Researchers set their sights on artificial intelligence to coordinate battlefield sensors

by J.R. Wilson

Washington —From one perspective, it might be called a dynamic database; from another, artificial intelligence (AI).

The latter term is somewhat ironic, in that it involves two distinct — yet here closely intertwined — definitions of "intelligence". One is the plethora of information now available in the digitized battlespace — human intelligence, signals intelligence, sensor intelligence, space intelligence, data transmitted to and from aircraft (manned and unmanned), satellites, land and sea systems. The other is how those data are sifted, sorted, fused, and distributed in a way that makes them useful.

The fact is, today's warfighter gathers and distributes more information on and around the 21st Century battlefield than ever before, and more is added to that tally every day. At the same time, we are developing newer, smaller, faster, cheaper, better ways to get information to everyone in the chain of command, from the President down to the rawest recruit is the forward-most foxhole.

And therein lies the problem: information overload.

Lacing this problem are several different elements of information surfeit and insufficiency, real-time and out-dated — more information than a human can process in the microseconds before the users want it.

Today those responsible for graphical representations of databases boast of their ability to update them in less than 36 hours, maybe less than 24. But for the special operations force about to make a clandestine raid on a hot target, anything more than a few minutes old could mean the difference between success and failure, life and death.

"The Holy Grail is to build a database in six hours and deploy anywhere on any platform," says Sandeep Divekar, president of MultiGen-Paradigm of Addison, Texas. "We aren't there yet for full-scale terrain and 3-D models — large-scale, complicated urban databases."

"Information" comes in all forms in the digitized battlespace. It may be a terrain map created by combining existing surface maps with images from space, then overlaying new infrastructure information from aerial reconnaissance and human intelligence. It also may refer to weapons types, quantities, positions; troop compositions and movements; military engagements; battle damage assessment, etc. Depending on source and timing, such information may be duplicative, complementary or contradictory; even if the latter, all of the data may have been accurate at the time it was acquired.

According to the Joint Chiefs of Staff's Joint Vision 2020, "advances in information capabilities are proceeding so rapidly that there is a risk of outstripping our ability to capture ideas, formulate operational concepts and develop the capacity to assess results. While the goal of achieving information superiority will not change, the nature, scope and 'rules' of the quest are changing radically.

"The joint force of 2020 will use superior information and knowledge to achieve decision superiority, to support advanced command and control capabilities and to reach the full potential of dominant maneuver, precision engagement, full dimensional protection and focused logistics. The breadth and pace of this evolution demands flexibility and a readiness to innovate."

In point of fact, some of those elements and all of the associated problems already are present, as demonstrated in Afghanistan, where the combination of overwhelming information and overwhelming long-range firepower accomplished in a matter of weeks what the former Soviet Union's conventional Red Army was unable to do in a decade, albeit the goals of the two military campaigns were markedly different.

But there were failures, some of which led to almost all of the U.S. military combat deaths — from "friendly fire" — in the Afghan pursuit of the Taliban and al-Qaeda.

"Who cares if you have situational awareness if you can't communicate it," notes Dr. Paul Kolodzy, program manager for the Defense Advanced Research Projects Agency's (DARPA) Small Unit Operations: Situational Awareness System (SUOSAS).

Research to find solutions to this problem is ongoing on several fronts, many of them within DARPA. For example, a new generation of "software agents" is being developed to automatically accept abstract tasking, get needed information, decide how to solve simple problems, help the user solve difficult problems and take action on the user's behalf.

"However, research and development are needed in the control of agent-based systems to mitigate dangerous and chaotic behaviors such as resource consumption, faulty communication, poor performance, system shutdowns and security vulnerabilities," notes Lt. Cmdr. Dylan Schmorrow of DARPA's Information Technology Office, where he is program manager for Control of Agent-Based Systems (CoABS).

"The CoABS program will develop and evaluate several control strategies that will allow military commanders and planners to automate relevant command-and-control functions, such as information gathering and filtering, mission planning and execution monitoring, and information system protection," Schmorrow says. "Through the effective control of agent systems, the intelligent agents will work in harmony to strengthen significantly military capability by reducing planning time, automating and protecting command and control (C2) functions and enhancing decision-making."

The goal of CoABS is to build taskable software robots that cut the amount of time warfighters spend manipulating information systems by a factor of ten, rather than focusing on the mission. In addition, information that otherwise might be overlooked has greater likelihood of being incorporated into the overall picture.

For example, "cooperating agents, by correlating reach-back and on-ship data, can exploit currently unused data to focus surveillance assets for more effective TBMD (theater ballistic missile defense)," says Dr. Ken Whitebread of Lockheed Martin Advanced Technology Laboratory Laboratories in Camden, N.J. Whitebread made his comments last February in a presentation on "Transitioning the Military to Agent-Based Computing" at the CoABS Transition Exhibition.

As long ago as 1996, five Air Force officers predicted U.S. armed forces in that timeframe would require an advanced information operations system that will generate "products and services that are timely, reliable, relevant and tailored to each user's needs. The products, which the officers outlined in a research paper on "Wisdom Warfare For 2025," must come from systems that are secure, redundant, survivable, transportable, adaptable, deception-resistant, capable of fusing vast amounts of data and capable of forecasting.

"The ability to fuse vast amounts of data from the multitude of sensors, automatically sort it, identify the essential pieces of information, and provide the right information to the right node in near real time is the goal," the officers wrote. "This represents one of the greatest challenges. The best system will be able to identify the relevant databases across dissimilar networks, search through and filter vast amounts of stored information and rapidly analyze and correlate data across distributed databases with thousands or millions of variables. The architecture must automatically maintain current information on designated target sets at all times and assist in targeting by presenting vulnerability, aim points and strike options. This process must remain effective even when incomplete or uncertain data are part of the underlying situation."

The obvious question now is: just how real is artificial intelligence? The hype has been high for years about what is to come, what is almost here, and what tremendous things AI will do, once we actually have it. But is AI real or is it just another form of vaporware?

The answer, according to researchers at Pathfinder Systems Inc. (Lakewood, Colo.), is ... mixed.

"An extreme case of these hopes, in our sphere of interest, is recognition of the evolving battle scene and directly generating doctrinally correct plans, operations orders, or frag orders from this scene recognition," according to a Pathfinder analysis of the current state and future prospects of AI. "While human 'command agents' indeed appear to work this way, we believe there is very little chance of achieving this in the near future, either by neural nets or by 'classical AI' methods.

"The fact is that the processing and storage power available in today's computers is wholly inadequate to approximate a 'neural net' model of the human brain's functions," the Pathfinder analysis goes on. "The human brain uses around 10 billion times less power per 'instruction' than modern processors; it has perhaps 100 to 300 trillion bytes of equivalent 'main RAM memory' and it learns about 1 million to 1 billion times faster than our best neural net or other software models." In contrast, a human child seems to generate approximately 1 million new neural connections per second — or perhaps 1 megabyte of new code per second.

Pathfinder forecasts the hardware necessary for human-like AI, if the current explosive growth in computing technology continues, evolving as follows:

• if only memory were the limiting factor, we would have AI around 2050;
• if heat/instructions were the limiting factor, we would have AI around 2070; or
• with learning were the limiting factor, we could not have AI before 2160.

Of course, researchers can do much in the realm of sorting, identifying, processing, and fusing crucial data — tasks that fall far short of human-level intelligence. It may be 2160 before AI reaches a level where it alone can make battlefield decisions — a prospect that human commanders probably will never welcome. But the ability to provide accurate, crucial data in real-time in a form those same human commanders can use to help them make battlefield decisions themselves should be much closer to reality.

In a paper on "Intelligent Mobile Agents in the Military Domain", scientists at Lockheed Martin agree that much remains to be done, including meeting a variety of technical requirements to prove and support widespread transition of agent technology into the military arena.

"Robust agent behavior control mechanisms must be developed and emplaced; the military must be convinced that agents are controllable tools rather than dangerous (in terms of security and bandwidth) and uncontrollable viruses. Conversely, resource control mechanisms must be instituted because agents have greater potential than human operators to overload legacy resources," according to the Lockheed Martin paper.

"For their part, military network designers must carefully define data and legacy-system access and release policies so that agents are not in danger of inadvertently disseminating sensitive information. From an agent technology standpoint, perhaps the most outstanding need is for a standard agent capability and data description language — a semantic framework that supports collaboration across many heterogeneous agent systems. DARPA's Control of Agent Based Systems program has put in place a JINI-based software infrastructure to support agent control and collaboration."

JINI is not an acronym, but the name of a Java network technology of Sun Micro Systems.

Kolodzy's program at DARPA is to devise an integrated mobile communication system with fast data throughput, optimized for restrictive terrain and using several different waveforms with self-configuring networking. According to DARPA, the top technical challenges they must meet before having a system suitable for field use by the Army, Marines, and Special Operations Forces are:

• low probability of intercept/low probability of detection (LPI/LPD) waveforms;
• |mobile ad hoc networking; and
• position and navigation in GPS-denied areas.

"SUOSAS was developed because individual warfighters are going to be disbursed in the future, as noted by a Defense Science Board report back in 1996," Kolodzy says. "So a lot of situational awareness information must be sent out to the individual warfighter. Some people refer to it as 'bits over iron', meaning bits sometimes have more power than iron. But it is hard to get a radio onto an individual warfighter and make sure no one else can hear him."

There is another problem — beyond security — that will restrict how information is disseminated and to whom: All that data must pass through a very small communications "pipe".

"So we determine what people should get with respect to their location, mission, or hierarchy," Kolodzy says. "Someone higher up in the hierarchy should have more information available than the individual warfighter. A battalion commander would have up to 1,000 square kilometers to watch, while the individual only has a small area, perhaps 1 km, around him. That reduces our data flow by a lot. We have a processing engine that figures out where you are and gets you only the information you need."

That is based on a data fusion or aggregation engine developed by Alphatech of Burlington, Mass., and SRI International in Menlo Park, Calif., that looks for redundancies. If three soldiers see the same tank, the data fusion engine will determine if it is the same tank and will only send out one message. If they have different reports, it will compromise, perhaps noting there are reports of one to three tanks at such and such location. If a platoon in a particular location needs information known to a nearby tank unit, that information would be sent to the group as a single message, rather than sending individual messages to each warfighter.

Alphatech's dynamic database effort involves an all source track and identity fusion (ATIF) capability to improve ground vehicle tracking and identification by fusing moving target identification (MTI) radar data with imagery intelligence (IMINT) and signals intelligence (SIGINT) data.

"We also have sensor fields out there and you don't want to send every single sensor report across the network. So those go through gateways that aggregate them into as few reports as possible. We are averaging a reduction of 90 percent," Kolodzy says. "If you have a field of unattended ground sensors, say 20 along a road, and a tank passes down that road, 18 of those sensors may detect the tank, with varying levels of confidence. The situational awareness engine would take those 18 reports and fuse them into a single report or possibly two, depending on the confidence level of the incoming reports."

The SUOSAS engine is similar to and comes from the same roots as the multitarget aircraft tracker, but must be much more tolerant because of the severe clutter of information in a ground environment. The processing also is distributed across several different nodes — instead of one processing engine getting all the information from all sensors, data feeds into multiple locations and through a distributed processing environment. In that way, if one sensor or even a node is lost, the SUOSAS engine could still fuse all the information received from surviving multiple nodes.

One fundamental requirement for SUOSAS is size — the current demonstration system weighs about 30 pounds, but DARPA experts are working toward a handheld version within the next three years. Because it has to work within the soldier's radio, they are using a very lower power general-purpose processor running Linux.

"One box does everything. The situational awareness systems processing is all truly embedded with the communications system, because the intelligence you need has to trade off with the size of the pipeline available," Kolodzy says.

Size, power, security and the ability to sort, process and discriminate information to the right people in the right place at the right time are difficult enough goals, but in order to truly do its job, the system must be fully automatic, requiring no direct action by any human being at any point.

Kolodzy compares it to the controversial Napster Internet-based system that allowed individuals to share copyrighted music: "All the data was distributed among users, each of whom had their own database of music. With our system, everybody has a little information that might be relevant to somebody else. So if someone goes into a new area, they can go out to everyone else and ask if any information is available. The most recent and best information will be returned, based on a comparison of databases by the processors.

"The soldier doesn't have to take any action at all. The system automatically updates itself as the warfighter moves from one location to another or as the environment around him changes. There are some automatic systems that will file information into the network, like bio/chem detectors. If a soldier uses a laser rangefinder, however, the soldier will need to make some manual input indicating what he's lasing - tank, building, etc. — but that's as far as he needs to go; the information will then be disseminated to whoever needs to know it."

Anyone going into a battle theater would be deployed with a basic database structure, which systems such as SUOSAS would then populate and update in real time. Information would go up the chain from individual soldiers to squad leaders, where it would be fused, then from the squad level on up unfused. Information coming back down the pipeline would be filtered for the best use of each level.

The system is designed to work according to where the recipients are, what their job is and whether the information is pertinent at any given moment. To do that, it takes into account command and control lessons and rules already built into software for the Army by SRI.

"We now have found a way to put those rules into algorithms. So if you have the rules, we know what to do with them and put them into software; others are better equipped to develop the rules and those may need to be changed from time to time, which is up to the warfighters and planners. Both the rules and the software can be modified in the battlespace," Kolodzy says.

SUOSAS is completing its third and final phase at DARPA with the development of a prototype to be tested in the field in August through October 2002. If all goes well there, it probably would be transitioned to the Army Communications Electronics Command (CECOM) for the final engineering and development phase.

"If all goes well, we're looking at about a three-year development program to make it small, then another two years or so before it would go into mass production and out onto the battlefield — say 2007," Kolodzy says, noting nothing coming from DARPA is likely to move into operational use at a faster pace. "DARPA tries to prevent technical surprise by working with high risk, high payoff technologies. If something were to go directly from DARPA to the field, it would mean we weren't stretching the technology far enough."