on 11-05-2009 01:20

Editor's rating

Average user rating    (0 vote)

Views 1114

Favoured 84

 

Human beings are just brilliant. We can dive to the sea bed in submarines, fly to the moon in rocket ships, design uber-powerful computers and so on and so forth.

But so far we've failed to deliver one of the most tantalising promises of the modern age. Personal domestic robots have remained irritatingly elusive.

Here are some of the reasons that making humanoid robots isn't as easy as we were led to believe by novelists, screenwriters and scientists.

 

The Asimo is Honda's effort at a robot

Walking is something that most of us achieve at about 12 months old and henceforth never give a second thought. Unless we're shopping.

But it's one of the trickiest engineering problems around - and robots can't do it. It sometimes looks like they can, but they can't - not really. Human walking is controlled falling. Robots doing controlled falling ends up in falling - but with a complete absence of control.

Much is made of the "walking" abilities of robots like the very groovy looking HRP-3, and Honda's very groovily named Asimo, but both these humanoids are designed to be stable at all times when walking, which paradoxically makes them fairly precarious creatures.

Legs seem like a brilliant idea to help a robot on its journey around our staircase world - but they've turned out to be trickier to control than anyone had imagined. Unfortunately for even the most advanced "walking" robots, the world outside its sterile hazard-free laboratory comfort zone remains a no 'bot zone.

 

Robots can't juggle eggs. Or even pick up boxes. If robots are to be of any use they will need to be able to handle things with the care and dexterity that we do.

There's a robot that can just about deal with staplers at the moment, but that's no use if what you want is a hot cup of tea. Or a cold beer - from a fridge.

The engineering and control problems of picking things up without crushing them are proving much trickier than science imagined. So rescue robots are out.

But those insanely complex issues pale into insignificance when you start attacking problems of perception and intelligence.

 

You can feed a robot with video, but you can't make it see. Not very well, anyway. Robot "vision" is very limited. Robots like MIT's beautiful looking Domo can respond to certain pre-programmed colours or shapes - and react in a few limited ways.

It's a start, but it's not the kind of useful sight that would enable Domo to make sense of the real world. Domo cannot react to non-programmed situations. This is because it doesn't actually know what it's looking at - it may look clever, but it cannot think.

 

If you thought making a robot see was hard, have a ponder about making one think. Human sight is very closely tied to our incredibly powerful cognitive function - in reality, the human brain does all the work for the eyes.

It took us millions of years to evolve to our current top-of-the food-chain, world-dominating status, so it's little surprise we haven't been able to create a robot in our own image in a handful of decades

Seeing is all about interpretation and being able to act on what it is you are seeing. Robots can't do this. However impressively they track objects or respond to certain words, they're only following orders.

Making a robot think would mean overcoming all the problems of perception, combining individual senses to give the robot a way of being "aware", and then giving the robot the capability of being aware of the fact that it is aware.

 

There is no single robot that can do everything - not even a female one. In fact there are very few that can do more than one thing with any degree of flair.

And the robot that can do much outside the lab where it was born and feels safe is a very rare creature indeed. Until we get the complete package, overall usefulness will be forever compromised.

This is the holy grail of robot design - if more skills become integrated the value of a robot around the home will increase exponentially.

 

It might be that we're trying too hard to create a robot that mimics or recreates human capabilities in a human way. This could be a futile endeavour.

It took us millions of years to evolve to our current top-of-the food-chain, world-dominating status, so it's little surprise we haven't been able to create a robot in our own image in a handful of decades.

Maybe we're just not capable of such a feat. Some people think that the only way to get a robot is to let them evolve - to live and learn. It is this approach that Prof Gerald Edelman is taking with his Darwin robots.

The Darwins learn and evolve, but progress is necessarily slow. In evolutionary terms, they're still very much in the swamp.

 

Robots are really slaves without the inconvenient ethical issues. The problem is though, that any robot worth having - assuming the whole walking, seeing, dexterity, autonomy, integrated function thing can be cracked - would have to perform at a level so indistinguishable from humans that all the troublesome "human" rights issues would irritatingly reappear.

In short a really useful robot would need to be - to all intents and purposes - a crying, talking, sleeping, walking, living doll. That could think. A human. Which defeats the object.

 

Because of the practical and philosophical problems with making humanoid robots, artificial people like C3PO are unlikely to be next year's Christmas "must have".

But all is not lost. Though the golden man is destined to stay on the screen and in our imaginations, his trusty sidekick R2D2 is on the point of bursting through the celluloid and into our houses. We've seen the future - and it's called Stair.

Although it wouldn't have impressed the makers of I Robot, it represents the first step on the road to robots. Stair doesn't think, it doesn't walk and it doesn't see. But thanks to the latest radar, infra red and ultrasonic sensors it can safely roll around rooms it's never been in before, pick things up, open doors and respond to simple voice commands.

According to its creator, Andrew Ng, in 10 years or so it will be able to clean and tidy our houses, put our clothes away, load and unload our dishwashers and empty the bins.

And that's what it's all about really isn't it? A metal servant to do the jobs we hate. And to fetch beer. Just don't expect walking. Or conversation.

Last update: 11-05-2009 01:31

Published in : Bot News, Personal Robotics