Evolution of the Cube

What could have been a more fitting time to study the genetic algorithm than the sesquicentennial anniversary year of Darwin’s publication of the Origin of the Species? We are now able to computationally simulate the same processes that Darwin studied, yet in a fraction of the time that they take in nature. This project attempts to test the potential of a genetic algorithm when aimed at design problems.

What is a genetic algorithm?

Initialization – make DNA
Best results from atomization (one random number for each gene)
Selection – test against fitness criteria
Best results from competing fitness criteria and clearly defined goals
Reproduction – crossover and mutate
Properly designed DNA allows crossover to preserve difference in the population
Therefore, mutation rates should be low (we used 1%)
Termination – reach generation limit or sufficient fitness level
Best results from a population of between 30 and 100, run for approximately 100 generations
More generations always give better results, but with demising returns

Experiment #1

DNA length, width, height
Fitness criteria equal edges

Experiment #2

DNA length, width, height
Fitness criteria equal edges

Cube from Points

DNA 24 point coordinates
Fitness criteria volume / surface area ratio