Evolution
Intelligent Design
Ant Navigation Fascinates Engineers
Watching ants optimize their search strategies is fun, and motivates biomimicry, but few are the biologists who consider what design requirements make it possible.
Research teams from two universities recently observed navigation strategies for different ants that live in very different habitats. Both were amazed at their skills and thought the findings would be helpful for robotics engineers. Both, however, simply did a lateral pass to Darwin to explain what they observed. It “evolved,” they said dismissively, thinking the explanatory work was done. Design theorists know better. They understand the hardware and software requirements to optimize a search algorithm.
Tree Ants
The first team, harking from Stanford and publishing in PNAS, studied arboreal ants. The arboreal turtle ant (Cephalotes goniodontus), common in Mexico’s subtropics, can solve the “shortest path problem” while navigating on the limbs of trees and bushes, these authors say. Finding the shortest path is a tricky problem facing service providers who deliver goods. The University of Texas explains the “Chinese Postman Problem” —
It is the problem that the Chinese Postman faces: he wishes to travel along every road in a city in order to deliver letters, with the least possible distance. The problem is how to find a shortest closed walk of the graph in which each edge is traversed at least once, rather than exactly once. [Emphasis added.]
In a tree, limbs can be considered “edges” and branching points can be considered “vertices.” Think of the complexity of branches going this way and that, with food sources at unknown distances from the nest. How do the ants minimize the energy cost of foraging, locating, and retrieving their food? And what do they do when obstacles are in the path?
Here, we investigate how the trail networks of the arboreal turtle ant (Cephalotes goniodontus) can solve variants of the shortest path problem, a basic optimization problem on graphs. Textbook algorithms for this problem find optimum solutions using knowledge of the entire network. Turtle ants nest and forage in the tree canopy of the tropical forest; their trail network is constrained to lie on a natural graph formed by tangled branches and vines (Fig. 1), and no ant has any global information about the network. Observations of turtle ants in the field show that a colony’s trail network approximately minimizes the number of vertices.
With no roadmap, each individual ant must somehow participate in the solution. The research team found, first, that the ants drop pheromones at each vertex and along edges. The strength of pheromones decays with time, giving them data about how recently another ant visited. The scientists noticed that two other data sources help the ants find the shortest path: (1) the bidirectional flow rate at a vertex, and (2) the “leakage” of ants as some leave the path to explore. The dynamics of these inputs helps the ants quickly converge on the shortest path.
In summary, our model for how ant trails change over time contributes to the synergistic exchange between biology and computer science, providing a plausible explanation for how turtle ant colonies can find paths that minimize the number of vertices, and suggesting a surprising algorithm for the shortest path discovery, by increasing the flow rate, applicable to distributed engineering systems.
How do they explain this surprising algorithm? “Evolution has led to natural algorithms that regulate collective behavior in many biological systems.” Enough said?
Rock Ants
Another team from the University of Arizona investigated a different kind of ant: a tiny species that inhabits rock crevices. “The ants go marching … methodically,” they say. Their work overturned an assumption about ant navigation: it’s more methodical than previously thought. Using a familiar analogy, they begin,
When strolling through an unfamiliar grocery store, you may find yourself methodically walking down each aisle to ensure you find everything you need without crossing the same path twice. At times, you’ll stray from this orderly process, such as when you see a vibrant “for sale” sign from across the store or realize that you forgot something. According to a study led by researchers at the University of Arizona, some ants go about their search for food and shelter in a similar manner.
Stefan Popp and Anna Dornhaus collected rock ants (Temnothorax rugatulus) near Tucson and set up an experiment in their lab where they could watch their movements with cameras and tracking software. This species does not form the familiar ant trails we find in our homes; rather, the individuals forage on their own.
Anecdotal evidence suggests they mostly forage on small living or dead arthropods and opportunistically lick up sugary liquids. Individuals can discriminate their own pheromone from that of other ants and are not attracted to or follow the trails of nestmates toward food. Their traits of foraging individually, having a relatively small range compared to other ant species, and small colony size make these ants a good study species to investigate search efficiency of Central Place Foragers.
Contrary to expectations, the individual ants did not move about in a random manner but instead walked in a back-and-forth motion as they explored. Hunters often train their dogs to catch scents using a similar method. The dog may need guidance from the hunter when blocked by an obstacle. These little ants, though, when faced with an obstacle, will switch to random mode to continue. How does this navigation strategy succeed in optimizing the path?
“Until now, the widespread assumption was that free-searching animals are incapable of searching for new resources methodically,” Popp said. “Most of the previous research on search behavior only focused on situations where the animal is already familiar with where it’s going, such as going back to the nest entrance or going back to a memorable food source.”
“Based on these results, many animals may be using complex combinations of random and systematic search that optimize efficiency and robustness in real and complex habitats,” Dornhaus said. “This discovery opens up a whole new way of looking at all animal movement.“
As in the previous research, this team believes what they discovered will help robot designers and other problem solvers. The ants’ strategy also “has the potential to unify different fields of science” and to provide “applications for real environments where a completely systematic search would fail when faced with an obstacle.”
So how does this team explain the origin of this robust, adaptable search strategy? It evolved. It evolved all over the world! “According to the researchers, the evolutionary advantage of meandering found in these rock ants could have possibly evolved in other species of insects and animals as well.”
Ant What It Used to Be
Is it helpful in science to toss the explanation for a complex ability to some unaccounted-for series of mistakes in the past? Can anyone really imagine some pre-ant without these abilities achieving the spectacular innovations that make their navigational skills so attractive?
Both teams realize, of course, that engineers who will use their findings to build robots or optimize search algorithms will have to apply their intelligent minds diligently to succeed as well as these tiny ants have. Undoubtedly, the engineers would have to draw up a set of requirements before creating a successful application. Are the evolutionary biologists overlooking the requirements for ant navigation?
In his excellent book Animal Algorithms, pp. 62-65, Eric Cassell shares additional astounding capabilities of ants. He mentions that an ant brain is one fourth the size of a bee brain, with about 250,000 neurons. Within that tiny brain, elegant software operates that can do landmark recognition, vector analysis, and path integration, using multiple sensory inputs: a sun compass, pheromones, and polarized light. The brain, furthermore, requires at minimum enough memory to store and retrieve the information, an odometer for measuring path length, and decision algorithms for chemotaxis. These all must be under central control to enable path integration. Without these requirements being met in each individual, the ants could not perform the wonders scientists admire. Cassell asks at the end of this discussion, “How did these complex programmed behaviors originate?” The question deserves a better answer than, “They evolved.”