Applying an evolutionary algorithm, first the morphology of a simulated passive dynamic bipedal walking device, able to walk down a shallow slope, is developed. Second, using such an evolved solution and adding minimal motoric and sensoric equipment to this morphology, a neural controller is evolved, enabling the walking device to walk an a flat surface with minimal energy consumption. The applied evolutionary algorithm fixes neither the size nor the structure of the neural controllers. Especially, it is able to generate recurrent networks, small enough to be any analyzed with respect to their behavior relevant inner dynamics. An example of such a controller is given which realizes also minimal energy consumption.
evolutionary robotics, genetic algorithms, artificial neural networks