The bright minds over at Fujitsu Other obstacles discovered while testing included the robot's inability to access individual items and navigation handling.
"Right now, the robot will guide you to the Colgate toothpaste shelf, but the user will have to grope for an individual toothpaste box," Kulyukin says. "Obviously, there is a chance that a wrong box can be picked."
The team is attempting to address the problem by integrating a small bar-code reader, which is already used by most grocery stores.
"What we anticipate is that once the robotic shopping assistant gets the user to a shelf with a bunch of individual items, the user will then use a miniature bar-code reader, essentially a pencil, to read the bar code on the shelf until the bar code of the right item is found," he says.But at least one organization wants to take a wait-and-see approach before getting too excited.Paul Schroeder, vice president of programs and policies for the American Foundation for the Blind, calls the research admirable but says the promise of technology doesn't always pan out in the real world.
"For this device to be helpful, they must conquer problems that would arise in an imperfect environment," Schroeder says.The $5,000 device is not ready to market or mass-produce, Kulyukin says. "But we would like to eventually see the Robotic Guide in other environments, like airports."The researchers built RG using existing technology. The work is funded by a $500,000 grant from the National Science Foundation and two grants from Utah State. Have rolled out an updated version of their HOAP (Humanoid for Open Architecture Platform) robot.
The HOAP-3 is a follow on to the HOAP-2 and HOAP-1, which was originally launched in 2001. After four years of development with Fujitsu's Automation and Laboratory divisions, the latest robot can now process and interpret sounds and visual images and has various sensors that let the robot produce high-level kinetic expressions.The HOAP-3 measures 23.6 inches tall and weighs in at nearly 20 pounds. The mobile robot runs on the RTLinux operating system over a 1.1GHz Intel Pentium M processor with 802.11g wireless capabilities. That means you can either tether the robot to the network or control it over a local wireless network.Fujitsu said it plans to ship 100 units over the next three years, targeting researchers in robotics and communication studies.
Other humanoid robots on the market include the mass-produced Sony QRIO and Honda Asimo as well as lesser-known models like the black chassis Chroino robot and the Repliee Q1, which looks like a human woman.
The robot hand is remotely controlled by means of a haptic "glove", in which each finger is connected to a motion-sensing device. The operator's hand movements are then measured and sent via a computer to the artificial hand, which almost instantly mimics the operator's moves.
The robotic grasper also measures the consistency of objects in its grasp by means of feedback to its motors. And this tactile information is fee back to the mechanical glove giving the wearer an artificial sensation of touch. The arm also incorporates an ultrasound sensor as well as three video cameras, to give the expert a good view of the procedure.
"Having the capability of ultrasound and palpation simultaneously is a major advantage," adds Ranjan Mukherjee, another member of the Michigan team. "Often the ultrasound and exam are done separately. But if the physician can look at the image and feel what he or she is seeing, that’s another huge advantage."
Mukherjee says the next stage is to conduct clinical trials, which should show how reliable the system is. But he believes it could be five years before the system becomes commercially available.
It is a realistic thing to try to do, says computer interaction expert Stephen Brewster, at the University of Glasgow, UK. Lots of telesurgery work is for the military so it is nice to see something coming into the standard medical domain.
But Brewster adds that the system would need to match a doctors bedside manner when dealing with patients. There are of course issues with powerful robots coming into direct contact with people,he told New Scientist. It would have to be very carefully designed and tested to make sure that it could not exert inappropriate forces and so cause harm.
But some experts are more doubtful that such a system would work in practice. "Tele-operated robots have a big problem with delay," says William Harwin, at the University of Reading in the UK, and the sensitivity of touch is not as good.
He suggests a better way to perform remote examinations is to perform the data gathering and expert analysis one after the other. "A more realistic model in principal is to gather data from the remote site, under direction from the clinical base," he explains, and then transfer it the remote expert, who could then use virtual reality tools to interpret it.