Landscape Changes and the Performance ofMapping Based Constraint Handling

The Mapping Based Constraint handling (MBC) method for evolutionary search methods was suggested and proved to be a new promising constraint handling method, 6]. The MBC method is based on an assumption that a bounded feasible solution domain can be mapped into a simple hypercubic domain. The result of applying MBC is to create a new unconstrained optimisation problem out of the originally constrained problem. An example of applying MBC is presented. One of the underlying assumption of the MBC is that feasible solution domain(s) can be decomposed and put together in any fashion so that less robust but computationally inexpensive mapping techniques also can be used. Such a decomposition changes the landscape of the problem both in the distribution pattern of search points and in the hamming distance between search points. Although it is very diicult to say in advance whether or not such decompositions will alter the search algorithm's performance, preliminary observations suggest that in practice it has little eeect.