14 DAY TRIAL // Investigators at the Arizona State University (ASU) are currently at the very edge of scientific research, conducting work that will ultimately result in the development of robots that are fully autonomous.
Even if some of today's most advanced robots may give the impression that they operate on their own, in all cases there is an operator somewhere. This is true for the most versatile of the machine.
Then there are those robots that only have very limited tasks, such as for example welding machines on assembly lines. These instruments cannot exit predefined operation patterns, and therefore cannot operate autonomously.
Even Predator drones and other unmanned robotic vehicles are not off the hook, so to speak. “Unmanned vehicles are not really autonomous. There is a pilot somewhere, making them operate,” says ASU expert Srikanth Saripalli.
He holds an appointment as an assistant professor in the School of Earth and Space Exploration at the university, where he is working on clearing the connection between man and machine.
The researcher says that there are two main obstacles that roboticians need to surpass in order to achieve the elusive goal of creating an autonomous machine.
First off, they need to develop a system that always tells the robot where it is. Next, a software program needs to be able to make instant decisions of how to use that information in planning future actions.
According to Saripalli, achieving simultaneous localization and mapping (SLAM) is very complex, mostly because of the difficulty of making machines understand what they see in video feeds.
“The biggest problem is that vision is a really rich sense, and while humans do a lot of the processing automatically, computers really don’t know how to incorporate all that data into something meaningful,” Saripalli reveals.
Researchers are currently also working on developing a way to emulate the behavior of creatures in the natural world, such as for instance ants. The level of cooperation these insects show is remarkable.
“One thing robots don’t do well is respond to unpredictable or changing conditions,” says ASU School of Life Sciences assistant professor of biology Stephen Pratt.
The researcher collaborates with colleagues from the University of Pennsylvania and the Georgia Institute of Technology (Georgia Tech) on innovating robotics.
“Ants are good at recruiting groups of two to 20 and working cooperatively to move large objects over rough terrain,” the expert concludes.