Engineering Animals

I read recently a book titled Engineering Animals that I picked up from the library after searching for book on emergent behavior. Within the text the authors shed light on several perspectives and applications within fields ranging from biology to engineering.

Among the topics covered were:

Structure and Movement:
Endo and exo skeletons, habitat building, movement through a medium (e.g. – land, sea, and air), and electrochemical (e.g. – how neurons work)

Sensors:
Chemical (e.g – smell or taste), mechanical (e.g. – hearing) and electromagnetic (e.g. – sight)

Combining the previous topics for more complex results:
Including Intelligence, emergent behavior, navigation, and communication

The chapters on the aforementioned subjects were the most salient to game design, particularly those on emergent behavior and navigation.

Please note, I am not going to summarize the entire chapter but am instead selecting the points that the exposed that which I found most intriguing to my study of Artificial Intelligence (AI).

Emergent Behavior

Most animals have three goals in mind; sustenance, security, and sexual drive, and it turns out the cockroach is no exception. Recently cockroaches have been studied as an analog for autonomous robots. Yet when scientists tried to model the cockroach they had difficulties in making the models match the manner in which the creatures behave in the real world.

Why? Scientists were using the wrong approach, originally working from a top-down model, they were asking themselves complex questions about the creatures goals, “What do cockroaches achieve?”. After numerous failed attempts to using this complex study to generate a realistic replica, the scientists changed their approach to the problem, by instead using bottom-up approach asking; “What do cockroaches do?”. The truth is not much, as cockroach behavior can be broken down into these simple rules:

1. If they are awake and next to a wall, they follow it
2. If they sense food, they head towards it and eat until full
3. If they are ready to mate and the sense a potential mate, they head towards it
4. If they are threatened, they run away
5. They avoid bright lights
6. If none of the previous cases occur then they wander aimlessly until something happens

What I find really interesting is that by linking all these simple rules together a remarkably accurate model of a cockroach was simulated. And it turns out that the same bottom-up approach can be applied to most types of eusocial insects (with few exceptions). The authors discuss termites, ants, bees, and wasps as well the original cockroach model.

I was particularly intrigued by a complex study on Ant bridges.

Here is the Ant Bridge Algorithm:

1. Avoid obstacles
2. Wander randomly but show preference to pheromone trails
3. If you encounter an edge, stop with half your body over the edge. Wait for some time. If nothing happens go back to wandering
4. If you find a stationary ant, climb on top of it and keep moving forward
5. If there is another ant hanging off of you don’t move, and emit a pheromone to call for help

Turns out that this simple algorithm is all that is needed to get ants to build bridges across gaps or streams. Talk about a cool upgrade to Sim Ant!

Birds of a Feather Flock Together

Flocking behavior is also surprisingly simple as can be seen with European Starlings flocks (which are often seen in North America and Europe). As an observer the patterns of the these birds appear to be very intelligent; the birds instantly change shape and direction.

But how do they do it?

The answer is another very simple series of rules that each bird follows:

1. Avoid obstacles (e.g. – trees, light poles, etc.)
2. Separation: avoid crowding neighbors by maintaining a minimum separation from each other
3. Alignment: steer toward the average heading of neighbors
4. Cohesion: steer toward the average position of neighbors

These limited rules at first appear to be almost too simple, yet as the book points out nature often chooses simplicity over complexity.

The rest of this chapter was on wolf pack tactics, which I didn’t find very compelling as wolves only have about a 30% success rate. The authors very briefly mentioned the subject of Cape Hunting Dogs, which have upwards of a 90% success rate. However, the listed information on the tactics of these hunting dogs is scarce on the web, so I have emailed around to see if I can get more information.

After reading it all, this book is a pretty good study for anyone interested in engineering, biology or game design. Particularly, those attempting to better understand the qualities that create the unique behavior of animals.