🔒 Locked

Experiment with a (real) ant.

Started by owl Nov 21, 2005 at 6:39 PM 78 replies 13.8k views

Original Post

Nov 21, 2005 at 6:39 PM

The other day I was sitting in the bathroom (don't laught) looking at the floor and then I saw a little red ant there. My first impulse was to kill it, but, as I was going expend some more minutes with my bussines I decided to do an experiment, which I'll describe next: I took the champoo and draw a circle (of about 4cm) around the ant, letting it locked up inside with the intetion of seeing how the ant would react to this. The ant began to follow the wall of champoo, apparently looking for a way out, going back and fort the exact center of the circle and staying there a while, like if it was deciding where to try next. It did three attempts to walk over the wall without success, it sank into it and slid inside again. Then it started to follow the wall again (kind of desperately). At the end it went to the center of the circle again and remained a few seconds there, then it walked directly into the wall of champu (this time underneath it) and, half walking, half swimming, it succeded to pass trough. I must confess that I did not expect the ant to be able to find a way to get out there. I belived ants where more like an instinct based automat than an intelligent animal. Conclusion: The ant was served with a problem, it tryied several (kind of ingenious) ways of getting out of there, and it did it. The ant solved the problem. Epiloge: Of course, I respect intelligence very much and for that reason I let the little ant continue it's way and live. Now I have an anthill in the kitchen. If it fits, I would like to hear your thougts about this and animal intelligence in general.

LessBread

Nov 21, 2005 at 6:46 PM

Quote:
Original post by owl
Now I have an anthill in the kitchen.

Um, maybe we should run some intelligence tests on you. [grin]

Trapper Zoid

Nov 21, 2005 at 6:50 PM

I only know a little bit about insect intelligence, but there is some overlap between this field and computer vision & robotics, as the CV/robotic researchers are interested in the instinct based algorithms that insects seem to follow, and the insect biologists are interested in the application work in various areas the computer science guys do (like in optic flow or behaviour-based robotic systems, for example).

From what I've heard and read, I think insects still seem to follow a set of coded rules (even if we don't know what all of those rules are), but the interplay between these rules is so deep very intelligent behaviour results. For example, one researcher has done tests on bees, getting them to fly through a maze with markers indicating the correct direction, and after some fly through the bees figure out what the markers mean and can correctly navigate through the maze, even when the maze is modified.

So yeah, there's a lot of interesting stuff that AI researchers can learn from insects. Simple code behaviours can build up to really awesome behaviour!

Caitlin

Nov 21, 2005 at 7:17 PM

Ants can "farm" and build ant bridges too :)

I_Smell_Tuna

Nov 21, 2005 at 7:17 PM

I think we all follow the same method of trial and error although species do it at different rates, or in this case rinse, lather, repeat.

Trapper Zoid

Nov 21, 2005 at 7:40 PM

Quote:
Original post by Caitlin
Ants can "farm" and build ant bridges too :)

I'm not sure if you are referring to ant farms or not with those quotation marks, but some types of ants really are farmers!

Caitlin

Nov 21, 2005 at 7:45 PM

:) Thats what I was referring to :)

Trapper Zoid

Nov 21, 2005 at 8:03 PM

Quote:
Original post by Caitlin
:) Thats what I was referring to :)

I thought so; but I think that ants cultivating fungus is really neat, so I wanted to post the link [smile].

To make this post somewhat on the topic of game A.I., it really is amazing what kinds of emergent behaviour can result from just implementing a series of simple rules. For example, just add a wall following routine and a light/noise avoiding routine to a robot control system, and you get a "cockroach" robot. Simple behaviour would be even easier to put into game agents, because you don't have to worry about the whole issue with sensors.

Caitlin

Nov 21, 2005 at 10:17 PM

Quote:

To make this post somewhat on the topic of game A.I., it really is amazing what kinds of emergent behaviour can result from just implementing a series of simple rules. For example, just add a wall following routine and a light/noise avoiding routine to a robot control system, and you get a "cockroach" robot.

I have done this in a number of autonomous systems I designed. The most advanced one wasn't really simple but worked quite well. It was a flight computer for R/C airplanes that could taxi, takeoff, fly, land, and taxi back to park. The system evolved aver about 8 months from a simple "plane saver" system that would return a plane back to radio coverage if the pilot let if fly out of radio coverage - that system was very simple. In the end the guys that came to me for the design decided that they didn't want to bother with it so it died.

Timkin

Nov 22, 2005 at 7:29 AM

Quote:
Original post by Caitlin
BA 0B 00 0E 1F B4 09 CD 21 CD 20 4B 6E 6F 77 69 6E 67 20 69 73 20 6E 6F 74 20 75 6E 64 65 72 73 74 61 6E 64 69 6E 67 2C 20 69 6E 74 65 6C 6C 69 67 65 6E 63 65 20 69 73 20 6E 6F 74 20 77 69 73 64 6F 6D 24

43 75 74 65 21 20 3B 29

Quote:
The most advanced one wasn't really simple but worked quite well. It was a flight computer for R/C airplanes that could taxi, takeoff, fly, land, and taxi back to park.

Why not just go out and buy a Piccolo avionics set? ;) I'm sure that your skills make your 8 months of time worth a lot more than the value of a commerical product for RC/UAVs!

Cheers,

Timkin

Caitlin

Nov 22, 2005 at 6:39 PM

For me it wasn't about making a commercial product, but rather to see if I could make it happen.

Timkin

Nov 22, 2005 at 7:50 PM

Quote:
Original post by Caitlin
For me it wasn't about making a commercial product, but rather to see if I could make it happen.

Fair call.

datxcod

Nov 22, 2005 at 9:01 PM

Can I ask you what were you doing sitting on the bathroom floor ..... naked (I assume) ?

etothex

Nov 22, 2005 at 9:17 PM

41 6C 6C 20 79 6F 75 20 6E 65 65 64 20 69 6E 20 74 68 69 73 20 6C 69 66 65 20 69 73 20 69 67 6E 6F 72 61 6E 63 65 20 61 6E 64 20 63 6F 6E 66 69 64 65 6E 63 65 3B 20 74 68 65 6E 20 73 75 63 63 65 73 73 20 69 73 20 73 75 72 65 2E

Reactive systems are a powerful way to simply solve complex problems. I took a class once whre we spent 1/3 the semester on AI through emergent behavior and reactive systems.

One way to do it is with something called a subsumption architecture(see Intelligence without representation). I'm sure there's a wikipedia entry somewhere or something but I'm not going hunting for it right now.

owl

Nov 22, 2005 at 9:31 PM

Quote:
Original post by datxcod
Can I ask you what were you doing sitting on the bathroom floor ..... naked (I assume) ?

That kind of question belongs to the lounge. :) I was actually sitting in the toilet, designing some stuff on paper and "taking out the trash". To me it's an inspiring moment. :) I mentioned it just to add a little fun to the point of my post which actually is about intelligent_behavior/problem_solving in nature.

To some people this little experiment may look trivial, but to me it says a lot about the capability of that particular ant to solve a problem.

The ant was probably a "explorer" looking for a place to stablish a colony or maybe looking for a source of food. It had assigned a particualar task, any of the two mentioned above. In the middle of this task, it was faced with a particular problem which wasn't directly related with it's main task.

The ant was able to determine the scope of the problem and try different approachs to solve it, discarding the ones that wherent useful "intelligently".

I've also noted that when the ant got out, it stopped like expecting a new wall of champoo to be in the way. Only after it made sure there wasn't anoter obstacle like that in the way, it continue with what it was doing before.

I'm not confident about this, but I suspect that the ant will remember this episode and will be able to pass trough an obstacle like that much more easily next time. And maybe it's descendency will too.

MTclip

Nov 22, 2005 at 9:42 PM

its been said that ants/ bugs arnt actually intellegent alone but in groups they become/ act more "intelegent" kinda like a brain cell.. one brain cell alone "pretty much worthless" but put 5 or 6 brain cells together and WHAM! you have a low class caveman ;P

check out Cellular Automata"

Electrovolt

Nov 23, 2005 at 6:12 AM

Good post. Ants are very interesting little creatures. I wish I new exactly how he escaped.

BTW, owl, don't know if you are aware of Ant Colony Optimization? You might find it interesting.

One aspect of it is based on the observation that ants always manage to find the shortest path to a food source from their nest. Apparently ants drop pheromones as they walk along, which can then be sensed by other ants. On return from the food source, an ant will follow the path with the most pheromones back. The shortest path very soon becomes apparent as it becomes populated with the most pheromones.

Unfortunately I don't have a link to a site for you, I learned it at varsity a few years, but I'm sure you can dig it up through google.

ketek

Nov 23, 2005 at 6:17 AM

You should elaborate a bit on yourself, if your first impulse was to kill,
next time before killing a bug or a small animal, or you when you feel the impulse to do some experimenting on something alive, slam an hammer on your hand and see if it hurts

me22

Nov 23, 2005 at 3:42 PM

Quote:
Original post by Electrovolt
One aspect of it is based on the observation that ants always manage to find the shortest path to a food source from their nest. Apparently ants drop pheromones as they walk along, which can then be sensed by other ants. On return from the food source, an ant will follow the path with the most pheromones back. The shortest path very soon becomes apparent as it becomes populated with the most pheromones.

http://david.gordon.name/projects/Ant-based%20Pathfinding.pdf -- Fairly interesting read, for an honours project :P

Made me think about using the pheremones idea to make an influence map for an RTS AI in a sort of weird pheremone/breadcrumb/flocking hybrid.

flashinpan

Nov 23, 2005 at 4:43 PM

Quote:
Original post by me22
Quote:
Original post by Electrovolt
One aspect of it is based on the observation that ants always manage to find the shortest path to a food source from their nest. Apparently ants drop pheromones as they walk along, which can then be sensed by other ants. On return from the food source, an ant will follow the path with the most pheromones back. The shortest path very soon becomes apparent as it becomes populated with the most pheromones.

http://david.gordon.name/projects/Ant-based%20Pathfinding.pdf -- Fairly interesting read, for an honours project :P

Made me think about using the pheremones idea to make an influence map for an RTS AI in a sort of weird pheremone/breadcrumb/flocking hybrid.

This discussion reminds me of a book I read by Michael Crichton (sp?) (of "Jurassic Park" and "ER" fame) called "Prey". The plot of the book is a bunch of nanites are bred with artificial intelligence and then escape from the lab (typical!!). Very interesting...and hard to kill the swarm since they are molecular in size. Bullets just go right through the swarm.

Topic Locked

This topic has been locked by a moderator. New replies are not allowed.