I am doing my thesis on Evolutionary Computation. In *n*-dimensional space point (let's call individual) has *n* variables associated with each dimension. Constraint is summation these variables must be around 1 (between .95 to 1.05). 

So the question is, how can I generate *n* random numbers whose summation will be around 1?

This isn't directly relate stats. But I would really appreciate any sort of help.

A fitness function evaluates fitness of every individual. Say, I have two individuals **a** and **b**. Variables associated with individual **a** will be adapted by some function (which I haven't figured out yet) of euclidian distance with **b** & difference of fitness value. These variables will be adaptive (by I guess something like covariance matrix). So if I increase value of some variables, values of some variables have to be decreased to maintain summation 1. Any idea or suggestion?

Sorry for not using latex and I don't have stronghold on stat.

You can have a look at [CMA-ES](http://en.wikipedia.org/wiki/CMA-ES).