Applying evolution strategies to neural networks robot controller

摘要: In this paper an evolution strategy (ES) is introduced, to learn weights of a neural network controller in autonomous robots. An ES used high-performance reactive behavior for navigation and collisions avoidance. The learned able solve the problem different environments; so, learning process has proven ability obtain specialized behavior. All behaviors obtained have been tested set environment capability generalization showed each No subjective information about “how accomplish task” included fitness function. A simulator based on mini-robot Khepera