Pilotless Chopper Finds Its Own Way

Forget model airplanes that need to be glued together or remote-control toy helicopters. UC Berkeley researchers have invented the real deal.

The BErkeley AeRobot (BEAR) is a small, fully autonomous helicopter capable of navigating and landing independently, requiring human intervention only to provide a destination.

The machine is being developed in the UC Berkeley Electrical Engineering and Computer Sciences (EECS) Department under the direction of Shankar Sastry.

Utilizing an embedded computer, the BEAR is designed to process information provided by the hardware components and to transfer data to and from the physical world, said John Koo, a visiting faculty member working on the project.

Koo described the development of reliable and attack-resistant computer software as the greatest challenge of the project.

To navigate, the BEAR uses Global Positioning Satellite technology to closely pinpoint its location. Landing requires the BEAR to use a camera and computer vision algorithms developed at UC Berkeley.

Improving the vision-based system is essential if the BEAR is to land on a moving or inclined surface, a task described by former group member and project co-founder David Shim as "a carefully controlled crash landing."

"Our goal is to make a robust, fully autonomous landing procedure that will work without human intervention," said graduate student Omid Shakernia.

According to Shim, unmanned aircraft have many advantages, the most notable being their small size, cheap construction, and disposability.

Conceivably, autonomous flying machines will be able to reach dangerous places and perform tasks ranging from power line repair to fire fighting.

While the operation of a typical helicopter is limited by the physiology of its human pilot, unmanned versions can push the envelope of helicopter capabilities.

The computer brain of the BEAR, unlike a human pilot, has a constant, unwavering attention level.

Modern vision technology can also deduce the 3D shape, location, and motion of the object being imaged without becoming tired, said Shakernia.

With a fleet of wirelessly linked aerobots, the possibilities abound. BEAR researchers currently run pursuit-evasion games in which a team of pursuing ground-based and aerial vehicles look for a moving target, called an 'evader.'

The helicopters search in a systematic manner, according to a probability map detailing the chance that the evader is at a specific location. The strategies being developed now could improve future search and rescue operations.

Using multiple autonomous helicopters at once has left the BEAR team wary of midair collisions. To combat this problem, the researchers have developed a vision-based formation flight control system.

"(The aerobots would) detect the relative motion of the formation partners and control their motion in order to keep a tight formation," said Shakernia.

Potential uses for this technology are not limited to unmanned flight, with the researchers seeing the project as an experimental platform for testing various technologies developed at UC Berkeley.

In one theoretical application, cars could be made to organize themselves into platoons to increase highway throughput, solving congestion problems in large cities.

"The exciting aspect of research is that as you develop a technology, more applications become possible than the original application you might have had in mind," said Shakernia.

Tags: