Optimizing the initial setting of complex adaptive systems–optimizing the layout of initial AFVs stations for maximizing the diffusion of AFVs

Abstract

There are occasions when people want to optimize the initial setting of a CAS (complex adaptive system) so that it evolves in a desired direction. A CAS evolves by heterogeneous actors interacting with each other. It is difficult to describe the evolution process with an objective function. Researchers usually attempt to optimize an intervening objective function, which is supposed to help a CAS evolve in a desired direction. This article puts forward an approach to optimize the initial setting of a CAS directly (instead of through an intervening objective function) by nesting agent‐based simulations in a genetic algorithm. In the approach, an initial setting of a CAS is treated as a genome, and its fitness is defined by the closeness between the simulation result and the desired evolution. We test the applicability of the proposed approach on the problem of optimizing the layout of initial AFV (alternative fuel vehicle) refueling stations to maximize the diffusion of AFVs. Computation experiments show that the initial setting generated with the approach could better induce the desired evolving result than optimizing an intervening objective function. The idea of the approach can also be applied to other decision making associated with a complex adaptive process. © 2015 Wiley Periodicals, Inc. Complexity 21: 275–290, 2016