| Vision
Chinese researchers have devised a neural network that enables a computer
vision system to detect faces and hands by skin color, thus enabling it to track
faces and hands as they move against a background of other colors. In tests, it
was able to track hand gestures with 96.25 percent accuracy. The system could be
used to enable human interactions with appliances and robots through gaze and
gesture.
Another approach to recognizing and tracking objects is to detect their
edges. A new silicon retina pre-processes an image, reducing it to a smaller
amount of edge information, then transmits the information optically for further
analysis and storage. The retina could be used in robots, smart sensor systems,
and remote monitoring cameras.
References: Unknown (2003). “Neural Net Tracks
Skin Color.” Technology Research News, September 8; Unknown (2003). “Vision Chip
Shines.” Technology Research News, September 5.
Wall Crawler
A recent robot demonstration for the US Air Force included a production robot
that can crawl up walls and across ceilings on six wheels, with adhesion
provided by a louvered fan that creates a low-pressure region between two
adjacent surfaces. It’s only a matter of time before wall crawlers are at work
cleaning or re-painting hospital walls and changing light bulbs on the ceiling.
Reference: Coburn, James (2003). “Robots display
force-protection prowess.” AFPN, August 12.
Conscious Robot
British researchers have received a grant to build a conscious robot
containing internal models if its “self” and of the world around it. These
models will be the basis from which the robot will build experience. By
experimenting with how its own body reacts with the world around it, the robot
will learn what is beneficial to it and what is not. Whether this amounts to
“consciousness” is as debatable as the concept of consciousness itself, but the
robot might shed some light on the age-old question.
Its visual systems may give clues of consciousness: By looking at how the
robot is representing visual stimuli to itself and how it reacts, researchers
hope to get some idea of what it is “thinking.”
MIT’s Rodney Brooks’ famous Cog robot, already several years old, does
not have the explicit goal of consciousness, but seems similar in some ways to
this project.
References: Jha, Alok (2003). “Will
fact match fiction as scientists start work on thinking robot?” The
Guardian, August 25; Romain, Gabe (2003). “Conscious”
Robot in the Works.” Betterhumans, August 25.
Female Robots Work Smarter
An Australian software firm has found a way of programming robots to
cooperate (hence, “female”) by sharing objectives, goals, and information about
problems they experience. Sometimes they can then work out their own solutions
to problems.
The software has been installed in three manufacturing robots that build
optical fiber components. Instead of executing single tasks sequentially, the
robots choose among tasks according to priority and circumstance. Instead of
setting up an entire production line for a single job, a single line can do a
range of low-cost jobs simultaneously, making small production runs involving
multiple different products feasible and economical.
Reference: Lowe, Sue (2003). “Enter the
female robots.” Sidney Morning Herald, September 9.
Grace: One Smart Robot
“Grace,” the Graduate Robot Attending Conference in Edmonton, was last year’s
winner of a challenge prize for artificial intelligence, having entered the
conference hall, registered at the desk, found the right meeting room, and
delivered a lecture about herself. Grace combines technology and concepts from
five collaborating labs, including Carnegie Mellon (overall hardware and
software architecture) and the Naval Research Lab (speech understanding
software). Other partners from academia, government, and industry helped solve
mobility, vision, and speech issues.
Grace was upgraded for this year’s AI conference in Mexico, and was joined by
George, an almost identical companion. The two are being designed to interact
with one another, and with people, socially.
Reference: Skirble, Rosanne (2003). “Robot
Challenge: Putting Artificial Intelligence to Work.” Voice of America,
August 4.
Linux Swarm
At this year’s LinuxWorld expo, SRI scientists were to extend robotic
cooperation even further, with up to 100 robots cooperatively mapping a series
of rooms and searching out designated objects. The robots’ intelligence came
from Linux and Sun’s Jini technology, with the robots themselves coming
from ActivMedia Robotics. Two types of cooperating robots are involved: one that
is good at mapping; the other at recognizing objects. One maps the territory to
be searched, and passes that on to the search robots, which then look for the
designated objects.
On January 15, 2004, the Centibots, as they are collectively called,
are scheduled to map an actual US military base.
Reference: Hachman, Mark (2003). “Linux-Powered
Robot Swarm Descends On LinuxWorld.” ExtremeTech, August 5.
Hair Transplant Robot
A stereotactic robotic arm under construction will take the pain, stress,
tedium, and labor out of hair transplant procedures. Stereotactics provides
three-dimensional mapping to guide a surgeon. The robot, says its neurosurgeon
developer, is practically a brain surgeon. The procedure of inserting hair
follicles into the scalp “is similar to inserting an electrode into the brain to
treat Parkinson’s disease,” he said. “But instead of one or two electrodes, you
are inserting 1,500 to 2,000 follicular implants.”
The robotic arm is being built by the company whose robots were featured in
the James Bond movie “Die Another Day.” It has two offset video cameras to make
a three-dimensional image of the patient’s head, and its “hands” hold canisters
of 50 hollow needles to extract the cores of the hair follicles. When enough
canisters have been filled, the robot implants the hair using the same needle.
The video cameras monitor the position of the head during the procedure and make
appropriate adjustments based on its stereotactic map.
The inventor claims his robot, when completed, will cut surgery time (often
up to eight hours) in half, and be much easier on the patient.
Reference: Riordan Teresa (2003). “Surgical robots promise
hair-raising innovation.” New York Times, September 15.
Robot Intelligence is in Its Genes
A segmented “snakebot” controlled by a genetic algorithm (GA) continues to
move toward it s goal even if some of its body segments are damaged. The
snakebot is made up of snap-on vertebral modules each containing shape-memory
alloy “muscles.” Switching current on and off to some combination of the muscles
causes them to contract and expand, moving the organism in a desired direction.
If a segment is damaged, the snakebot’s GA figures out a new combination of
muscle movements in the remaining good segments to keep it moving toward its
goal.
When activated by a segment failure, the GA begins to evolve and randomly
mutate a population of 20 digital “chromosomes” — initially random bits that
correspond to a muscle switched on (1) or off (0). The chromosomes that make the
bot move furthest are selected for reproduction in the next generation. After
multiple generations, optimum movement is achieved.
A snakebot with segments deliberately disabled was able to move with “an
ungainly, dragging gait,” but it reached its destination.
Reference: Graham-Rowe, Duncan (2003). “Robot spy can
survive battlefield damage.” New Scientist, August 20.
Robotic Exoskeleton On Sale
A “robot suit” that helps the elderly or disabled walk, get up the stairs or
seat themselves to relax without a chair will go on sale for about US$8,500 in
Japan next Spring. “HAL-3” (Hybrid Assistive Leg) consists of a backpack
containing a computer and batteries, and four motorized actuators attached
around the knees and hip joints. The computer analyzes electrical signals in the
leg muscles to determine the wearer’s intention, and the actuators move the legs
accordingly. A wearer can walk smoothly at about 2.5 mph.
The suit was developed with Japan’s aging population in mind. If it succeeds,
there will clearly be a market for it in the United States and other countries
facing the aging of a boomer generation.
Unknown (2003). “Japanese
ready to launch `Robolegs’ for commercial use.” Taipei Times, August 22.
See also article in the September issue of
HFD about a US military /
MIT project to create an exomuscular uniform.
Today; Gemologists. Tomorrow: You?
A rough gem is placed it into a box attached to a desktop computer. An image
of the diamond appears on-screen with lines showing the optimum cut for this
particular stone. If it is desired to produce several small gems out of one
large stone, the computer will give appropriate guidance.
The computer’s instructions are passed to a robot appliance smaller than a
home sewing machine, which cuts and polishes accordingly. The process takes
about an hour. It would have taken a human gem cutter/polisher all day. Even at
its current cost of $1-million, the robot is a bargain, at least in high-wage
countries; replacing dozens of people by working faster, all day, every day.
Robots could quickly end the careers of diamond cutters and polishers all
over the world. But why stop there? HFD carries plenty of evidence, including in
this issue, that robots are developing the ability to perform surgery
unaided.
Reference: Macklem, Katherine (2003). “Robots
with a keen eye for shape and colour.” Macleans, September 8.
One Million, and Counting
The International Federation of Robotics counted “at least” 757,000
industrial robots in use at the end of 2001, assuming an average working life of
12 years. (If their working life is taken to 15 years, the number would be 1.02
million.) Japan had at least 360,000, Europe 219,000, the US 97,000, Korea
41,000, and Australia 2,900.
IFR predicts the world market for industrial robots will rise from 78,000
units installed in 2001 to just over 104,000 in 2005 — an annual growth of 7.5
percent. It predicts a total of at least 965,000 by the end of 2005, with a lot
of the growth coming from Europe.
Reference: Blake, Vincent (2003). “At
your robotic service.” Australian IT, September 2. |
