Towards practical evolutionary robust multi-objective optimization

Multi-objective optimization methods focus towards finding the high-performing Pareto-optimal solutions, without considering their sensitivity to minor deviations from their original values. It is a fair assumption that practical realization of optimal solutions is often accompanied by minor differences from the exact numerical results produced by an optimizer. Taking this factor into account, Robust Optimization methods seek to find high-performing solutions which are also less sensitive to such deviations. In this work, we have proposed strategies to minimize the number of function evaluations (which can be an expensive enterprise) to enhance one of the earliest proposed methods for robust Multi-objective Optimization. Our focus is on constrained Multi-objective optimization problems and hence we make use of the Infeasibility Driven Evolutionary Algorithm (IDEA), as the Evolutionary Multi-objective Optimizer. We take up three constrained Multi-objective engineering design optimization problems from the literature as the test-bed for our experiments and present results on the same.