Integrating Objective Functions in Self-Modeling Systems

This paper is about balancing the competing demands of goals within and among SISSY Systems. Two of the chief design difficulties for operations in a dynamic environment are that no system or group of cooperating systems has just one goal, just as no system or group of cooperating systems has completely accurate information about the environment in which it is expected to operate. This paper addresses relevant mathematical issues involved in balancing competing goals in an uncertain dynamic environment, and proposes a new approach to study the interplay between changing goals and a dynamic environment, by focusing on the objective functions that are used to implement the goals. We also argue that classical methods of optimization are not the right approach, because they take too long and are all extremely sensitive to initial conditions, which is a problem because the systems cannot know the environment with enough precision. As a consequence of this realization, our approach balances the derivatives of the objective functions to determine a direction in which the system should change. The paper concludes with some remaining difficult research problems, and some next steps to take.