I'm going to play biologist for a moment, and talk about a species other than humans or nonhuman primates. First, imagine that you're about 10 mm long, a couple mm high, and you're stuck in the middle of the Sahara desert. Eventually you've got to find food, so you leave the comfort of your burrow to forage for food that could be many meters away. When you find food, you then have to find your way back home. And all you have to do this is a brain that weighs 0.1 mg (see the image below). If you're a member of the genus Cataglyphis, you do this on a daily basis, and you do it so well that you're a puzzle that researchers have been attempting to solve for more than a century. These researchers have attempted to answer three types of questions: What information do these ants use? How do they use it? And how did their navigational system evolve? Since my ability to play biologist only goes so far, I'm not even going to attempt to answer the third question, but I will try to explain what I know about the current answers to the first two.
From Wehner (2003, p. 580; see footnote 3 for reference)
OK, I know what you're thinking. You're thinking, "Chris, you don't study ants!" That's true, but thinking this you've seriously underestimated the heights of my nerdiness. Let me explain how I came to love desert ants (despite never having actually seen one). I once took a course on spatial navigation in humans and robots, and strangely enough, the first readings for the course weren't about humans or robots, but about Cataglyphis. It turns out that human navigation is very difficult to study (and model in robots), because we've got these really big, complex brains, and we're inherently neurotic beings, which means we tend to produce really noisy data. Ants, on the other hand, have little tiny, relatively simple brains (see the picture below), and are rarely very neurotic (though they are more likely than most human subjects to bite experimenters), so they make for much simpler research subjects, and much cleaner data. Furthermore, the strategies they use for navigating appear to be analogous to those found throughout the animal kingdom. Therefore, studying how they navigate their environments can give us clues to possible navigation strategies that we, with our big, complex, neurotic brains might use in ours. After reading about the little bug(ger)s for the course, I became fascinated with them and their impressive feats of navigational skill, and I've been following the ant navigation literature ever since. And now I share my wealth (cough) of knowledge with you, though it wouldn't be a stretch to say that I'm just writing this post for myself. So let's get started.
The Compass in the Brain
The first key to navigation is having a sense of direction (something completely absent in me, I might add), so that you can determine your heading. For this, you will need a compass that utilizes some feature of the environment to give you absolute directional information. We humans have developed the magnetic compass that utilizes the Earth's magnetic field. Ants, however, have yet to invent the magnetic compass (and they don't have one in their brains, as some bird species do). What other feature of the environment might be utilized by a compass? Well, since most (if not all) Cataglphyis species forage during the day, the best source of directional information is the sun. There are a couple ways to utilize the sun do calculate headings. One is to use its azimuthal position of the sun (you can compute the azimuth of the sun for your location here). Ants do appear to use this information, but they rely more heavily on a more reliable source of directional information: the pattern of polarized light (E-vectors) in the sky 1. Here are the basics of how this works2:
As shown by extended series of parametric behavioural tests, bees and ants are programmed with a strikingly simple internal representation, or template, of the E-vector patterns in the sky (Fig. 1). While the actual patterns change with the elevation of the sun, the insect invariably applies its hard-wired internal template. It might do so in a template-matching mode. At any one time, the best match between the internal template and the external pattern is achieved when the insect is aligned with the solar (or antisolar) meridian. At this point, maximal overall responses summed over all detectors of the neural template occur. The match decreases as the animal deviates from this reference meridian (0 Â° or 180 Â°, respectively, in Fig. 1). (p. 130)
In other words, in the ant brain there is a "template," or a representation of the E-vector patterns of the sun, and this pattern is matched to the actual E-vector patterns in the environment. Doing this allows the ant to determine his orientation relative to the position of the sun, and thus his heading.
How does the ant's little tiny brain do this? Well, it appears that inside the brain there are compass neurons that receive as input information from groups of neurons that are specifically designed to detect E-vector patterns, which act on opponent processing principles much like the cones in human eyes (see this post for a description of opponent processing in human color vision). This means that a particular group of E-vector pattern detecting neurons will be the most excited when the E-vector patterns indicate one direction, and be the most inhibited when they indicate the opposite direction (180 degrees from the exciting direction). These neurons in turn receive their input directly from the eye >. The compass neurons use information from the E-vector pattern detectors to indicate the ant's heading at any particular moment3.
This compass system is dependent on lighting conditions. If, for example, it is overcast when the ant begins a foraging trip, its compass system calibrates itself to these conditions. Ordinarily, since the foraging trips of ants are of relatively short durations, this is sufficient. If lighting conditions change, however (e.g., the clouds dissipate), the ant's compass system is thrown off, and it will not be able to determine its heading4.
The Odometer in the Brain
Knowing which direction you're facing is great and all, but by itself it won't help you get anywhere. You also have to know how far you've traveled. You need an odometer. To measure how far they've traveled, Cataglyphis uses "self-induced optic flow"5. In other words, it they use the speed with which images pass across their retinas (this strategy is also used by honeybees
Putting the Compass and Odometer Together: Path Integration
Once you've got a compass and an odometer, you can use them together. When sailors do this, it is usually called dead reckoning. When animals do it, it's called path integration. Darwin was one of the first to hypothesize that animals might navigate using dead reckoning8 (see, I can play biologist really well; I cited Darwin). He wrote:
Whether animals may not possess the faculty of keeping a dead reckoning of their course in a much more perfect degree than can man; or whether this faculty may not come into play on the commencement of a journey when an animal is shut up in a basket, I will not attempt to discuss, as I have not sufficient data.
(See, I even quoted Darwin... I am a biologist!) Anyway, here's how path integration works in ants. As soon as the ant leaves its nest, it begins to keep track of a vector that points straight home. Wherever it goes, it maintains this vector, constantly updating it based on its current heading and the distance it's traveled. When it decides to go home, it follows this vector until its current position matches its target position represented in its hea. If, while on its journey home, a cruel experimenter decides to pick it up and place it somewhere else, then upon being placed on the ground, it will immediately begin to follow the same vector, believing that it still points home9. When it gets to the position where it thinks its home should be, it will begin to spin, turn, and search around looking for it. After ants have discovered food sites, they will also store vectors that point to them as well.
For some species of Cataglyphis, path integration is the only method available to navigate their environment, because there are few if any objects between the nest and food sources10. This is the Sahara, after all. Other species live in areas with some vegetation. These species can use landmarks as a supplement to path integration. When the ant encounters a landmark, it stores an image (or snapshot, as some researchers call it) of the landmark from a particular perspective11. Associated with that snapshot will be a motor program that specifies the movements the ant should take, i.e. how it should adjust its heading and perhaps how far it should move12. The ant then uses the landmark on subsequent trips by moving towards the landmark until the current image on its retina matches the snapshot stored in memory.
Combining path integration and landmarks, ants can store vectors associated with individual landmarks. Thus, when they reach the landmark (when the retinal image matches the snapshot), they can switch to a new vector and begin to travel along its trajectory. Researchers learned this by training ants with a set of objects that they could use as landmarks and then, in the test phase of the experiments, spreading the landmarks out. The ant would travel to the first landmark and, after reaching the point where it matched the snapshot, traveled in the usual direction until they reached the point at which the vector ends. Ordinarily, they should have been able to see the next landmark (or the food, or home, or whatever they were looking for), but since the landmarks had been spread out, they now began searching (as they do when they're picked up while traveling on a route) until they caught sight of another landmark, at which point the process began all over again13.
So, that's how ants, with their little tiny brains, navigate their environment. They use sun's E-vectors for headings, retinal flow and the number of steps they've taken to compute distances traveled, and combine these for dead reckoning, or path integration, so that at any given point on their trip they know the direction and distance they need to travel to reach their goal. In addition, they associate particular movements with landmarks in their environment (which are represented as visual snapshots), so that they can travel from one landmark to another (or to the goal). And all of this, including the vectors and images stored in memory, is accomplished with a brain that weighs 0.1 mg. I don't know about you, but I find that pretty amazing.
1 Wehner R (1994) The polarization-vision project: championing organismic biology. In: Schildberger K, Elsner N (eds), Neural Basis of Behavioural Adaptation, pp 103-143. Fischer, Stuttgart.
2Wehner, R., Michel, B., & Antonsen, P. (1996). Visual navigation in insects: Coupling of egocentric and geocentric information. The Journal of Experimental Biology, 199, 129-140. Figure 1 from p. 131.
3Wehner, R. (2003). Desert ant navigation: how miniature brains solve complex tasks. Journal of Comparative Physiology A: Neuroethology, Sensory, Neural, and Behavioral Physiology, 189(8), 579-588.
4Wehner, R. (1994).
5Ronacher, B., & Wehner, R. (1995). Desert ants Cataglyphis fortis use self-induced optic flow to measure distances traveled. Journal of Comparative Physiology A: Neuroethology, Sensory, Neural, and Behavioral Physiology, 177(1), 21-27.
6Esch, H.E., & Burns, J.E. (1996). Distance estimation by foraging bees. The Journal of Experimental Biology, 199, 155-162.
7Wehner & Michel (1996).
8Darwin, C. (1873). Origin of certain instincts. Nature, 7, 417-418.
9Collett, T.S., Collett, M. (2002). Memory use in insect visual navigation. Nature Reviews Neuroscience, 3(7), 542-552.
11Collett & Colett (2002).
13Collett & Colett (2002).
Since you were talking about robots -- has anybody ever actually built a robot that emulates these ant algorithms and does so successfully?
You know, I don't know of any robots that simulate ant algorithms specifically, but there are a lot of robots that use image matching (snapshots of landmarks), and some use dead reckoning as well.
Neat post, thanks very much for the info!
I always found the topic of animal orientation fascinating. I was fortunate to see both Wehner and Collett give talks a few years back, but it is nice to get an update like this every once and a while, as I am not keeping up with that literature very well these days.
Bora, you know, I find most ant navigation boring: lay down some pheromones as you go, and the rest will follow. But these desert ants, along with bee species that fly around looking for flowers, are fascinating, because they have to complete complex navigational tasks with so few neural resources. Plus, they make for cool robots.
Oh, yeah, following pheromones is so pedestrian. I want to see some cool brain-use! Admittedly, I focused more on the birds and the bees (not in THAT sense!), but an occasional look at desert ants is really cool. And the Wehner talk was amazing!