A robot is nothing but a baby that needs to be trained to handle objects, this is what researchers at Cornell University Personal Robotics Lab believe. The team led by Ashutosh Saxena is working on making robots understand the objects around them.
This they believe will help them handle the objects according to their understanding to help them get adapted to the environment. They are trying to make the robots understand the nature of objects and their utilization like infants learn on their own.
Learning to Place New Objects (http://pr.cs.cornell.edu/placingobjects)
An Indian Institute of Technology-Kanpur alumnus, Ashutosh Saxena devised the experiment where the robot is trained to look at cups of all shapes and sizes and see what is common to them all. A similar process is to show the robot how to find the handle of the cup and hold it properly.
Also they must learn how to place object correctly. The robot can pick up objects with more ease than they can place them down. For instance Saxena saw that the robot was able to place the cup upright on a table but when it went to the dishwasher the robot placed the cup upside down.
The research hopes to train the robots to make the right decisions on their own. Saxena said that they just show the robot some examples and it learns to generalize the placing strategies and applies them to objects that were not seen before.
The researcher had found that the robots were capable of understanding the simple logic behind things and were found to be performing the task accurately 98% times, if they already knew about the object. If the object was new to robots, they still performed 95% times in a correct manner.
It learns about stability and other criteria for good placing for plates and cups, and when it sees a new object — a bowl — it applies them, said Saxena.