Use of concurrent processing with the adaptive complex method for global optimization of large dynamic systems

摘要

A modified version of the adaptive complex method for global optimization is described that parallelizes a procedure that is computationally serial in nature. The modified method assigns n/sub p/ processors for parallel execution. Simulation tests and an implementation with parallel hardware indicate that speed-up factors with the modified method are nearly linear with n/sub p/ up to about 0.1k where k is the number of vertices in the optimizing complex. Since k increases linearly with problem size, the modified method can significantly reduce computation times for large optimization problems.