next up previous
Next: Brick Problem Description Language Up: Discovery is creativity: EvoCAD Previous: Discovery is creativity: EvoCAD

Evolutionary Algorithm

Instead of initializing the population with a single brick, as in previous experiments, here we want the current design to be used as a starting point.

To begin an evolutionary run, a starting structure is received by the ED engine, consisting of one or more bricks, and needs to be ``reverse-compiled'' into a genetic representation in order to seed the population.

Mutation and crossover operators are applied iteratively to grow and evolve a population of structures. The simulator is run on each new structure to test for stability and load support, and compute a fitness value. Evolution stops when all objectives are satisfied or when a timeout occurs.

Pablo Funes