Wine Tasting Robots From Japan

The Japanese have a national love affair with robots of all kinds. One of the coolest things I ever waited on line for five hours to see was the Tokyo Robot Show, where along with tens of thousands of cameraphone toting young Japanese I got to see the debut of Sony's Qrio and Honda's Asimo bipedal robots.

While perhaps not so glamorous as a robot that can walk around, play soccer, and dance, the latest robotic technology out of Japan is perhaps more practical. NEC corporation has developed a robotic sensing apparatus that is capable of analyzing wine samples accurately enough to determine the varietal and the place of origin. While this sort of technology isn't entirely new, the method used by this robot is somewhat unique. Unfortunately this thing isn't cool enough to slurp it up and then spit it out again, but it does use laser beams!

OK. Maybe they're only infrared LED beams. But that's still pretty cool, isn't it?

Anyhow, these LEDs reflect light off the wine sample and then spectrographic analysis is done on the reflected light to analyze the sample. The applications of this system, which can identify about thirty different variations in flavor profile, are limited given the hundreds of thousands of wines out there in the world. However, this technology proves the possibility of developing sensors to help with specific incidents of fraud that are more common in the wine world.

I'm sure Domaine Romanee Conti has already placed an order, or will shortly, as they seem to be a particular target for fraudulent bottles, given the astronomical prices they command at auction.

A robotic wine taster, capable of distinguishing between 30 different varieties or blends of grape, has been developed by engineers in Japan.

The idea is to automate wine analysis so that retailers and customs officials can easily check that a wine is indeed what its label declares.

The wine-bot was developed by scientists from NEC's System Technologies laboratory and Mie University, both in Japan. It is about twice the size of a 3-litre wine box and consists of a microcomputer and an optical sensing instrument.

For analysis, a 5 millilitre sample of wine is poured into a tray in front of the machine. Light emitting diodes then fire infrared light at the sample and the reflected light is sensed by an array of photodiodes.

By identifying the wavelengths of infrared light that have been absorbed by the sample, NEC says the wine-bot can correctly identify the unique organic components of 30 popular wines within 30 seconds.

Grape fraud

Because the combinations of these components are unique to certain wine-making regions, NEC says the wine-bot can even tell where the wine came from. The company promises to extend the number of wines the device can recognise before it is commercialised, but has not revealed when this will happen.

John Corbet-Milward, head of technical and international affairs at the UK's Wine and Spirit Trade Association, says deliberate mislabelling of wine is a serious problem.

"There always has been such fraud going on," the told New Scientist. "If there has been a bad grape-growing year, for instance, it's quite tempting for people to make up a blend of wines that's difficult to spot."

Currently, fraud detection is performed through human tasting and careful analysis of a vineyard's records. "Any new machine that can identify wine fraud is potentially very interesting if it can speed up detection processes and reduce costs," Corbet-Milward adds.

Refined palate

But he also notes that a wine-bot would need to be capable of distinguishing between more than 30 flavours of wine to be truly useful. This is because the global wine market includes thousands of wine varieties and blends. "The acid test would be to test a machine against a team of qualified tasters," Corbet-Milward says.

He is not alone in having reservations. "There have been many attempts to do this but the calibration that is involved is truly enormous," says Geoff Taylor, managing director of Corkwise, a company that analyses wine to assess its quality.

"You have to teach the sensor all the parameters of all varieties and blends of wine. Then you have to include all the subtle differences caused by the use of different yeasts and maturation environments," Taylor adds. "It's a massively complex picture."