Rozenblit, a professor of Electrical and Computer Engineering at the University of Arizona in Tucson, is developing 3-D decision support software to help field commanders visualize the battlespace for combined land, sea and air operations.

"We call it symbolic visualization because our purpose is to create new visualization concepts that are different from high-resolution displays that create photo-like simulations," Rozenblit says.

"Here our approach is the opposite of realistic-looking, high-resolution displays. We are using a set of abstract symbols that are quite simple in terms of how they are depicted, yet have rich semantics behind them."

A battalion, for instance, might be represented by a cube, which has five usable sides, the sixth side being its connection to the underlying terrain. The cube represents the unit, but each side could carry information about such things as group strength, fuel supply, fatigue or training.

Similarly, terrain is shown symbolically. Depicting terrain with topographic accuracy isn't necessary and can actually become confusing, says Liana Suantak, a Ph.D. student who is working on the project with Rozenblit.

Avenues of approach, choke points -­ such as mountain passes -­ and lines of defensible terrain don't have to be represented with a high level of detail.

In addition to symbols that represent terrain and units, the simulation uses graphs to display other data. Instead of giving commanders columns of figures, the simulation displays bar graphs that show data at a glance, such as the relative strengths and weaknesses of units.

The result is a simulation that can be configured rapidly and run on a laptop computer. It gives a commander a compact, uncluttered depiction of the situation that's simple and easily grasped, but rich in the information it provides.

The researchers couple this display with genetic algorithms to not only give battlefield commanders a clear view of the situation, but scenarios for addressing it.

To search every possible course of action would take forever, of course. So the idea is to find a method that avoids unpromising strategies, Suantak says.

Genetic algorithms look at what's called the "fitness function" of possible solutions and cross the most promising ones, in much the same way that a geneticist might cross fruit flies to develop certain physical attributes.

"You cross them and come up with an offspring that's similar to the parents, but still different," Suantak explains.

By running through 50 or 100 generations, the genetic algorithms are designed to produce a strong solution. Fewer generations will produce an answer faster, she adds. There's always a tradeoff between speed and the robustness of an answer.

The algorithms also consider mutations in which genes are flipped and the answer is quite different from the original parent solutions.

The symbolic visualization software is well adapted to stability and support operations such as refugee relief, peacekeeping and similar operations, such as supporting voting processes when establishing a new democratic government, Rozenblit says.

"Traditionally, there has been no good way to represent situations in cities, where ethnic strife may be a problem. We can represent a lot of shades, where it is not really clear who the factions support and where the allegiances are shifting."

The software has civilian applications as well, particularly in disaster relief.

"It can be used to represent situations where you need to evacuate people because of flooding, hurricanes or loss of power," Rozenblit says. "It allows agencies such as FEMA to get a snapshot of what's going on. And it supports decision making. Where do you send supplies first? Which route do you take? And so on."

Rozenblit has been working on this symbolic visualization software for about six years and now has a product that the U.S. Army will be using for training and assessment.

The work has been funded by the U.S. Army Research Laboratories. - By Ed Stiles

[Contact: Jerzy Rozenblit, Ed Stiles]

22-Jan-2002