An Adaptive Penalty Function Method for Constrained Continuous Optimization in Population-Based Meta-Heuristic Optimization Methods

The main difficulty encountered in applying the Penalty Function method in handling constrained continuous optimization problems, especially equality constraints, consists in the setting of the penalty coefficients. The paper is proposing a novel Adaptive Penalty Function (APF) method which can be generally applied in conjunction with any population-based meta-heuristic optimization method and which makes the constraints handling process virtually parameter free. The proposed APF method was implemented in conjunction with the 1P-ABC optimization method and was compared with the highly competitive SRES method and with a known dynamic penalty function method on the known G set of COP test problems. The comparison results proved the effectiveness of the proposed APF approach.