Monkeys can control a robot arm as naturally as their own limbs using only brain signals, a pioneering experiment has shown. The macaque monkeys could reach and grasp with the same precision as their own hand.
"It's just as if they have a representation of a third arm," says project leader Miguel Nicolelis, at Duke University in Durham, North Carolina. Experts believe the experiment's success bodes well for future devices for humans that are controlled solely by thought.
One such type of device is a neurally-controlled prosthetic - a brain-controlled false limb. Nicolelis says his team's work is important because it has shown that prosthetics can only deliver precision movements if multiple parts of the brain are monitored and visual feedback is provided.
Gerald Loeb, a biomedical engineer at the University of Southern California in Los Angeles, says the new experiment already has some parallels in everyday life. For example, he says, when you drive a car it becomes an extension of your body.
But Nicolelis says the monkeys appeared to be treating the robot arm as their limb, not an extension. "The properties of the robot were being assimilated as if they were a property of the animal's own body."
Arm waving
The core of the new work is the neuronal model created by the researchers. This translates the brain signals from the monkey into movements of the robot arm. It was developed by monitoring normal brain and muscle activity as the monkey moved its own arms.
The task involved using a joystick to move a cursor on a computer screen. While the monkey was doing this, readings were taken from a few hundred neurons in the frontal and parietal regions of the brain. The activation of the biceps and wrist muscles was monitored, as was the velocity of the arms and the force of the grip.
Once the neuronal model had developed an accurate level of prediction the researchers switched the control of the cursor from the joystick to the robotic arm, which in turn was controlled by the monkey's brain signals. At first the monkeys continued moving their own arms whilst carrying out the task, but in time they learned this was no longer necessary and stopped doing so (see Flash animation.
For Nicolelis, the end goal is to help people with paralysis by bypassing brain lesions or damaged parts of the spine. Initially patients would control robotic aids, such as a mechanical arm attached to a wheelchair.
But eventually the signals could be used to stimulate the nerves controlling a patient's own muscles. Nicolelis and his team have already begun to testing this approach on people, but he says it is too early to discuss this research.
Journal reference: Public Library of Sciences Biology (Vol.1, Issue 2, p.1).
Monkeys can control a robot arm as naturally as their own limbs using only brain signals, a pioneering experiment has shown. The macaque monkeys could reach and grasp with the same precision as their own hand.
"It's just as if they have a representation of a third arm," says project leader Miguel Nicolelis, at Duke University in Durham, North Carolina. Experts believe the experiment's success bodes well for future devices for humans that are controlled solely by thought.
One such type of device is a neurally-controlled prosthetic - a brain-controlled false limb. Nicolelis says his team's work is important because it has shown that prosthetics can only deliver precision movements if multiple parts of the brain are monitored and visual feedback is provided.
Gerald Loeb, a biomedical engineer at the University of Southern California in Los Angeles, says the new experiment already has some parallels in everyday life. For example, he says, when you drive a car it becomes an extension of your body.
But Nicolelis says the monkeys appeared to be treating the robot arm as their limb, not an extension. "The properties of the robot were being assimilated as if they were a property of the animal's own body."
Arm waving
The core of the new work is the neuronal model created by the researchers. This translates the brain signals from the monkey into movements of the robot arm. It was developed by monitoring normal brain and muscle activity as the monkey moved its own arms.
The task involved using a joystick to move a cursor on a computer screen. While the monkey was doing this, readings were taken from a few hundred neurons in the frontal and parietal regions of the brain. The activation of the biceps and wrist muscles was monitored, as was the velocity of the arms and the force of the grip.
Once the neuronal model had developed an accurate level of prediction the researchers switched the control of the cursor from the joystick to the robotic arm, which in turn was controlled by the monkey's brain signals. At first the monkeys continued moving their own arms whilst carrying out the task, but in time they learned this was no longer necessary and stopped doing so (see Flash animation.
For Nicolelis, the end goal is to help people with paralysis by bypassing brain lesions or damaged parts of the spine. Initially patients would control robotic aids, such as a mechanical arm attached to a wheelchair.
But eventually the signals could be used to stimulate the nerves controlling a patient's own muscles. Nicolelis and his team have already begun to testing this approach on people, but he says it is too early to discuss this research.
Journal reference: Public Library of Sciences Biology (Vol.1, Issue 2, p.1).
Brain implants have been used to "read the minds" of monkeys to predict what they are about to do and even how enthusiastic they are about doing it.
It is the first time such high level cognitive brain signals have been decoded and could ultimately lead to more natural thought-activated prosthetic devices for people with paralysis, says Richard Andersen project leader at the California Institute of Technology, in Pasadena, US.
By decoding the signals from 96 electrodes in a region of the brain just above the ear – called the parietal cortex - the researchers were able to predict 67 per cent of the time where in their visual field trained monkeys were planning to reach.
They also found that this accuracy could be improved to about 88 per cent when the monkeys expected a reward for carrying out the task.
The team were even able to predict what sort of reward the monkeys were expecting - whether it was juice or just plain water – from their brain signals.
"In the future you could apply this cognitive approach to language areas of the brain," says Andersen. By doing so it may be possible to decode the words someone was thinking, he says.
'Reach region'
Previous research by Miguel Nicolelis at Duke University in Durham, North Carolina, has shown how electrodes implanted in the motor cortices of monkeys can be used to control a robot arm. But this involved recording signals used to control muscles to move the monkey's arm.
The new findings could in theory make this simpler by allowing, say, a paralysed patient to merely specify which object to reach for, and let the robot worry about how it gets there.
The monkeys were trained to think about a particular point in their visual field before reaching for it while the researchers recorded signals in an area Andersen calls the "reach region".
This area is associated with planning, he says. "It takes information from the sensory system and forms early plans for intention."
Previously it has not been clear whether these signals were cognitive or simply related to where the monkey was looking, says John Chapin, at State University of New York who is carrying out related work using a different part of the brain.
Andersen believes this work shows the signals are cognitive because the monkeys were trained not to move their eyes during their experiments so the signals are not linked directly to sensory input.
Ultimately the only way to be really sure, says Chapin, is to try it on humans.
The work was also carried out with researchers in Canada and Switzerland.
Swarms of independently-minded collaborative robots are no longer the stuff of science fiction - they may soon be patrolling national borders and exploring space
James McLurkin has a novel party trick - he can coax 20 small autonomous wheeled robots to form herds, disperse again, wheel in neat circles, sing a harmonic rendition of the theme from Star Wars, and automatically recharge from a power station.
McLurkin, a postgraduate student at the Massachusetts Institute of Technology, is trying to design robots that will work together and make collective decisions. If he succeeds, swarms of robots could one day be put to work in the home, in space and by the military. "A swarm or a team can collaborate to overcome what a single robot might not be able to do," explains Paolo Gaudiano, who works on swarms at Icosystem in Cambridge, Massachusetts.
Soon teams of up to 40 robots could be employed as border security guards and outside airports. Frontline Robotics in Ottawa, Canada, has installed collaborative software on its vacuum cleaner-sized PC-bots and ...
A battalion of 120 military robots is to be fitted with swarm intelligence software to enable them to mimic the organised behaviour of insects.
The project, which received funding this week from the US Defense Advanced Research Projects Agency (DARPA), is aimed at developing ways to perform missions such as minesweeping and search and rescue with minimum intervention from human operators.
The project is run by US software company Icosystems, which specialises in creating programs that mimic behaviours found in nature. Their software will use simple rules to co-ordinate complex behaviour among the robots.
"We will be addressing some fundamental questions about control strategies for robotic swarms," says Paolo Gaudiano, vice president of technology for Icosystems.
The robots' behaviour has been modelled in a computer environment by Icosystems but the company will now be able to test different approaches in the real world. The 120 robots were built for the US military by I-Robot, a company co-founded by robotics pioneer Rodney Brooks.
"Pathological configurations"
Swarm intelligence describes the way that complex behaviours can arise from large numbers of individual agents each following very simple rules. For example, ants use the approach to find the most efficient route to a food source.
Individual ants do nothing more than follow the strongest pheromone trail left by other ants. But, by repeated process of trial and error by many ants, the best route to the food is quickly revealed.
Eric Bonabeau, chief scientist for Icosystems, concedes it is possible that some unforeseen circumstance could throw the robots into chaos. This occurs in natural systems when, for example, ants become isolated from their group and end up running around in a circles, following an every stronger trail of pheromone, until they die of exhaustion.
"There may be some pathological configurations and we need to investigate that," he says. "But I think that it applies to virtually every man made system that has to operate in the real world."
Gaudiano notes that a key goal of the project will be to develop measures which can be used to evaluate when and if robotic swarms could usefully be deployed, and which control strategies are best suited to specific missions.
A battalion of 120 military robots is to be fitted with swarm intelligence software to enable them to mimic the organised behaviour of insects.
The project, which received funding this week from the US Defense Advanced Research Projects Agency (DARPA), is aimed at developing ways to perform missions such as minesweeping and search and rescue with minimum intervention from human operators.
The project is run by US software company Icosystems, which specialises in creating programs that mimic behaviours found in nature. Their software will use simple rules to co-ordinate complex behaviour among the robots.
"We will be addressing some fundamental questions about control strategies for robotic swarms," says Paolo Gaudiano, vice president of technology for Icosystems.
The robots' behaviour has been modelled in a computer environment by Icosystems but the company will now be able to test different approaches in the real world. The 120 robots were built for the US military by I-Robot, a company co-founded by robotics pioneer Rodney Brooks.
"Pathological configurations"
Swarm intelligence describes the way that complex behaviours can arise from large numbers of individual agents each following very simple rules. For example, ants use the approach to find the most efficient route to a food source.
Individual ants do nothing more than follow the strongest pheromone trail left by other ants. But, by repeated process of trial and error by many ants, the best route to the food is quickly revealed.
Eric Bonabeau, chief scientist for Icosystems, concedes it is possible that some unforeseen circumstance could throw the robots into chaos. This occurs in natural systems when, for example, ants become isolated from their group and end up running around in a circles, following an every stronger trail of pheromone, until they die of exhaustion.
"There may be some pathological configurations and we need to investigate that," he says. "But I think that it applies to virtually every man made system that has to operate in the real world."
Gaudiano notes that a key goal of the project will be to develop measures which can be used to evaluate when and if robotic swarms could usefully be deployed, and which control strategies are best suited to specific missions.
发消息
加为好友
