Complex Behavior in Evolutionary Robotics

6 Data Driven Success Prediction of Evolution in Complex Environments

Preamble: In contrast to classic Evolutionary Algorithms which use completely goal-based selection operators, selection in nature is largely environment based in the sense that there is no global goal populations are pushed toward by an external force. In fact, fitness is defined by biologists by means of the amount of an individual’s offspring capable of surviving to adulthood which is just complementary to a classic computer scientist’s point of view. With Evolutionary Robotics, a computer scientist has to face reality as it introduces an intermediate case between the two extremes: there is a desired goal toward which selection is supposed to push, but there is also a complex ...