During the height of the wars in Iraq and Afghanistan, thousands of robots zipped over rugged terrain and zoomed through battered walls and doorways.
In most cases, as one military officer stood guard, the mission's operator would sit nearby, with a joystick in hand, controlling a small, motorised robot that would gather intelligence and help advance covert operations.
To some, it was sort of like playing a videogame. The difference, though, was that there was no "pause" button or break in the action. "For that reason, whenever you had an operator deployed, you would have a guard to keep the operator safe, because he has no idea what's going on around him," says Henrik Christensen, the chairman of robotics at the college of computing, as well as the director of the centre for robotics and intelligent machines at the Georgia Institute of Technology in the United States.
Now, robotics engineers, scientists and computing experts like Mr Christenen and his team are jumping in to further help mission operators by giving them control of robots from farther away and more detailed analysis before those robots roll into danger zones, both in wars abroad and during domestic disasters or terrorism situations.
They are also tackling another challenge: finding ways to offer robotic operators short breaks so they do not have to remain stressfully glued to their screens throughout the entirety of a mission.
"You'd like to get to a place where they're not exhausted after one of these missions," says Mr Christensen. "So you'd like to have a degree of autonomy, where they could move small distances and say: 'go to this doorway or follow this hallway to the end' and the robot would do that autonomously, and while it drives, the operator could take a small break."
To get to this point, experts developing military robotics are making some models smaller - palm-sized, even - while others are making bigger ones, flying machines capable of surveying a semi-demolished building then relaying the perilous path a tiny device might need to navigate to gather secret data.
Research and development efforts behind these kinds of unmanned vehicles designed for use on the ground, in the air or even under water are increasing around the world.
The global market dedicated specifically to robots that will be used by the military on the ground is expected to grow from US$4.5 billion this year to $12bn by 2019, according to a report published last month by WinterGreen Research.
Regardless of their final destination, these sophisticated machines share a common aim: to assist or replace people caught in battlefield or other dangerous scenarios. So far, more than 50 countries are in the process of creating or acquiring robotic systems for their military programmes and some are already using these technologies, according to market data from ABI Research.
Many governments are turning to this technology to reduce - or, ideally, prevent - casualties during military operations. But some are also seeking to cut the amount they spend on defence and, analysts say, the costliest robotic systems are still less expensive than vehicles operated by people.
"I think it'll be enormously important in the future," says Susan Eustis, the president of WinterGreen Research. "All the defence companies are in it because of the robots used in Iraq and Afghanistan and as they pull out the market is going to increase because of the rise of terrorism."
Back in 1958, the US defence advanced research projects agency (Darpa) was created to help prevent strategic surprise attacks while still maintaining the advanced technological presence of the country's military. As part of its mission, the agency established a challenge to US congress that mandated one-third of all military land vehicles to become autonomous by 2015.
In April, Darpa awarded a $1 million prize to a team of individuals who created the best "fast adaptable next-generation ground vehicle." While this armoured machine may never make it to mass production with its exact design, it is an example of how even agencies such as Darpa are pushing to innovate unmanned vehicles.
All told, the US army is planning to upgrade around 2,700 robot systems for training or deployment purposes, while another 2,400 robots will be divested and shared with department of defence partners and other agencies, WinterGreen Research says.
University researchers, including some who have partnered with government agencies or private companies, are behind some of the robots that are currently being deployed in this space.
Boston Dynamics, for one, is an engineering company that spun off from the Massachusetts Institute of Technology. It developed BigDog, among other robots, which weighs 240 pounds and climbs, runs and carries heavy loads over rough terrain. Originally funded by Darpa, the machine only hits speeds of up to 4 miles per hour but can scale slopes as steep as 35 degrees and lifts up to about 154 kilograms, Boston Dynamics says.
At Georgia Tech, Mr Christensen is working on a series of projects, including a multi-year, ongoing programme led by BAE Systems, a global defence and security products company. This particular one combines the use of unmanned aerial vehicles that relay to ground-based models where they should travel. The goal is to get the robots to within 100 metres of a target site, without raising suspicion.
"This is typically where you have the most casualties," says Mr Christensten. "In the past, we've primarily been working on the ground vehicles. Now we're looking at how do you do the combination of aerial and ground [robots] to provide a complete solution. You want to make sure it's very safe because very few military commanders want to get this deployed until it's not exposing the soldiers to an increased level of danger."
The technology is currently being tested at a facility within Georgia, at a site with a mix of regular urban buildings as well as some that have been partially destroyed. While the different models of Mr Christensen's robots, as well as those across this sector, vary considerably in shape and size, many units these days are combining software as well as robotic platforms.
More models are also employing the use of sensors and batteries. In other words, some of these robots share their ancestry with another technology: mobile phones.
"We have platforms where we're using cell phone technology," says Mr Christensen. "One of the platforms we're using right now has a CPU [central processing unit] and camera on board and the cost is $80," he adds.
"Today, the cell phone revolution has really brought the cost down."
Indeed, researchers are trying to address all sorts of challenges that may hamper growth in this field.
Battery life for these robots typically only lasts between one and six hours and some experts have only just begun working on more advanced features, including "wall-penetrating radar" that would permit soldiers to "see" through buildings without having to enter them.
Mr Christensen says mission operators and soldiers are already stressed enough within their environment, so it is crucial to minimise the amount of training they would need to operate one of these robots - hence the videogame-like design for control systems.
It turns out that a US soldier has played an average of 10,000 hours of computer games.
"So if you can use that experience as a skill set then it's much easier for them to pick up [a robot] these days", says Mr Christensen.
But with the length of time it takes to develop these robots, cost remains the biggest challenge for this market's growth.
"We need to get these down so they're cheap enough that you can deploy them without having to worry about the overall cost," says Mr Christensen. If smartphones continue to develop and sell the way they are, that may not be an issue in the future.