Learning Robots

carnun writes "The NewScientist is reporting on a new snakelike robot that uses genetic algorithms to change its motion pattern so that it can still function after sustaining damage." Roland Piquepaille sends in a report about another "learning" robot, named Adam, which is designed to seek out and feed from "flowers".

Comments

Worms

[rf0]

I wonder if they could extend this to create a worm type robot that if it got cut in two it could still carry on. Say a brain in each segment

Rus

Very Worthwhile

[boogy nightmare]

Damn Damn Damn, try as i might i cant find anytihng funny to write for this. I think the snake idea is inspired (imagine the fun aspects of a robo-whacking day hehehe)

Imagine and earthquake scenario or anything along those lines (911) a van turns up and releases several hundred robotic snakes all able to sense heat, movement etcetc..

Bless the Brits and their fabulous snakes :)
see got a funny line in the end

S

Self-healing robots

[rainstorm]

Given that we will increasingly use robots to do the work that's too dangerous for humans to do, self-healing robots that can get the job done, damage or no, sound like a nice step forward.

WISOR

[JohnGrahamCumming]

If you find that interesting it's worth reading about a robot called WISOR that was built by a company called Honeybee Robotics. WISOR is uses for inspection and repair of high temperature and pressure steam pipes under the city of New York. It moves through the pipes like a very large inch worm.

There's even a movie (a really odd movie in fact) about it.

John.

Self-healing and redundancy a must for robots

[thepacketmaster]

While it is certainly natural for all living creatures to learn to cope with an injury, I think this is far more important for robots, since they don't have a survival instinct.

Humans will go to great lengths to avoid getting injured (well, most sane ones). However, robots will just do what they are told. While they may be told to avoid any hazards, they just don't have that instinct that says I *really* don't want to get my limb chopped off today.

So to compensate, all robots in hazardous or isolated situations should have this self-healing built in, as well as redundant limbs. A robot could be built with 2 legs, but why not make it 4 or 6 and have a couple to spare. The same goes for sensors. Only two are needed for stereoscopic vision, but 4 or 6 would be better.

Fitness Measurement for the GA?

[wedgehead]

This article is unclear on the nature of the fitness measurement used by the GA. The following quote leads one to believe that the GA is running onboard the snake's processors (rather than in simulation):

"Once the robot was mobile, the team disabled some of its segments to see if it could adapt to injury. Initially it was immobilised, says Bentley, but as the GA continued to try to improve the locomotion, it gradually worked out how to move again, albeit more awkwardly and with an ungainly, dragging gait - but it was still good enough to get the robot to its destination."

If the GA is running onboard the robot, then how does it know whether it's successfully moving (let alone in the right direction)? Is there a sensor that's providing information on distance to a target?

Building the "terminator" into robots is true AI

[adzoox]

This is essentially building "survival" into robots. Although fictitious, The Terminator's function was to complete it's mission, it's processing would route power or find power from other limbs, etc etc.

I think that true AI will result from this "survival instinct" - because robots will eventually learn that in order to survive/continue they will need to be loved, to fix themselves (eat, be healthy), and to have shelter (come out of the rain)

Re:Self-repairing robots have been around for a wh

[Ruds]

So? Sounds like a good solution to the problem. In this case, the goal is to get to the end of the maze, not to see the maze.

Many AI and robotics researchers seem to fall into the trap of modelling human cognition rather than designing an intelligent artifact. Both approaches have their place, but it's a shame when someone who's trying to design AI gets blinded by experience as a human being and tries to program an agent to make decisions the same way.

That being said, computer vision is an important field in AI, but vision isn't necessarily the best way to avoid walls.

Matt