A method for behavioural organization for autonomous robots based on evolutionary optimization of utility functions

摘要

A method for behavioural organization for autonomous robots, based on evolutionary optimization of utility functions, is introduced and illustrated through several simulation examples. The method is shown to be efficient in generating behavioural organization systems that are both flexible and robust to noise. With this method, the amount of hand-tuning of parameters is minimized and, in principle, the user is only required to define fitness functions for the behaviours directly related to the task of the robot. The utility functions representing the beliefs and intentions of the robot are, in general, obtained through evolutionary optimization. However, if desired, the user also has the freedom to specify utility functions by hand.