Challenge 5

You have available in this simulation an example of Darwinian / Wallacian Natural (actually Artificial) Selection which operates with most of the characteristics of evolution in the real world. It is a blind mechanism that only operates on local interactions to produce a global (and seemingly intelligent) result. It also doesn't necessarily find the optimum "solution" to a problem, but a suite of "reasonable solutions" that are quite good.

Evolution acts by encouraging individuals and populations to expore a "fitness landscape." Like a landscape composed of valleys and hills, gentle and craggy mountains, and peaks ranging from moderately high to "the highest," populations evolve by "hill climbing." Some get stuck on the lowest peaks (local optima) while others may eventually make it to the highest peak (the global optimum).

I would like you to investigate the course of evolution (the nature of the fitness landscapes) on an interesting pattern of a relatively high number of cities. Devise an interesting pattern that will be difficult for evolution to solve completely (one where the global optimum is not uniformly reached). Use a relatively high number of cities, since low numbers like 3 are trivial and 4 or 5 are probably not that much better. I would suggest between 20 to 30 to make the result both non-intuitive for a human confronted with the same problem and to give the software a workout.

Try to characterize the fitness landscape by using screen-shots of characteristic solutions. Also look at and take screen-shots of how evolution has proceeded and gotten "stuck" on a few local maxima. Please document your investigation and explain your procedures and results.