h a l f b a k e r yVeni, vidi, teenie weenie yellow polka dot bikini.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
If you ever had a spare moment and watched ants tracking the scent of an ant trail, you will see that they criss-cross the path, sometimes losing the scent and then picking it back up again.
I wondered how ants knew which way to go? If they came to a path at an angle, they seemed to very quickly
decide the right direction. Left or Right?
I see two possibilities:
1. The trail gets weaker in one direction. This doesn't make much sense as the ants are going along it in both directions. I presume they re-mark the trail as they go, but how do they do it in both directions with the right reduction in strength?
2. The trail is coded.
Imagine the ant leaves a series of dots on the ground. The dots need to be deposited with the right code or it would read the same in both directions.
For example:
dot - space - dot - space - dot - space - dot, reads the same in either direction. Let's abbreviate it using Morse code:
.-.-.-.
So what is the shortest, most easiest code that surely defines the direction?
I propose:
.-..--
But it seems strange to think that ants can lay down scent like this.
Maybe more than you wanted to know.
http://www.ted.com/...rdon_digs_ants.html
[2 fries shy of a happy meal, Jan 17 2013]
What's the gradient in this case?
http://www.youtube....watch?v=nGlzoT0sLaM
(Ant mill) [Ling, Jan 17 2013]
An even bigger ant mill
http://www.youtube....watch?v=prjhQcqiGQc
Quite incredible... [Ling, Jan 17 2013]
[link]
|
| |
I think I once read that the second ant, following the explorer, also adds to the trail. And so does the third ant. The net effect is that the trail becomes narrower and stronger in the direction that the most ants go, making it relatively easy for the nth ant to follow (also adding to the trail). |
|
| |
It's possible (and I suspect) that there are actually two trails being deposited, one being a kind of identity marker generated at the nest (which you might label "home") and another which would be developed once a food-source is found (which you might label "food"). |
|
| |
Every ant, whatever state it's in leaves behind a "home" trail. There are more ants at the nest, and so generally, the scent gradient will be stronger there, branch out over established trails, and it's strength in any location could be described as a function of the number of ants that passed by in the last 5 minutes. |
|
| |
An ant will adopt one of 4 different behaviours - "search", "gather", "deliver" and "return". |
|
| |
An ant in "search" mode will follow a "home" trail, in the direction of a dissipating gradient. Where the ant gets to the end of such a trail, it might wander about a bit, before either finding food, or going back to the nest, empty mandibled. This would encourage a fractal search pattern to emerge which would be what you'd want/expect from a fit search strategy. |
|
| |
If it finds food, it picks it up, emits a "food!" pheremone, switches to "deliver" mode and reverses direction, following the "home" trail, heading in the direction defined by the stronger scent (i.e. it goes back "up" the gradient). However, it leaves behind an additional scent gradient ("food!") as it goes home. |
|
| |
Any ants in "search" mode might be influenced by detecting a "food" pheremone and switch to "gather" and follow paths defined by an increasing "food" gradient and a decreasing "home" gradient. When they find the food, they pick it up, switch to "deliver" and reverse direction, going down the food gradient and up the home gradient. In doing so, it also releases a "food!" pheremone which reinforces the "food" trail further and encouraging other ants to switch to "gather" mode. |
|
| |
When the food is all gone, the "food" trail starts to dissipate and that particular trail goes cold over time. |
|
| |
In terms of defining a direction, I think it can only be defined in terms of local strength gradients. |
|
| |
[zen_tom], I think there must be a little more to it. The [2FSOAHM] link shows that some sort of high five goes on between the ants when they meet. But check out the ant mill...
There is little of the antenna touching, and yet they are all going in the same direction. |
|
| |
[Vernon], understood, but how does that show *direction* to/from the nest? |
|
| |
Well I didn't know the details that [zen_tom] provided. But an ant can almost always follow its own trail back to the nest. If it encounters some other trail from the same nest, while exploring, I suspect it doesn't matter which way it turns. One direction will lead back to the nest and the other won't. |
|
| |
I do think ants have enough memory capacity to remember such junctions when they encounter them again, and it may specifically be able to recognize its own trail, distinct from another ant of the same nest. That would certainly help avoiding exploring where it had already gone. |
|
| |
//So what is the shortest, most easiest code that
surely defines the direction?// |
|
| |
A dot is a dab; a dash is a space and a space is a
space. |
|
| |
You can't use 3 'markings'. |
|
| |
But if I misread and you meant dash dot dash dot
dash dot, I can string those together and read the
same backwards. |
|
| |
Huh? Do you guys seriously mean you can't hear them
talking to each other about where the food source is;
what they think of nest politics, etc.? |
|
| |
Ants, and other members of the Order Hymenoptera,
employ a wide range of chemical communications signals. |
|
| |
It would be conceivable they lay (and trail from each
individual) a trail of "out" scent or a trail of "home" scent
as they pass along the trail. |
|
| |
Maybe, but then if an ant crosses the two parallel
scent trails, how does it know to turn left or right?
Same problem. The strength of scent idea doesn't
fully hold up; how would the strength be varied? Time?
Number? |
|
| |
It doesn't but if they're both parrallel(at least in "ant"
geometry) then they both lead to or from the nest in the
same direction, so it doesn't matter. For each individual
ant they just follow up or down the chemical gradient, in
the same way self propelling bacteria, or sperm, do in
terms of self orientation. |
|
| |
//You can't use 3 'markings'. // Why not? You could
have two chemicals (A,B, gap, A, B, gap), or two
intensities (strong, weak, gap, strong, weak, gap). |
|
| |
I'm having trouble with the gradient idea. If a single
ant lays down scent as it forages, then returns to its
track 5 mins later, the difference in strength of the
scent over 10cm would be: |
|
| |
Assume scent life is 1 day
Assume ant travels at 1cm/sec
=862m scent length |
|
| |
Difference in strength over 0.1m is 1/8620. |
|
| |
A quick check and our well developed eyesight is
good to see the difference between around 500
shades of grey *if they are next to each other*. |
|
| |
[MB] Fair enough. Sounds even more far fetched than
my proposal. I get the point. |
|
| |
However, considering some of the other info, I wonder
if there is just one scent. It doesn't really matter about
gradient or the direction if the ant blindly follows it
and interrogates ants going the other way.
"You have food and I'm foraging" = carry on in this
direction etc. This would work in the nest runs, too,
where the scent trails might be very confusing. |
|
| |
Hey Ling, try the same thing with different
parameters and it works out. |
|
| |
Scent "strength" at release = 60
Scent decay rate @ -1 per minute
Ant travel rate 1cm/second = 60cm/minute
Difference in scent strength = 1/cm
|
|
| |
Q: how does the scent trail last for more than an
hour?
A: Lots of Ants |
|
| |
So now, while the scent trail is decaying rapidly -
if you've got 100 ants following it, reinforcing it as
it goes, the trail "stays up". In the early stages
when only a small number of ants are out foraging,
the trails themselves are probably not that
important - if a returning ant loses its way, as long
as it has a rough idea of which direction it came
in, and as long as there are other ants out there
out foraging, it shouldn't be too long before it
stumbles back onto a live trail to give it a good
idea of which way is back home. If the trails decay
rapidly, then the gradient sensing issue should be
ameliorated. |
|
| |
The beauty of the model is that with 1000 ants all
reinforcing valid trails (i.e. those trails that lead
to food sources) those trails will stay up and
running for as long as the food remains available
at each location. After the food is all collected,
the trail needs to rapidly go cold, otherwise
you've got a bunch of ants trooping out
unnecessarily in the wrong direction. |
|
| |
Looking at the system from a distance, it ought to
resemble a few wavy tendrils reaching out, until a
food source is found, at which point that tendril
gets thicker, takes a steadily more direct route
until the food is gone. It's the same pattern you
see in inter-neuron connections, pre-lightning
strikes (though in lightning, the search tendrils
reach out from both poles) slime-moulds and
other surface-mapping functions that self-build
from interacting components. |
|
| |
[edit] woah, never seen an ant-mill before - but it
seems reasonable to assume this is just a
pathological "bug" in the gradient-following
strategy employed by the ants. It's really quite a
profound example of nature deploying heuristics
that while they work 99% of the time, can still
cause very dramatic - and imperfect results. |
|
| |
[post edit-edit] Whoever wrote the software that
these ants run on, needs to review their issues-
list! |
|
| |
//our well developed eyesight is good to see the
difference between around 500 shades of grey//...
but 50 is enough for most purposes. |
|
| |
daaaang... ants are the new lemmings. |
|
| |
[Zen_tom], thanks for taking the time to explain. It
does sound plausible. By the way, have you noticed
that ants touch antennae of almost all oncoming ants
on the trail? |
|
| |
On another point, just out of interest, what *is* the
shortest code, using two states, that can define
direction? I had a go, but I'm wondering if there might
be a shorter one. |
|
| |
//what *is* the shortest code, using two states, that
can define direction?// |
|
| |
That is a very interesting question. I can't do better
than [Ling]'s. Unless one of the states is '>'. |
|
| |
I have a hypothesis, but it's very quarter-baked. Consider a human running at top speed through snow. At first, the footsteps will be far apart, but as the person gets tired, he/she will run slower and leave prints that are closer together, thereby, without "trying" to, naturally indicate a directionality. |
|
| |
With ants, it could be the same thing. As an ant walks, its scent muscle pops out and in, leaving dots of chemical when it's out, but the gland/muscle naturally gets tired and, assuming the ant walks at a constant speed, leaves dots more sparsely as it goes; this creates the direction. It's not "intelligence", just a natural biological function of getting tired. |
|
| |
I imagine by the time they've walked all the way to the food and eaten their fill, they are totally exhausted (I would be!) and leave almost no scent trail on the way back. So the directionality remains intact. |
|
| |
Ants usually don't eat when they get there. They take the
food back to the nest, where it's used by the whole nest. |
|
| |
It was probably a pile of ant exoskeletons. |
|
| |
[MB] it is quite strange that the example I gave is not
beaten, yet. Beyond my ability to prove it is the
shortest. |
|
| |
With a humerous twist, some of the people in South
East Asia go foraging for ants (red tree ants), and their
eggs. When asked if they ever walk around in circles
instead of going home, they said "Only after too much
rice wine". |
|
| |
Who's giving the ants rice wine? And why? |
|
| |
The answer they give is: "Yes". |
|
| |