Artificial Life Research

Every year the [College of Information Science] offers the "Special Exercise" class, in which 2nd year students can propose any year-long implementation project they want to do, so long as a professor is willing to supervise that project.

For some reason, three students this year have asked me to supervise projects on Artificial Life (AL) for them (which is awesome, by the way!). All of them asked me whether it was possible to eventually extend the implementation project into a research project on AL, and because I rather liked the answer that I gave to them, I'll summarize it below.

I can see a master graduate student using Artificial Life in their project for one of these three high level directions:


1. Simulating the real world to support decision making

This is the most straightforward kind of research. The goal is to make an AL system that mimics some aspect of the real world, that serves as a decision making tool. A classical example is a simulation of people fleeing from a Movie Theater room. In this case, the simulation data can be used by an architect to decide where to put the emergency exits to maximize the number of people who successfuly escape.

I have recently supervised a student on a similar project. He developed a disaster evacuation simulator which reproduced the movement of evacuees during the Kobe and Tohoku earthquakes. Using this simulator, it is possible to answer questions regarding the number and location of evacuation sites and broadcasting stations, as well as questions regarding the road network layout of a city.

2. Simulating the real world to understand it

This case is similar to the previous one in that we are trying to reproduce a system from the real world. But here, there is some aspect of the system which is not completely known, and we want to improve our understanding of it by creating implementing rules in the simulation, and seeing how these rules affect the simulation behavior.

One example of this idea is the simulation of astronomical bodies, such as the solar system. By setting up different starting conditions and letting the system evolve using the known laws of physics, it is possible to compare different hypothesis about how the solar system was formed. The same principle could be applied to the study of animal migration, the development of the continental plates, and the behavior of botnets.

3. Developing an interesting, completely artificial system

Here we depart from the restriction of the real world (although it will probably still be used as inspiration). We create a completely artificial system, such as a swarm of particles or robots, and either observe its behavior by adding complex rules, or try to have it achieve a particular desired behavior.

The goals for this type of the study vary widely, but I usually see this in the development of automatically generated content (for games and/or computer graphics), or to try and develop truly artificial life forms (soft bots).


Of course, this is just a very broad overview. There is a lot here which I am not covering, but for someone who has no idea of what Artificial Life research could be useful for, I think this is not a bad starting point.