When future generations boldly go where no man has gone before, they will likely receive a boost from Ed Luke.
An associate professor in computer science, Luke has spent nearly 15 years developing and expanding Loci, a high performance computing framework that helps engineers build predictive models. Most recently, he has been awarded part of a $2.2 million grant to help NASA design a spacecraft capable of carrying humans deeper into space.
Announced in January, the new project will use Loci-CHEM, a fluid modeling expansion of the basic framework, to study the turbulent mixing of chemicals that occurs when a space-bound vehicle is launched.
“Launches involve a violent mixing of materials, which can be very hard to accurately model,” Luke said. “Our work will focus on developing algorithms to make that modeling possible.”
Luke developed Loci while at Mississippi State earning a doctoral degree in a program that combines engineering, applied mathematics and computer science. He said it was this multidisciplinary training that helped him develop the tool.
“Sometimes, if you have an engineer, a computer scientist and a mathematician in the same room, they can be talking about the same thing and not realize it because they aren’t speaking the same language,” Luke said. “Each of these areas is rich with computational problems and solutions and, by combining knowledge from each, you can get tools like Loci.”
Deriving its name from locus, the mathematical concept on which it’s based, the Loci framework provides a way to input data so that the computer can assemble it to create a model and provide proof that the result is logically consistent.
“It’s a logic-based system. You give it a lot of different information—sort of mathematical relations—and it finds the model that satisfies them all,” Luke said. “Basically, it’s a code for building codes, so that the user doesn’t have to worry about the high performance computing issues or complex computational problems.”
Loci uses principles from artificial intelligence, which allow it to assemble models using logical inferences from two things. First, it maintains a rule database that describes how to make inferences about model variables. Second, it maintains a database of facts known by the user. The system can inspect those facts and rules to assemble the model or simulation that satisfies a goal requested by the user.
The system is designed in such a way to not accept ambiguity. Luke explained that, sometimes unknown to the user, similar programs can produce results that should not be trusted. Loci, however, will simply not run if it detects logical inconsistencies.
“It wants every ‘i’ dotted and every ‘t’ crossed and if that doesn’t happen, it tells the user it cannot provide a meaningful answer,” Luke said. “That makes it much harder to make an accidental mistake.”
While initial funding for Loci came from the National Science Foundation, the project has received millions of dollars in grants over the years. The additional funding helps ensure continued advancement that includes expansions like Loci-CHEM and Loci-Blast, a system currently in development to model explosions such as those from roadside bombs.
“As time has gone by the range of uses has expanded,” Luke said. “A lot of NASA contractors have started using it, as has the Air Force and other government units.”
He continued, “Boeing actually used one of our Loci systems to evaluate the re-entry risks associated with a damaged tile on the space shuttle. They actually took pictures of the damage while it was in orbit and used Loci to construct a model of the damaged tile to simulate the effects of the damage on re-entry heating. This result was used to assess other models and help in the process of deciding if the shuttle would be able to return to Earth safely.”
As the list of companies, organizations and projects that have benefited from Loci keeps growing, Luke said he is confident more and more engineers will explore the potential uses of the program. He explained that it is distributed under an open source license to make it widely available and encourage application of the system to a more diverse range of problems.
“Having a tool like Loci is only part of the solution,” Luke said. “Like any computer simulation program, engineers have to know how to apply it and trust that its results are accurate. And as people continue to use it, it could become as ubiquitous as the older engineering methods we rely on today.”