R2-D2, C-3PO move beyond science fiction in Israeli industry

JERUSALEM -- Driverless trains, automated packing plants and mechanized fruit-pickers are marvels of modern technology, but according to robotics scientists, they're destined to take their place way down the robot evolutionary scale.

If the roboticists succeed, today's automatons will simply be the primitive forebears of machines that operate in unknown, unstructured and changing environments. In short, they could be ancestry to machines that think.

Robotics is a hot arena of research in Israel. For industrially developed nations with acute labor shortages, the creation of intelligent machines prepares for the future.

"Robots may be the only way to keep Israeli agriculture profitable," says Yael Edan, chair of the Paul Ivanier Center for Robotics Research and Production Management at Ben- Gurion University of the Negev.

"They have the potential to raise the quality of fresh produce, lower production costs and reduce the drudgery of manual labor."

She and her colleagues are attempting to make an existing robotic melon-harvester economically viable. They want it not only to pick melons, but also to guarantee that each fruit picked is sweet and tagged to indicate its level of ripeness or the exact date on which it will ripen. And what's more, they want the robot to transplant melons, and selectively spray the weeds that grow around them -- thus spraying only one-tenth of the melon field.

A robot field worker moving along the open furrows is a relatively easy notion to stomach, but what about relying on a robot to drive your car?

This is no future vision; a robo-car is already here, according to Reuven Segev, professor of mechanical engineering in Ben-Gurion University's Autonomous Robotics Laboratory. Embellishing upon standard cars, he and his team have equipped vehicles with video cameras to gather data, enabling them to follow the road. There are also automatic laser sensors that continuously measure the vehicle's distance from different points (such as the car in front of it), and a control system to process this incoming information and command the actuator pedals.

Although the car exists, it isn't in Israeli showrooms just yet.

"It'll be a hard sell convincing people that a car doesn't need a driver," said Segev, pointing out that there will be safety issues, insurance problems and cost factors.

The team anticipates robotic vehicles will first gain public confidence through handling hazardous materials or carrying important loads into dangerous areas.

If people are reluctant to place their lives in the hands of robot drivers, will they be more receptive to robot surgeons?

Yes, says Moshe Shoham, a professor at Technion-Israel Institute of Technology in Haifa. "Especially when they learn that 10 percent of the 200,000 knee replacements annually performed in the United States by human hands must be repeated because of surgical error."

Shoham, whose specialty is mechanical engineering, has developed a robot that specializes in knees and is now being tested. "Orthopedic procedures like this, involving rigid bone, lend themselves to automated surgery," he said.

Shoham's robot performs entirely alone, unlike those in use for hip replacement. From a preoperative CT scan, the surgeon feeds precise information about the operation site into the robot's computer; the unshakable robot hand does the rest. Sensors monitor the "hand" to ensure it exerts a force no greater than that of a human surgeon, and a "joystick" allows the physician to take over from the robot at any time.

"Surgeons from other specialties have approached me about adapting this robot for use in their areas," said Shaham. "I believe it has application in the brain, eye and ear, nose and throat surgery. At the moment, I'm reworking it for spinal surgery, where there's no room for error...a tiny slip of the hand can permanently paralyze a patient."

One key to how well a robot functions is how efficiently it "sees" or senses. A robot's sensors must be capable of both recognizing objects and guiding the robot along changing paths.

At the Weizmann Institute of Science's Moross Laboratory for Vision Research and Robotics in Rehovot, a new and sophisticated computerized vision system is emerging from a collaboration between mathematicians and brain researchers.

The catalyst was the recent discovery that visual perception in the human brain is a two-way process: Incoming visual information flows simultaneously from the perception centers to memory centers and back again at high speed, enabling the brain to compare large numbers of pictures and identify similarities between them.

Weizmann mathematician Shimon Ullman has imitated this process to "teach" robots to see: He has created a system in which robot sensors compare incoming visual data with photographs, so they can not only navigate changing paths, but also recognize different car models and even different expressions on a human face.

Scientific imitation of the human brain is also leading the way to more precise motor control in robots. Professor Tamar Flash of Weizmann's math and computer science department had been comparing motor performance of healthy humans and those suffering from various neurological disorders, when she realized her work had application to robotic systems as well.

"I'd attached sensors to different parts of the human arm and discovered that the brain always chooses the smoothest movement possible," she said. "This taught me that the brain 'thinks' of movement in geometrical form, and constantly creates internal models of smooth mechanical behavior."

What's more, that discovery led Flash to develop algorithms that improve the movement of robotic arms. "The better we understand how the brain plans our own movement, the better equipped we will be to plan the movement of industrial robots."

If the sci-fi nightmare of evil rampaging robots taking over the world remains distant, the dream of a proficient, tireless, uncomplaining, non-unionized workforce is edging perceptibly closer.