BEACON Researchers at Work: Exploring the evolution of navigation

This week’s BEACON Researchers at Work blog post is by MSU postdoc Frank Bartlett.

Frank BartlettOver the years, my research has focused primarily on understanding mechanisms of navigation behavior. My interest in navigation probably stems in part from my complete inability to navigate, often getting lost on campus or in my own back yard. Up until the last two years my investigations have addressed questions about visual navigation in hymenopteran insects -questions such as “What are the contents of visual spatial memory? “ and “How are these memories used to revisit familiar locations?” Such questions have a rich scientific history with published experiments dating back nearly a century. Much is now known about how a variety of animal taxa, ranging from roundworms to humans, learn about and move through their environments.

As much as we currently know about how navigation behavior works, very little is known about how it evolves. What constitutes the early beginnings of complex navigation strategies such as path integration and landmark guidance? What kinds of selective pressures influenced the evolution of navigation behavior? Questions of this nature have gone largely unasked. This is not surprising since the adaptive nature of directed movement often appears self-evident. In the case of attraction or repulsion in response to some environmental stimulus such as light or gravity, it is rather easy to imagine how and why such an ability evolved. But how do we go from simple oriented movement, such as phototaxis, to more sophisticated behavior such as path integration, which allows an organism to compute the direct homeward path from anywhere along the outbound route? (This ability has been demonstrated in a variety of animals ranging from insects to humans.) Organisms such as honeybees, ants, rats, pigeons, etc. display elegant behavioral solutions to deal with navigation in complex environments. However, it is very difficult to use biological systems to track the evolution of these abilities.

In order to formulate and test ideas about the evolution of navigation behavior, I have packed away my video camera and sunscreen and loaded Avida onto my laptop. The digital organisms in Avida provide an opportunity to examine the evolution of behavior over the course of hours and days rather than decades. It also provides explicit control over the environment allowing us to carefully define the behavioral task organisms must solve.

The first obstacle in my pursuit was to determine the physical and sensory capabilities of the organism. If one wishes to observe navigation, the organisms must first be able to move. To evolve any kind of interesting movement behavior the organism should have the ability to assess its environment including current position in the environment, condition or quality of the current position, the direction of movement, etc. To accommodate these requirements, we equipped Avidians with instructions that allowed forward movement and rotation as well as instructions to sense their current position in the environment. In addition we added an instruction that allowed the Avidians to sense their current directional heading.

Having settled on the organisms’ sensory and locomotor capabilities it was time to define the environment and the task. Avidians live in a grid world where they compete for space. They also compete for cycling time on a common processor, which allows them to execute their behavior. Under standard circumstances, all Avidians live and behave in a common grid. In our experiments the organisms lived and competed for space in a population grid but performed their behavior in a separate “state-grid.” The state grid was implemented to remove collisions and other complicating interactions among neighbors. We set up a rather simple foraging task defined by a single boundary. The state grid was divided into a northern resource half and a southern reproductive half. If an Avidian occupied a pink, northern grid cell and performed the sense instruction, it received a reward in the form of extra processor cycles. This extra processing power allows an Avidian to execute its behavior faster. A faster organism can move and reproduce more quickly than its neighbors and gain a competitive advantage. Collecting resources, however, was not enough to be successful at our task. Here is the rub: resource cells did not allow for reproduction. In order to successfully divide and generate offspring the Avidian had to occupy a southern, white grid cell. These cells provided no resource. So, the task required the organisms to evolve a move/sense behavioral strategy that put them into the pink zone to collect resource followed by traveling to the white zone to reproduce. This task is a simplified version of central-place-foraging seen in many biological organisms where the critter has to venture out to collect food and return home. Rather than “home” being a single point in the environment it is south of a “line in the sand”. Such a task mimics a intertidal zone where aquatic animals have to travel on shore to collect food but return to the water to avoid desiccation. Of course our digital organisms neither swim nor get wet!

We started the population with a single Avidian, placed facing north at the boundary of white and pink. This organism was equipped with a simple strategy (see Movie 1, above). In the movie the Avidian is represented with a dot for a head and a line for its body. When it flashes red it is executing a sense instruction. Our starting organism simply moved across the boarder and sensed to receive resource. It then turned around and moved to the white zone to divide. We turned this organism loose in the environment for 60,000 generations of mutation and natural selection. This took about 6 hours to complete, equating to about 1.5 million years in human generations. Movie 2 (below) shows the most successful behavior from this evolutionary run. This Avidian exhibits what folks here call “cockroach” behavior; following the walls of its environment and occasionally moving across the diagonal. Although this appears similar to the edge following behavior seen in many biological organisms the Avidian performs this strategy by simply moving without using a sense of direction.

It was instantly clear to us that these Avidians evolved a very simple strategy that exploited the geography of the state grid. To counter this chicanery, we modified the environment to provide no resource or reproductive area along the edges and diagonals of the state grid. The south is still the reproductive zone, represented by grey regions, while resource remains pink. White areas are a no-mans-land where the organism could neither collect resource nor reproduce (Movie 3, below). In addition, Avidians started their lives from random locations in the grid with a random facing. We seeded the evolutionary run with the cockroach from the last experiment and another 60,000 generations later a more interesting Avidian evolved. The dominant organism from this run evolved to use a sense of direction and ha
d multiple dist
inct behaviors. Movie 3 shows the behavioral trace of this organism. Every time the Avidian flashes red, it is trying to collect resource. When it flashes green, it is sensing its current direction. How many separate movement behaviors can you categorize? (This organism evolved 4 separate behavioral modules)

In addition to the short generation time, another advantage of studying digital organisms is our ability to open up the hood and figure out exactly how they work. For example, even though our organism shown in Movie 3 appears to seek the northeastern corner of the state grid, it has no internal representation of the corner of its world. It simply executes a loop that moves it in the northeastern direction exactly 50 times. Although there is information from the environment via the directional sensor that could indicate when it reached the corner, this Avidian did not evolve to use it. Instead, both the corner seeking behavior and the resource collection behavior, which it spends the bulk of its time performing, are timed by a mechanism that monitors the continuous growth of its offspring. Each behavior is terminated when the Avidian’s offspring has grown to a particular length. So, rather than orchestrating behavior using cues from the environment, many of the Avidians evolved to make navigational decisions based on their life history.

These experiments are the beginning exploration of how complex navigation behavior might evolve. In these cases, both the organism’s sensory/motor capabilities are quite minimal and the environments of evolution are oversimplified. Currently we are developing more elaborate sensing and more intricate environments to develop hypotheses about how simple behavioral systems may evolve to solve more complicated tasks.

To learn more about Frank’s work, contact him at bartle47 at msu dot edu.

This entry was posted in BEACON Researchers at Work and tagged , , , , . Bookmark the permalink.

Comments are closed.