ASIMO to the rescue - Embedded.com

ASIMO to the rescue

While surfing the web, Michael Barr, the editor in chief of Embedded Systems Programming, came across a project undertaken by Honda. The goal of the project was to build a robot to help people. The result was ASIMO, a human-sized, human-shaped robot. ASIMO gets a lot of mainstream press coverage and he represents an impressive accomplishment, but our stable of experts, being engineers, raised a number of practical problems. Here's their exchange on ASIMO, the morality of robots, and what technology can do to better people's lives.


Michael Barr
I'm trying to figure out how ASIMO will protect itself when it's out in the “real world.” Say, for example, you are disabled and send ASIMO out for some groceries from a nearby market. ASIMO heads out, but never comes back. Or ASIMO gets its wallet stolen. Or it buys the food, and then the food is stolen from it.

And if it will have a self-protection mechanism, couldn't that be misused by a malicious owner (or robot with emotional difficulties)? For example, I might send my ASIMO to kill you, or at least to stun gun you, or whatever.

How can one idea seem so cool and yet so impractical at the same time?

Is ASIMO an acronym?

Bill Gatliff
ASIMO = Artificially Sentient, In My Opinion?

Michael Barr wrote:
I'm trying to figure out how this thing will protect itself when it's out in the “real world.” Say, for example, you are disabled and send ASIMO out for some groceries from a nearby market. ASIMO heads out, but never comes back.

“We hAvE yOuR ASIMO. sENd 1000000Y iN sMAll biLLS oR wE'LL uNPLug hIM.”

Seriously though, Michael, the answer is part of the Japanese Way: “Build it, and they will come.” In other words, work blind to the questions for a while (like, ten years in this case). But I suppose if the same company makes automobiles, power tools, generators, robots and who knows what else, they can afford to have a few skunk-works projects going on the side.

Michael Barr wrote:
And if it will have a self-protection mechanism, couldn't that be misused by a malicious owner (or robot with emotional difficulties)? For example, I might send my ASIMO to kill you, or at least to stun gun you, or whatever.

Forget that, what if a coding error causes it to stun-gun its owner? In the words of Atticus Finch, “the best way to get shot is to carry a gun.” Or now, apparently, to hang out with armed robots.

“Watch out, he's packing heat! And he's powered by Windows CE!!”

Michael Barr wrote:
How can one idea seem so cool and yet so impractical at the same time?

Ain't our job great?

Larry Mittag
Actually, ASIMO is probably most of a tribute to Isaac Asimov, who also considered some of the questions that Michael raises. His Three Rules of Robotics provide a remarkably sound “moral” base for intelligent robots. They were (from memory, so please forgive if I get it wrong):

  1. Do not harm human beings
  2. Do not through inaction allow harm to come to human beings
  3. Protect yourself

Note that these rules are in priority order. Maybe someone should get to work coding these up.

Michael Barr
I did some more digging (on Honda's website) and found this particular robot is not self-controlled. You have a wireless remote control that looks very much like a Nintendo controller. You turn it, speed it up, make it shake hands or wave or pick something up yourself. So they may be a few years away from dealing with these issues…

Larry Mittag
So what you are saying is that the hardware is ready but the software is falling behind… 🙂

Bill Gatliff
Er, so to help a disabled person it has to have an enabled person at the helm? Yeah, sounds like they have a few years to go.

Maybe a better idea is this one:

www.cs.Washington.edu/homes/kautz/ac

It describes an AI for PDAs that tries to figure out what you're doing, and then helps remind you when you forget. Intended for Alzheimer's patients.

Example: a GPS receiver tells it that you're waiting at a bus stop. Using an onboard copy of the bus schedule, the time of day, and your previous habits, the device concludes that you're headed to a particular destination (a doctor's appointment perhaps). If the patient suddenly realizes she doesn't know why she's on the bus, the PDA can tell them. Or, at least, give them directions back home.

Very, very difficult stuff. And not as sexy as a robot, either. But probably more practical and achievable.

There are a couple of disabled people in my neighborhood, and my grandmother died after suffering from Alzheimer's for a number of years. It's obvious to me that even with all the modern gadgetry around us, unless you have two eyes, two ears, and ten working fingers, you may as well be living in the stone ages.

Except for the joystick on my neighbor's wheelchair. Man, that's one *smart* peripheral. It takes his jerky, MS-laden commands and translates them into exceedingly precise movements that allow him to navigate close spaces as well as I do. Amazing.

And yet, there's no equivalent device that lets him play video games. Or do email. Or drive a car. Or prepare a meal. Or use the bathroom. Or … you get the idea.

There's a whole class of problems like that that nobody seems to be looking at. I'd love to, just tell me where to start!

Larry Mittag
Interesting stuff. Intended for Alzheimer's patients, but do it right and it also benefits everyone else. Everything from forgotten items on a grocery list to a Wizard that can help you drive your car (Do you really want to change lanes here, Dave? That truck is a lot bigger than you are). Put software like this that can derive intent in charge of that robot and things could get interesting.

I know what you mean, Bill. Someone in my company just lost half a leg to a motorcycle accident. He was considering a prosthetic with an embedded control system, but only if he could get the source code so some of our people could make improvements.

Bill Gatliff
My brother was in a serious automobile accident a few years ago. (He survived, but it wasn't pretty for a long time.) Thomas struck a dump truck, head-on. The dump truck had crossed the centerline of the two-lane road doing about 45 MPH, but since both vehicles were rounding a curve at the time, my brother couldn't tell there was a problem until he was too close to avoid it. The driver of the truck swerved as my brother crossed the centerline to avoid him, but that only caused the dump truck — and attached trailer — to suddenly block both lanes.

Fortunately, the EMT station was about 100 yards away, and they heard the impact and were on the scene in seconds…

Since then, I've been mentally mulling over an automotive collision-avoidance system that's completely visual. All the pieces needed to produce it are in place, and are even being tested elsewhere (like at CMU's Field Robotics Laboratory), but nobody seems ready to put it in a car yet.

The system would use two digital video cameras positioned at the upper corners of the car's windshield, facing forward with a 180-degree view each. A microprocessor would correlate the images in real-time, and identify moving targets that had the potential to cross the path of the host vehicle. A heads-up display would paint a dot on the windshield at exactly the location where the driver needs to look to see the object in question.

All the math to do this is pretty straightforward, and well within the confines of some not-so-high-end embedded processors. So why don't we have a system like this yet? And why is the next best thing, which is an RF unit produced by Eaton, not even close? Beats me.

Maybe I need to quit waiting for everyone else to get to work on it, and get it done myself. I mean, that's what consultants are supposed to do, right? Who's with me?

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.