Forgot your password?
typodupeerror
Robotics Technology

Monkey's Thoughts Make Robot Walk 146

Posted by samzenpus
from the surrender-now dept.
geekbits writes "For all those who have at one time or another been too lazy to get up off the couch and go to the fridge and get a beer, heat up some pizza, or change the channel when the remote is missing, we may be one step closer to being able to keep our tushes parked just a little while longer. There may also be some slightly more noble implications here. According to an article in The New York Times, in an experiment at Duke University, a 12-pound, 32-inch monkey made a 200-pound, 5-foot humanoid robot walk on a treadmill using only her brain activity. She was in North Carolina, and the robot was in Japan."
This discussion has been archived. No new comments can be posted.

Monkey's Thoughts Make Robot Walk

Comments Filter:
  • monkey business (Score:4, Insightful)

    by seanadams.com (463190) * on Wednesday January 16, 2008 @09:43PM (#22075118) Homepage
    Being able to read the monkey's brain sounds like the only innovation here, not making the robot walk. Reading between the lines, it doesn't sound like the monkey is really controlling the robot in any real sense at all.

    Several things make me question that. One, why is the robot in north Carolina and the monkey in Japan? It's just for show. Nothing of scientific significance is being demonstrated by that. We all know that internet can connect two gizmos across large distances. The experiment could have been conducted much more simply at one location and made no less effective a point (except to clueless investors maybe).

    Secondly, because of the distance, there is a significant delay (TFA says 250ms, about what I might have guessed.) This would seem to preclude the monkey being able to control the robots actuators in any direct sense. I.e. lift thigh, swing lower leg forward, position foot, lower thigh, positioning body over front leg. Walking is a "controlled fall". No way you could issue all those commands 250ms ahead of seeing or feeling their effect. You'd trip and fall.

    So, what is the monkey really doing? I doubt if he is even thinking "left, right, left, right" because even that would be hard to coordinate with so much lag.

    Finally, why is there a damn robot in the first place? Wouldn't it be much easier to have the commands control a computer animation? You could do that in such a way that the model would look much more interesting to the monkey... it could look like another monkey, a giant walking banana, whatever.

    My guess is that they are simply getting a binary command value from the monkey: "walk" or "don't walk". And the whole robot thing is just for effect. I hate to be such a cynic but this looks like showmanship, not science. If that is the case then this is equivalent to the simple video games that have been demonstrated using brain control.

    However, I could certainly imagine that the journalist totally failed to understand the experiment and maybe something important was lost in his explanation of it.

  • Re:monkey business (Score:5, Insightful)

    by Puff_Of_Hot_Air (995689) on Wednesday January 16, 2008 @10:13PM (#22075476)
    The article is not claiming that the Monkey is directly controling actuators. The monkey is walking, and the signals from it's brain (several hundred neurons), are being used to control the walking motion of the robot. Obviously the processing of monkey neurons response -> robot control is being performed in software specifically tuned to this one monkey with implants. The interesting thing about this experiment is not that they trained a monkey to walk a robot (they didn't), but rather that the monkey was able to keep the robot walking after it had stopped moving itself. This means that all those neurons that the researches are triggering the walking motions from are still going when the monkey thinks about walking but doesn't actually do it. We've known the brain is capable of this for some time, but this is the first experiment I have seen that appears to involve more than a simple "go stop" form of response. If they have tapped enough neurons to control all the actuators required for a robot to walk, then this is news indeed.
  • by megaditto (982598) on Wednesday January 16, 2008 @10:22PM (#22075576)
    The basic idea is quite simple: start by sampling a whole bunch of neurons (usually a local EEG or fMRI of some sort). Then,

    In humans, obtain two recordings (one blank and one while thinking about doing X), then diff the two and map to X'.
    In monkeys, also get two recordings (one blank and one while doing X), then diff the two and also map to X', hoping that doing X reads the same as thinking about doing X.

    You'd need to repeat these steps a bunch of times to get good signal to noise, and also need several controls (thinking about Y, Z) to make sure the mapping is specific enough. Normally, the technique is just good enough to allow quadriplegics to click buttons and such, but takes lots of effort and patience (and lots of costly equipment).
  • Re:monkey business (Score:2, Insightful)

    by LaskoVortex (1153471) on Wednesday January 16, 2008 @10:45PM (#22075822)
    > Several things make me question that. One, why is the robot in north Carolina and the monkey in Japan? It's just for show. Nothing of scientific significance is being demonstrated by that. We all know that internet can connect two gizmos across large distances. The experiment could have been conducted much more simply at one location and made no less effective a point (except to clueless investors maybe).

    If you had built a robot in Japan and your friend figured out a way to read a monkeys thoughts in NC, why would you ship one operation to the other location when you can link everything up with wires. Sounds like they saved a lot of money here.

    > Secondly, because of the distance, there is a significant delay (TFA says 250ms, about what I might have guessed.) This would seem to preclude the monkey being able to control the robots actuators in any direct sense. I.e. lift thigh, swing lower leg forward, position foot, lower thigh, positioning body over front leg. Walking is a "controlled fall". No way you could issue all those commands 250ms ahead of seeing or feeling their effect. You'd trip and fall.

    The conscious information required to generate human walk is tiny in comparison to its complexity at the physical scale. Most movements required for walk are processed between the muscle and the spinal chord and never involve the brain at all. This is called "reflex". It is legitimate to locate the reflex action in the robot when one considers the actual physiology of walking.

    > My guess is that they are simply getting a binary command value from the monkey: "walk" or "don't walk". And the whole robot thing is just for effect. I hate to be such a cynic but this looks like showmanship, not science. If that is the case then this is equivalent to the simple video games that have been demonstrated using brain control.

    The use of a robot is *proof* of principle. You can not simply model a phenomenon on a computer screen and claim success. We live in a physical world and so we need to perform physical experiments to validate our theories or test our systems. This is why Nobel prizes go to experimentalists as a rule (or at least to the people who acquire funding for the experiments).
  • Re:monkey business (Score:5, Insightful)

    by yet another coward (510) <yacoward@[ ]oo.com ['yah' in gap]> on Wednesday January 16, 2008 @10:53PM (#22075892)

    Being able to read the monkey's brain sounds like the only innovation here, not making the robot walk.

    No, reading the monkey's brain has been done many times before. This report is gee-whiz, but nothing in it is very innovative.

    So, what is the monkey really doing? I doubt if he is even thinking "left, right, left, right" because even that would be hard to coordinate with so much lag.

    When you walk, you don't think "left, right, left, right." A lot of the rhythm generator is accomplished by central pattern generators, many of the ones involved are in the spinal cord. The same way the brain engages the walk routine built into downstream parts of the nervous system, the brain can engage the walk routine built into a Japanese robot.

    Finally, why is there a damn robot in the first place?

    There is a robot because this group's ultimately goal is to develop neural prosthetics. They have done experiments controlling computer animations, as have quite a few other research groups.

    My guess is that they are simply getting a binary command value from the monkey: "walk" or "don't walk".

    You have a good point here. How finely grained is the monkey's control of the robot? The article does not tell us. I looked unsuccessfully for a corresponding scientific publication. I hope this study is published soon with more details about how specific and how precise the control really is.

Top Ten Things Overheard At The ANSI C Draft Committee Meetings: (9) Dammit, little-endian systems *are* more consistent!

Working...