Approximation assisted multi-objective collaborative robust optimization (AA-McRO) under interval uncertainty

Existing collaborative optimization techniques are predominantly focused on singleobjective deterministic optimization. However, a large number of engineering problems involves disciplines or subsystems that are each multi-objective, constrained and have uncertainty. The literature reports on a few Multi-objective Multi-Disciplinary Optimization (MMDO) techniques. However, these approaches in general require a large number of function calls and their computational costs can be exacerbated under input uncertainty. In this paper, we present a new approach which is an improvement to a previous MMDO method. Specifically, the focus of this new MMDO is on the use of approximation under interval uncertainty. The proposed approach is called Approximation-Assisted Multiobjective collaborative Robust Optimization (AA-McRO). AA-McRO is aimed at obtaining Pareto optimum solutions for an MMDO problem whose system performance is relatively insensitive to input uncertainties. An important characteristic of AA-McRO is the use of online meta-modeling for objective and constraint functions to perform system robustness evaluation and subsystem level optimization. The optimal solutions from robustness evaluation and subsystem optimization are used for sampling new points and for updating the meta-models. The advantage of such a sampling technique is that a good approximation of objectives and constraints is achieved with a reasonable number of sampled points. Based on our current results, AA-McRO can be significantly more efficient than a previously reported MMDO method against which a comparison is made using three numerical and engineering examples.