Interactive robots to aid
laborers in auto industry
By Chris Baumbauer
Staff
Writer
There is a little office in the Electrical Engineering
annex where wonders are at work. This place is the Robot Vision Laboratory.
The Robot Vision Laboratory exists to help robots
see the world around them and interact with that world. The robots in
the vision lab are being developed, with help from Ford Motor Company,
to interact with the auto industry.
Avi Kak, a professor of electrical engineering
who is in charge of the robot vision lab, said Ford helps guide the
research and will ultimately benefit from it.
"Many of the goals of the lab are decided in joint
consultations," he said. "Ford would gain not only fantastic technology,
but also be able to eliminate jobs that are too painful for the factory
workers — that are difficult or downright dangerous."
According to Guilherme DeSouza, a doctoral student
in electrical and computer engineering, the lab is working on three
projects: line tracking, tire unloading and visual recognition.
Line tracking allows robots to work on a moving
assembly line. Right now, when a robot works on an assembly line, the
line must stop while the robot performs its task, such as painting a
car. What the robot cannot do that may seem simple to humans is to work
on a moving object.
DeSouza said this is because people are able to
see the world and interact with the environment with touch. Robots,
however, do not have the ability to see the world around them, and,
as a result, cannot perform tasks immediately when a situation arises.
The second project is designing a robot that can
unload tires from a truck and place them on a conveyor belt so they
can travel to where they will be attached to vehicles.
Today, workers unload tires from a truck by pulling
an approximately 10-foot tall stack of tires down and loading them on
a conveyor belt. Then the workers must wait to do it all over again.
In this situation, people are waiting up to 10
minutes for the next batch of tires to load and running the risk of
the tires falling on them.
The problem a robot would face in this situation
is that the tires are not loaded in a perfect column and would need
to be loaded onto the conveyor individually, requiring the ability to
recognize what a tire is and where it is.
The third project is to design a robot with the
ability to recognize specific objects in a pile of other objects. For
this to happen, the robot needs to learn what the object looks like
without a drawing of the object already stored in its processor.
At the vision lab, the robot's ability to recognize
shapes and colors is researched using cameras. In the first project,
two sets of cameras are used — one to identify where the vehicle
is with respect to its area and another to position the robot for work.
In the tire example, the robot will be able to
reposition itself to align a screw with a wheel, regardless of where
the wheel is positioned.
The lab is working on a way to allow the robot
to predict where an object will be after it sees the object. This is
because, in the amount of time it takes for the robot to see an object
on a conveyor, that object is gone before the robot has a chance to
use it.
According to Kak, the final product should be ready
in three to four years.
|
