What if robots could find out more on their own and share that knowledge with each other?
Many of the jobs humans would like robots to perform, such as packing items in warehouses, assisting bedridden patients, or aiding soldiers on the front lines, aren’t yet possible because robots still don’t recognize and easily handle common objects. People generally have no trouble folding socks or picking up water glasses, because we’ve gone through “a big data collection process” called childhood, says Stefanie Tellex, a computer science professor at Brown University. For robots to do the same types of routine tasks, they also need access to reams of data on how to grasp and manipulate objects. Where does that data come from? Typically it has come from painstaking programming. But ideally, robots could get some information from each other. That's the theory behind Tellex's "Million Object Challenge". The goal is for research robots around the world to learn how to detect and handle simple elements from bowls to bananas, upload their data to the cloud, and allow other robots to analyze and use information.
Robot
Tellex’s lab in Providence, Rhode Island, has the air of a playful preschool. On the day I visit, a Baxter robot, an industrial machine produced by Rethink Robotics, stands among oversized blocks, scanning a small hairbrush. It moves its right arm noisily back and forth above the object, taking multiple pictures with its camera and measuring depth with an infrared sensor. Then, with its two-pronged gripper, it tries different grasps that might allow it to lift the brush. Once it has the object in the air, it shakes it to make sure the grip is secure. If so, the robot has learned how to pick up one more thing.
The robot can work all day, often with a different object in each of its grippers. Tellex and his graduate student John Oberlin have gathered-and are now sharing-data on about 200 articles, beginning with such things as child's shoe, plastic boat, rubber duck, garlic press, and other kitchen utensils , And a sippy cup that originally belonged to her three-year-old son. Other scientists can provide data for their own robots, and Tellex hopes that together they will build a library of information on how robots must handle a million different articles. Eventually, robots facing a crowded shelf will be able to "identify the pencil in front of them and pick it up," says Tellex.
Projects like this are possible because many research robots use the same standard framework for programming, known as ROS. Once a machine learns a certain task, it can pass the data to others, and those machines can load comments that in turn refine the instructions given to later machines. Tellex says data on how to recognize and understand any given object can be compressed to only five to 10 megabytes, about the size of a song in your music library.
Tellex was an early partner in a project called RoboBrain, which showed how a robot could learn from someone else's experience. His collaborator Ashutosh Saxena, then at Cornell, taught his robot PR2 to lift small cups and place them on a table. Then in Brown, Tellex downloaded that information from the cloud and used it to train his physically different Baxter to do the same task in a different environment.
Such progress may seem incremental now, but over the next five to 10 years, we can expect to see "an explosion in the capacity of robots," says Saxena, now CEO of a startup called Brain of Things. As more researchers contribute and refine cloud-based knowledge, he says, "robots must have access to all the information they need, at their fingertips."
The robot can work all day, often with a different object in each of its grippers. Tellex and his graduate student John Oberlin have gathered-and are now sharing-data on about 200 articles, beginning with such things as child's shoe, plastic boat, rubber duck, garlic press, and other kitchen utensils , And a sippy cup that originally belonged to her three-year-old son. Other scientists can provide data for their own robots, and Tellex hopes that together they will build a library of information on how robots must handle a million different articles. Eventually, robots facing a crowded shelf will be able to "identify the pencil in front of them and pick it up," says Tellex.
Projects like this are possible because many research robots use the same standard framework for programming, known as ROS. Once a machine learns a certain task, it can pass the data to others, and those machines can load comments that in turn refine the instructions given to later machines. Tellex says data on how to recognize and understand any given object can be compressed to only five to 10 megabytes, about the size of a song in your music library.
Tellex was an early partner in a project called RoboBrain, which showed how a robot could learn from someone else's experience. His collaborator Ashutosh Saxena, then at Cornell, taught his robot PR2 to lift small cups and place them on a table. Then in Brown, Tellex downloaded that information from the cloud and used it to train his physically different Baxter to do the same task in a different environment.
Such progress may seem incremental now, but over the next five to 10 years, we can expect to see "an explosion in the capacity of robots," says Saxena, now CEO of a startup called Brain of Things. As more researchers contribute and refine cloud-based knowledge, he says, "robots must have access to all the information they need, at their fingertips."
0 comments:
Post a Comment