The Wars of the Virtual Worlds

October 18, 2004

Every few months, a continent crackles into life on linked supercomputers. It is a place of huge cities and surrounding countryside with an intricate road network on which tens and even hundreds of thousands of trucks, tanks, mopeds, pedestrians and other vehicles move.

The newly created world is an electronic arena in which top military officers of the U.S. Joint Forces Command can develop tactics for the future an arena that has taken a significant leap in complexity in the past three years, facilitated by the computer skills supplied by the University of Southern California Information Sciences Institute.

Using techniques developed over decades of the use of high performance supercomputers in academic research in the physical sciences, the ISI-JFCOM team was able to make the simulation software scalable, i.e. not artificially limited to a small number of simulated participants.

The ISI scientists collaborated with Caltech and the technical personnel who were active in the JFCOM effort, many of whom are employees of such companies as Lockheed Martin, Alion Sciences, SAIC, and Toyon.

One experiment in this series, "Urban Resolve," began October 12. Set in the year 2015, in JFCOM's description, "it involves a U.S.-led coalition force that must confront and overcome a skilled adversary who is equipped with modern capabilities and is operating in an urban environment."

Two groups of officers, the blue team leaders of the coalition, and the red team leaders of the adversary, control their forces from separate command posts, rooms full of monitors at which specially trained aides enter make the moves ordered by commanders.

These aides are called "puckers" - a holdover from the days when military exercises were conducted on huge tables or floor areas, and soldiers pushed token ships or tanks, ("pucks") around with sticks.

Two other computer control rooms complete the set up. A green team controls the "clutter"- vehicles, pedestrians, and other facets of the civilian population, not part of the forces of either side. Finally, a white room for the experiment umpires synthesizes a combined view of operation - a so-called "angel's eye view."

Puckers for green, red and blue teams add vehicles to the world by selecting them from a menu of thousands pre- written units of software code, each describing the behavior of a specific vehicle - taxi, tank (not just generic, but specific model), city bus, that have been created over the course of more than a decade.

Some of the vehicles have very complex behavior sets, but even the simple 'bots "know" how fast to go on which roads, to turn corners, to avoid collisions and to stay on the roads, as well as being time-sensitive, e.g. they know to crowd the roads during morning and afternoon rush hours.

A select few - most of them combat units - are far more complex, endowed with artificial intelligence that can respond and react to changing circumstances in complex ways.

Simple or complex, the population of the arena world used to be much smaller.

"For a long time," says ISI project director Dan Davis, a marine veteran who has turned his combat-zone experience to good use in the virtual world, "there was an unacceptable ceiling for the number of the vehicles that could be simulated on individual workstations on a local network - they couldn't get much above about 30,000."

Dr. Robert Lucas, director of the computational sciences division at ISI, led an effort that definitively broke this barrier, with one major event coming in 2002. That event was the record for SemiAutomated Forces so far: 1 million entities. "Now," says Davis, "while we are hesitant to say just exactly where the final limits may be, we see no immediate constraints on the delivered scalability."

The three-part experiment now taking place will use about 100,000 entities, employing 128-node Linux cluster supercomputers located at two military supercomputing centers, ASC in Ohio and MHPCC in Maui. Both Linux clusters were provided to JFCOM by the Defense High Performance Computing Modernization Program (HPCMP).

The Joint Experimentation Directorate (J9) directs the exercise from Suffolk, Virginia. Controllers and analysts also work at the U.S. Army Topographic Engineering Center at Fort Belvoir, Virginia, and at the Space and Naval Warfare Systems Command facilities in San Diego, California.

ISI computer scientists Dr. Ke-Thia Yao, Gene Wagenbreth, Brian Barrett and John Tran also participated in the effort. Tran, with degrees from Notre Dame University, joined the team two years ago, and may have a good opportunity to see some of the issues he has been simulating taking place in real life, as he has just received orders to reserve duty for six months. The JFCOM experiments are designed to help the U.S. better understand how to engage in combat in urban settings.

"Phase I will focus on using human intelligence," reads the Urban Resolve description, "along with advanced intelligence, surveillance, and reconnaissance (ISR) technologies, to gain comprehensive situational awareness and situational understanding of the urban environment and the adversary forces."

"In ... Phase II, the friendly force will continue to employ leading-edge ISR capabilities to find and track the adversary. In Phase III the U.S.-led coalition will employ a fully equipped, combined or joint task force with modern air, land, sea, and space capabilities to maneuver effectively in the urban battlespace."

The current experiments, which are scheduled to be completed October 22, are part of a continuing program.

"The outcomes of this joint and multinational experiment will expand our understanding of future urban conflict, from pre- crisis to post-conflict, while providing insights into today's urban warfighting challenges. Without the use of high performance computers, these experiments would not have been conducted at the scale necessary to reflect the realities of 21st century urban environments," the JFCom description concludes.


Pucker-eye view: a segment of the urban battlefield, with (left hand side) control toolbar. Note the green arrow at 10 o'clock inside the yellow circle, just above the roadway on the far side of the canal. (Click on image to enlarge)


The view from the green arrow viewpoint, The vehicles seen as dots on the display above resolve into individual cars, motorcycles and pedestrians. (Click on image to enlarge)