An improved differential evolution for constrained optimization with dynamic constraint-handling mechanism

In this paper, an improved Differential Evolution (DE) with a self-adaptive strategy to determine the control parameters is proposed to solve constrained real-parameter optimization, combined with the dynamic constraint-handling mechanism. It is implemented by restating the single-objective constrained optimization as a set of single-objective unconstrained problems and dynamically assigning to the individual adaptively as its fitness, and the self-adaptive strategy of control parameters based on the intrinsic structure analysis of differential vectors is use to solve each unconstrained optimization problem individually. This approach is tested on a suit of test problems proposed for CEC2010 competition and special session on single objective constrained real-parameter optimization. The result indicates the combination of dynamic constraint-handling mechanism and self-adaptation of control parameters in DE will outperform using the former solely for constrained optimization.