Coevolutionary architecture for distributed optimization of complex coupled systems

A coevolutionary architecture for distributed optimization of complex coupled systems is presented. This architecture is inspired by the phenomena of coevolutionary adaptation occurring in ecological systems. The focus of this research is to develop flexible design architectures for addressing the organizational and computational challenges involved in optimization of large-scale multidisciplinary systems. In the proposed design architecture the optimization procedure is modeled as the process of coadaptation between sympatric species in an ecosystem. Each species is entrusted with the task of improving subdomain specific objectives and the satisfaction of subdomain constraints. Coupling compatibility constraints are accommodated via implicit generalized Jacobi iteration, which enables the application of the proposed architecture to systems with arbitrary coupling bandwidth between the disciplines, without an increase in the problem size. A domain decomposition approach is presented for distributed structural optimization to construct a class of test problems. Numerical studies are presented to demonstrate that convergence to an optimal solution satisfying the subdomain and coupling compatibility constraints can be readily achieved