Rediscovering the wheel with genetic optimization.
This project uses PyBox2D and PyGame to create a simple simulation environment consisting of an inclined plane and polygons. Square boxes representing the population are released on the outer left side of the inclined plane, with the goal of reaching the other side. The boxes are represented as polygons described by a list of vertices. These vertices can be modified during a genetic process to find polygon shapes that can move further down the inclined plane.
| Generation 1 | Generation 1000 |
|---|---|
 |
 |
In the genetic optimization method, mutation is used as the only genetic operator, as this has proven to work very well. The fitness value is simply the distance traveled in the positive x-direction. Since the number of simulation steps for each generation is limited, the optimization process can also select for the fastest wheels.
The rendering can be turned off by pressing the spacebar to speed up the optimization by several orders of magnitude.
With a population size of
MIT