On-Line Box-Complex Optimization

This paper discusses a study of an on-line Box-Complex optimization routine and the stability of the algorithm in the presence of measurement noise. Specifically, the paper explains modifications which must be made to the published versions of the algorithm to allow it to function in a real-time environment. The modified algorithm is then used to optimize a simple three-dimensional function with, and without, measurement noise. The impact of constraints on the optimizer is also considered. Based on this data, it appears that the modified algorithm can, in fact, be used successfully for closed-loop evolutionary optimization (EVOP). It appears that the presence of moderate levels of noise does, as would be expected, require retuning the optimizer, but does not appreciably affect the potential performance.