The Simulated Binary Crossover (SBX) is a well known variation operator for genetic real-valued optimization. It's surprising that I never came to implement this before, since I've been working with GAs for three years.
This is a remarkably elegant operator which was specifically designed to preserve the properties of the one-point crossover used for binary-coded GA optimization:
1) The average of the generated values is approximately the same of the original values before crossover; and
2) The probability of a spread factor value of betha = 1 is higher than for other spread factor values, what yeilds a stationary crossover operator.
That's such a great thing that I'm finally merging it into my NSGA2 implementation for numerical optimization benchmark. To find out more, go through the following slides: