- The Robot, Developed At Cornell University, Uses Kinect Sensors, 3D Cameras And A Database Of Videos To Work Out What Its Owner Wants
- In Tests, The Robot Correctly Anticipated Its Owner's Needs 82% Of The Time..
A beer-pouring robot that can read your body movements and anticipate when you want another drink has been developed by American students.
Researchers from Cornell University used Microsoft Kinect sensors and 3D cameras to help the robot analyse its surroundings and identify its owner's needs.
The robot then uses a database of videos showing 120 various household tasks to identify nearby objects, generate a set of possible outcomes and choose which action it should take - without being told.
As the actions continue, the robot can constantly update and refine its predictions.
As well as fetching drinks for thirsty owners, the robot can also work out when its owner is hungry and put food in a microwave, tidy up, make cereal, fetch a toothbrush and toothpaste, open fridge doors and more.
 |
The Cornell robot uses sensors and a 3D camera to analyse the depth of its surroundings (left). The view seen by the robot in the right-hand picture shows how it anticipates its owner's actions. It compares the actions against a database of household task videos and chooses what it thinks is the most appropriate response. The more actions the robot carries out, the more accurate its decisions become. |
The view seen by the robot in the above picture shows how it anticipates its owner's actions.
The Cornell robot uses sensors and a 3D camera to analyse the depth of its surroundings.
Ashutosh Saxena, Cornell's professor of computer science and co-author of a new study tied to the research: 'We extract the general principles of how people behave.
'Drinking coffee is a big activity, but there are several parts to it'.The robot builds a 'vocabulary' of such small parts that it can put together in various ways to recognise a variety of big activities.'
'Even though humans are predictable, they are only predictable part of the time,' Saxena said.
'The future would be to figure out how the robot plans its action.
Right now we are almost hard-coding the responses, but there should be a way for the robot to learn how to respond.'
Saxena and Cornell graduate student Hema S. Koppula will they present their research at the June International Conference of Machine Learning in Atlanta.
They will also demonstrate the robot at the Robotics: Science and Systems conference in Berlin, Germany, also in June.
VIDEO: Could this be the future? Robot learns how to pour you a beer