The Berkeley Aerial Robot (BEAR) project passed a significant milestone earlier this month, when a 130-pound model of a helicopter successfully guided itself through a course that included random obstacles that weren't on its internal map--a first, according to the university.
The project, funded in part by the Defense Advanced Research Projects Agency (DARPA), is part of a larger effort to create robots that can get to places too dangerous or difficult for humans to go.
John Deere and iRobot, for instance, are working on an autonomous ground vehicle that will be able to bring supplies to soldiers at the front lines. Next year, the U.S. Army will deploy a robot car with a machine gun that will drive itself--but a human will be in control of the gun.
While the military is sponsoring much of the research, advocates assert that these vehicles will also be used to deliver medical supplies, phones or food to individuals stranded by an environmental disaster or trapped in a mine shaft.
The recently conducted test was performed to test the system's laser obstacle avoidance system and see how well it would perform at low altitudes in an "urban canyon" environment. In one run, the helicopter guided itself through a grid of 10-foot-high tents. The tents were on the helicopter's internal map, but the machine had to figure out how to avoid them.
In a second run, the location of the tents was not included on the flight path. The laser system feeds information on the new structures to the helicopter's navigation system, which then takes actions to avoid the problem.
Last year, BEAR researchers flew two helicopters at each other in a game of chicken. "They flew toward each other, sensed each other and adjusted their course," said a UC Berkeley spokeswoman.
Experiments are being conducted at various institutions with autonomous fixed-wing planes and helicopters. Each has their advantages and disadvantages. Helicopters can take off from a stationary position. However, because of the environments in which they would be used, the navigation systems need to be more precise.
"Fixed-wing UAVs (unmanned aerial vehicles) can tolerate less precision in control algorithms because they can glide for a time if something goes wrong," Shankar Sastry, UC Berkeley professor of electrical engineering, said in a prepared statement. "In contrast, helicopters are inherently unstable, so the control inputs need to be exact, and they need to come quickly to keep the helicopter from falling from the sky."
While the obstacle avoidance system tested this month relies on lasers, researchers will start to dedicate more energy to computer vision systems. In these, sensors feed digital images to onboard computers, which then, through probability and artificial intelligence, try to chart a safe course. Computer vision, while more difficult to achieve, potentially will work better because it can provide information about the structure or character of a landing surface.
The BEAR project was founded in 1996 and began to obtain DARPA funding a few years later.
