Showing posts with label Learning. Show all posts
Showing posts with label Learning. Show all posts

Friday, September 12, 2014

Robot Boris-2 learning to load a dishwasher - Video



University of Birmingham scientists teach robot to load dishwasher. 

Credit: University of Birmingham

Researchers at the University of Birmingham in the U.K. have set themselves an ambitious goal: programming a robot in such a way as to allow it to collect dishes, cutlery, etc. from a dinner table, and put it in a dishwasher.

While not quite there, the team recently held a demonstration at the British Science Festival meant to show how far along they've progressed.

Professor Jeremy Wyatt with Boris-2, the robotic manipulator

Getting a robot to grab something and pick it up isn't hard anymore, what is difficult is getting a robot to pick up almost any object, regardless of shape, size, texture or delicacy and do something with it.

But, that's the goal of the team at Birmingham, one they've been working towards for five years (and spending £350,000 in the process), they are only just now at a point where they can demonstrate their robot, named Boris-2, picking up several different types of objects.

The team reports that Boris's abilities are part memorization, part processing. The robot was taught how to pick up several objects by manually manipulating its grippers.

After that, to pick up something not seen before, the robot runs though the process of picking up known objects and must attempt to apply its stored knowledge in a new way.

To that end, it must determine which parts of the object to grasp, and how much pressure to exert, no small feat, the team acknowledges.

The real goal of the project is not to build a dishwashing robot, of course, it's to figure out how to build robots that are capable of picking up a variety of objects in a variety of settings, skills that would be useful in factory settings, military operations and perhaps someday, in dining rooms and kitchens.

Such skills, the team notes, mean moving past second generation robots (those that can operate in a relatively non-changing human environment, e.g. drones, self-driving cars).

The next generation will have to be able to adapt to changes made to the environment by humans, or others such as pets or other robots.

They'll need to be able to work or perform duties alongside humans or in some cases, when they are not around at all.

Tuesday, February 18, 2014

NASA Robonaut 2: Humanoid Learning Medical Skills for Space Emergencies - Video



NASA is training Robonaut 2, a humanoid space robot to pull double duty as an emergency doctor in space — a surrogate physician that could one day be controlled by experts on Earth to help sick or injured astronauts.

The $2.5 million Robonaut 2, nicknamed R2, is designed to work alongside the astronauts and even take over some of their more tedious duties inside and outside the International Space Station.

The new NASA training is adding telemedicine skills to that mix.

In a new video of Robonaut 2's telemedicine training, the automaton performed an ultrasound scan on a mannequin and even used a syringe like it would to administer a real-life injection.

The tests were performed using a ground-based version of R2 robot, the mechanical twin of the one currently aboard the space station.

Zsolt Garami
"I would say that within an hour I trained him more than with other students I'm working for a week, so I think that he's learning really fast," Dr. Zsolt Garami, of the Houston Methodist Research Institute, says in the video.

Far from earthly hospitals, astronauts who currently live on space station, typically in six-month-long stints, must be trained in basic surgery and medical procedures in case of an emergency.

But Robonaut 2, which has a camera-equipped head, could administer care to spaceflyers, controlled by doctors on the ground.

So far, tests with Robonaut 2 have shown that human controllers can perform tasks "correctly and efficiently by using R2's dexterity to apply the appropriate level of force and can track their progress using R2's vision system," NASA officials explained in a video description.

Garami said the robot might eventually be able to learn to do some tasks on its own.

NASA's Robonaut 2 and Dr. Zsolt Garami, of the Houston Methodist Research Institute, practice telemedicine techniques during a test of the robot's bedside manner. 

This image is a still from a NASA video documenting the Robonaut 2 test. 

Credit: NASA

Thursday, February 28, 2013

Sleep Reinforces Learning: Children’s Brains Transform Subconsciously

During sleep, our brains store what we have learned during the day a process even more effective in children than in adults. 

Credit: Monkey Business / Fotolia

During sleep, our brains store what we have learned during the day ‒ a process even more effective in children than in adults, new research shows.

It is important for children to get enough sleep. Children's brains transform subconsciously learned material into active knowledge while they sleep -- even more effectively than adult brains do, according to a study by Dr. Ines Wilhelm of the University of Tübingen's Institute for Medical Psychology and Behavioral Neurobiology. Dr Wilhelm and her Swiss and German colleagues have published their results in Nature Neuroscience.

Studies of adults have shown that sleeping after learning supports the long-term storage of the material learned, says Dr Wilhelm. During sleep, memory is turned into a form that makes future learning easier; implicit knowledge becomes explicit and therefore becomes more easily transferred to other areas.

Children sleep longer and deeper, and they must take on enormous amounts of information every day. In the current study, the researchers examined the ability to form explicit knowledge via an implicitly-learned motor task.

Children between 8 and 11, and young adults, learned to guess the predetermined series of actions -- without being aware of the existence of the series itself. Following a night of sleep or a day awake, the subjects' memories were tested.

The result: after a night's sleep, both age groups could remember a larger number of elements from the row of numbers than those who had remained awake in the interim. And the children were much better at it than the adults.

"In children, much more efficient explicit knowledge is generated during sleep from a previously learned implicit task, says Wilhelm. And the children's extraordinary ability is linked with the large amount of deep sleep they get at night.

"The formation of explicit knowledge appears to be a very specific ability of childhood sleep, since children typically benefit as much or less than adults from sleep when it comes to other types of memory tasks."

Monday, December 10, 2012

Brain pacemaker delivers constant Deep-Brain Stimulation (DBS) for Alzheimer’s patients


Credit Image: Functional Neuromodulation

Once implanted into the brain, a pacemaker-like device delivering electrical stimulation could help improve the memory of Alzheimer’s patients. Technology Review reports.

Using electrodes, Deep-Brain Stimulation (DBS) is already used to treat patients with Parkinson’s, epilepsy, and obsessive-compulsive disorder.

In this Alzheimer’s trial, co-chaired by Constantine Lyketsos of Johns Hopkins, the device was placed into a region of the brain involved in learning and memory.

In Alzheimer’s patients, brain tissue atrophies and the reduction of memory and thinking skills increase over time.

According to the Johns Hopkins team, electrical shocks could stimulate critical neural networks disrupted by Alzheimer’s.

Recent trials for potential Alzheimer’s drugs have failed to halt or stave off cognitive decline.

Now, in a pilot study with these deep brain stimulators, after one year of constant stimulation, brain scans of six Alzheimer’s patients showed signs of increased neuron activity in areas involving learning and memory, although, it could be unlikely to actually reverse the effects or damage to the brain caused by Alzheimer’s disease.

The researchers are recruiting patients into the new trial initiated by Toronto-based Functional Neuromodulation. The trial will track patients who have the device for a year using doctor observations and brain scans.

[Via Technology Review]