Cognitive Networks: Towards Self-Aware Networks

1.4. Evolutionary and Adaptive Systems

So far we discussed that biological systems operate with feedback-based control loops to adapt to changes in the environment. This phenomenon can be seen in the natural selection process as well. Species evolve over many generations to be able to better survive in a dynamically changing habitat by adopting new features. Influences from mutation introduce diversity and adaptability to new challenges.

Genetic algorithms (GA) are a popular class of evolutionary algorithms as a heuristic method for the solution of optimization problems. While they are not necessarily related to networking problems, they are mentioned here because they nicely illustrate the operation of genetic evolution, which is a common principle in biologically inspired networks. In GA, a problem is encoded as (binary) individuals (chromosomes) and its evolution is performed over many generations applying the typical operations of natural selection, mutation and reproduction. Selection is the process where the fitness of individuals is evaluated upon which the new generation is selected. For the pool of selected individuals, parents are chosen which then produce a child by crossover and mutation, see Figure 1.9. In crossover, parts of the chromosomes of the parents are broken and recombined to form new different chromosomes. Mutation further adds to the diversity by modifying the genetic material. This process repeats itself until a termination condition is reached. The resulting ...