Who's driving? Robots aid soldiers, but can be irritating.

The prevailing thinking in the Army about deploying unmanned combat vehicles is that they keep soldiers out of harm's way.

That premise continues to fuel research into dozens of military programs seeking to develop robotic systems that could be sent into combat without human operators.

Feedback from soldiers who participated in a recent Army experiment at Fort Knox, Ky., however, challenges some of the conventional wisdom associated with the use of robots in combat.

The six-week experiment--involving two unmanned Stryker light armored vehicles outfitted with advanced sensors--was intended to measure soldier workload and performance while operating robots. It also shed light on what soldiers like and don't like about robots.

Their observations are leading researchers to believe that robotic technology has the potential to become a huge assist in combat operations, but that too much automation may not be desirable.

As part of a reconnaissance mission during the experiment at Fort Knox, two soldiers rode in the back compartment of one of the two unmanned Strykers. They were not controlling or driving the vehicle, which was on autopilot mode and had been programmed to follow a particular route.

"It's what we call a manned robot," says Patrick Nunez, a scientist at the Army Tank-Automotive Research Development and Engineering Center, who oversaw the experiment.

"This is one of the first times we've done an experiment with a manned vehicle on autopilot," he says. "The soldiers don't have a direct view out of the vehicle. They have cameras that see."

The other Stryker served as a "follower" robot and did not have a crew inside.

The lead vehicle was charged with locating enemy forces downrange and assessing their strength. The soldiers inside the robot's crew station relied on digital maps and data to familiarize themselves with the local area. They also received intelligence from their command center.

The experiment intentionally was designed to overload the soldiers' senses and overwhelm them. While they were busy planning their route, red icons--indicating enemy combatants launching rocket-propelled grenades--popped up on their intelligence terminals. "We measured their ability to react and plan a reconnaissance of the red force," Nunez says. "They needed to find a route to sneak up and look at the red force. They needed to think about the geography, topography and figure out the best route."

When multiple targets were spotted, the crew station's automated technology prioritized the enemies and pointed to the soldier which weapon would be most effective to defeat the threat.

The two soldiers did the work that normally would require a four-person crew. The Army is banking on robotic technology to help reduce manpower, says Nunez. The goal is tied to a broader Army initiative--under the Future Combat Systems program--to shrink the size of a heavy brigade by about 10 percent, or 326 soldiers.

The sophisticated software installed aboard the Stryker robot is what creates "crew aiding behaviors," Nunez says. It automates common tasks such as route planning and the selection of observation and fire positions. The experiment showed that the software "significantly reduced engagement time and improved the way the soldier controls the robot."

Defying expectations, the soldiers trusted the automation software to get the job done, he adds.

But they weren't as keen about the autopilot.

"They liked the freedom of not driving but didn't like not being in control," Nunez says. Like most people, they weren't comfortable driving on cruise control, and found it irritating.

The upshot for robot developers is that "you need to have a human in the loop with the robotic control," he says. But it would be counterproductive to give the soldier more control over the robot and, in the process, increase his workload, because the point of having a robot is to reduce the burden.

Current technology does not provide a happy medium--somewhere between 100 percent in control or 100 percent on autopilot.

"We'd like to have a 'supervised driver,'" says Nunez. "We need to find a way that even in autonomous mode, the soldier still has a sense of control and is able to tell the robot what to do."

The notion that robots help keep soldiers safe, away from enemy tire, also is becoming outdated as current conflicts attest that the battlefield is "non-linear," and troops have to be prepared to respond to attacks anywhere. In the experiment, the scouts inside the robot were operating in a hostile environment populated by simulated RPG throwers.

If a situation was deemed too dangerous for the crew, the operators inside one robot could choose to send the other crewless robot forward, by itself.

In a broad sense, the experiment was successful, according to Nunez. But he cautioned that the technology still has a ways to go before it is ready for real-world combat.

Officials at Fort Knox are evaluating the results of the experiment, particularly the effects of automation, Nunez says. "The jury is still out."

The Army wants to apply the lessons from this and other experiments as it designs the Future Combat Systems family of advanced vehicles. While tank platoons today have four vehicles, an FCS reconnaissance platoon would have three armored vehicles--two manned and one robotic.

The robotic vehicles would be smaller than a Stryker, but bigger than a Humvee.