Evolutionary computation at the edge of feasibility

Numerical optimization problems enjoy a significant popularity in evolutionary computation community; all major evolutionary techniques use such problems for various tests and experiments. However, many of these techniques (as well as other, classical optimization methods) encounter difficulties in solving some real-world problems which include non-trivial constraints. This paper discusses a new development which is based on the observation that very often the global solution lies on the boundary of the feasible region. Thus, for many constrained numerical optimization problems it might be beneficial to limit the search to that boundary, using problem-specific operators. Two test cases illustrate this approach: specific operators are designed from the simple analytical expression of the constraints. Some possible generalizations to larger classes of constraints are discussed as well.