Web Concepts

Posts Tagged ‘human-robot interaction’

Excuse me? Hello?

Simon the robot, developed at Georgia Tech's Socially Intelligent Machines lab, uses a camera to figure out when a person is interested in interacting with it. It watches their behavior and uses some form of gesture, like waving its hand, to get noticed.

Then it has to figure out whether it actually captured the person’s attention. Using only the camera as a guide, Simon achieved about 80 percent accuracy in determining when he had gotten someone’s attention, according to Aaron Bobick, professor and chair of the School of Interactive Computing at Georgia Tech's College of Computing.

Check out the video below in which a volunteer constructs something out of colorful blocks, under Simon’s watchful eyes. Simon’s timid wave interrupts his concentration, and he waves back.

The goal is to help robots understand when to sit and be quiet and when it’s appropriate for them to get a human’s attention. People can figure this out pretty easily with simple body language, but it’s not as obvious for a robot. Remember Furby and its insistent, insipid desire to play, even when you didn’t feel like it? Simon is apparently a little more polite.

“In order for these robots to work with us effectively, they have to obey these same social conventions, which means they have to be able to perceive the same things humans perceive in determining how to abide by those conventions,” Bobick explains.

Other roboticists at Georgia Tech are helping robots learn whether they want to deceive other robots or humans, and then carry out a deceptive strategy to that effect. No word on whether these technologies would ever be combined, so robots could figure out when we humans aren’t paying attention and then carry out their devious world-dominating plans.

[International Business Times]

Safety fears have been a major obstacle keeping the personal robotics revolution from fully blooming, as it can be dangerous for humans to work alongside large, strong robots that could potentially injure them. After all, robots don’t know pain and have trouble assessing dangerous situations in which their actions might harm a human being. In industrial settings, this often keeps robots cordoned off in their own spaces where they can’t accidentally hurt someone.

So Festo set out to create an arm that is both strong and dexterous as well as safe for humans working alongside it. Using 3-D printing technology, the engineers (and their partners at the Fraunhofer Institute) created lightweight and soft trunk segments that are powered by pneumatic artificial muscles that drive some serious force underneath. The trunk segments are littered with resistance sensors that help it to be aware of contact with people and objects around it. A three-fingered gripper at the end is also designed to need little force to grasp a range of objects.

The result is a multifunctional gripper ‘bot that could be applied to a range of industrial applications, as an extra set of hands on the lab bench, or even as an assistant in the operating room. And since it can interact with other Bionic Handling Assistants to pass objects back and forth, multiple robo-trunks could be deployed in arrays that assist both humans and each other. No word from Festo on whether it will work for peanuts.

[New Scientist]

Using a powerful Japanese-built industrial robot, researchers at the University of Ljubljana are conducting the study on their own arms. Using two different tools – one blunt and one sharper – the robot struck each participant a total of 18 times at different impact energies. The subjects were then asked to rank the degree of pain on a scale ranging from “painless” to “unbearable.”

By doing so, the researchers hope to arm future robots with a sense of what kinds of motions and impacts cause varying degrees of human pain. As robots become more uniquitous, it’s inevitable that they’ll be interacting with human more often and in closer quarters. The robot in the study was actually programmed to move toward a certain point in space beyond the human arm, pushing the arm aside as it moved, the idea being that a robot and human working in close proximity will eventually come into contact and the robot needs to be aware of its own strength and the human’s tolerance for it so no one meets a mechanically violent end.

Most of the subjects described the robots jabs as mildly to moderately painful, so it seems the robot is playing nicely enough. And while it might seem silly or even simplistic to sink research dollars on a people-punching robot, the field of human-robot interaction is actually quite challenging. Israeli researchers are offering a cash prize for the best robotic hand capable of performing a handshake, a task that sounds simple but in reality is seriously complex. If we’re going to live side-by-side with our future robot companions, we’ll first have to teach them how to be a little more like us.

[New Scientist]

The game pairs up two human players, one playing the role of a human astronaut, the other a robot. The team must then collect the proper tools and supplies from around their space station environment to fix their faulty generator; one player can't beat the clock on his or her own, so the two must interact to complete all the given tasks in the allotted time.

By studying that human teamwork and social interaction, the group hopes to better hone the autonomous behaviors of its robot Nexi, a humanoid 'bot that MIT is developing to explore better means of social learning and human-robot interaction. The team will then recreate the Mars Escape game environment and display what Nexi has learned from it at the Boston Museum of Science.

Check out the game here. As long as you're playing games at work, you might as well be aiding the pursuit of better robotic intelligence. Just be careful not to overdo it; the more you play, the closer you get to losing your job, and the closer Nexi gets to replacing you.

[MIT Personal Robotics Group via Hizook]