To Kill a Flocking-Bird.

Team: 70

School: Los Alamos High

Area of Science: Natural Science

Interim: Problem Definition: Which (well-known) optimization method best optimizes flocking behavior in the least time?

Flocking behavior, a phenomenon of several objects acting as a whole, is very rarely optimized. Usually a simulation is coded, and the operator adjusts until a flock is formed, but no one has gone to the trouble to define what a good flock is and optimize it. We plan to fill this void in the area of flocking by comparatively running optimization algorithms to find the one best suited to the field. Since this is a very sparsely researched area of optimization, a diverse arsenal of techniques will be employed.

The goal of this project will be to find the optimization best suited to the behavior of flocking. The algorithm will quickly and efficiently adjust the several variables involved, such as field of vision, adhesion, separation and cohesion. Since the algorithm will be unavoidably costly on processing power, it will be a perfect candidate for a supercomputer.


Problem Solution:

To solve this problem, we will use different optimization techniques such as genetic algorithms, directed random search, simplex method, linear programming, steepest descent, and conjugate gradient. Methods may be added or deleted if necessary. We will integrate the previously mentioned algorithms into our flocking model and evaluate their performance using an algorithm that judges quality as equal distance between neighbors and maximum neighbors.


Progress to Date:

Presently, we have thoroughly researched the topic of optimization methods in programming. We have also refined our resources for the tools in optimization. During the kickoff, we were presented with various optimization tools to help us in project. We have further researched the topics and tools given to us at the Kickoff. We have and worked with the code for flocking systems in NetLogo.


Expected Results:

For this project, we hope to get a flocking pattern that carries the optimization best suited to the behavior of flocking.

References:
1. Stephanie Forrest: Genetic Algorithms: Principles of Natural Selection Applied to Computation, 2007
2. Isaac Councill, Lee Giles : Random Search For Multiple Layer Perceptron 2005
3. István Maros : Computational Techniques of the Simplex Method (International Series in Operations Research & Management Science) 2004
4. R. Fletcher: Practical Methods of Optimization 2005
5. Nigel Gilbert: Agent-Based Models in NetLogo (Quantitative Methods in Science) 2008

Team Members:

  Stephanie Djidjev
  Mei Liu
  Peter Ahrens
  Victoria Wang

Sponsoring Teacher: Lee Goodwin

Mail the entire Team