISR model, and adds that the top-down EAGLE ISR model and the bottom-up SPAAT solution “would complement each other nicely.”
The GMU group ran several evaluations and “what if” scenarios to test its model’s effectiveness, Hammett says. The model included peacetime, humanitarian assistance, major theater war and other scenarios. The peacetime scenario required fewer intelligence-based requests and produced the highest satisfaction ratios, while a global war scheme tested the limits of the model and produced the lowest requirement fulfillment rates, Hammett says. “It gave us exactly what we expected,” he adds.
In other tests, the group introduced a new kind of UAV asset, added a sensor to an existing asset and increased the quantity of the most popular asset to gauge the model’s response to realistic adjustments. The model also measured rates for low, medium and high mission priorities; figured percentages of assets assigned; and displayed surplus, and perhaps less valuable, assets. Of the group’s six designated asset types—three manned and three unmanned—three were used much more frequently than the others.
Some results surprised the model’s designers. For example, they found that the unavailability of a popular UAV asset type caused a higher rate of unfulfilled high-priority missions. When the group increased that asset type and ran the model again, the fulfillment rate increased. The result indicated the asset’s value and tagged it as an attractive acquisition candidate. “You could look in a matter of minutes at what asset is in the greatest demand,” Hammett says. “So if you were going to procure assets, those are the ones you would buy.”
Regner says mathematical optimization models such as SPAAT were originally formulated to evaluate how new technologies might overlap or enhance current capabilities. “You can think of it as technology insertion, but it immediately becomes an acquisition problem,” he adds.
All three operations researchers—Regner, Loerch and Hammett—agree that those charged with assigning global ISR assets should focus on sharing data and acquiring optimization tools that assist decision making. Says Regner, “By re-organizing themselves, STRATCOM has indeed recognized the problem and is trying to have the tools that span that space.”