Robust optimization (RO) of designs becomes indispensable to deal with inherent manufacturing uncertainties. Yet, RO is still too expensive to solve complex problems such as the design of helicopter rotor blades. A promising way to reduce this cost is to use the multi-fidelity meta-model-based optimization (MFMBO), which combines low fidelity models (LFM), high fidelity models (HFM), meta-models, design of experiment (DoE), and optimization techniques. The field of MFMBO is not been completely explored despite the high number of proposed strategies. This paper contributes to the MFMBO cost-effectiveness improvement by proposing a new strategy. The main feature of the latter is the mutual adaptive refinement of the HFM and LFM DoE sets. A new idea is proposed to generate initial nested DoEs, which is implemented using an ordinary and an optimal Latin Hyper Cube method. These DoEs are refined adaptively and mutually using an improved MaxMin-distance criterion. The expensive HFM is used only to define a correction C of the LFM. C is used to adjust LFM evaluations during MBO. Meta-models are constructed using the Radial Basis Functions technique. The strategy is tested with two RO techniques, namely, the Non-Dominated Sorting Genetic Algorithm (NSGA-II) and the Multi-Objective Evolutionary Algorithm Based on Decomposition (MOEA/D). The implemented strategy is validated on three different mathematical benchmarks, compared with four different MFMBO strategies and applied on a helicopter rotor blade. This validation shows that the proposed MFMBO is competitive and can produce solutions as accurate as the HFM-based MBO while reducing significantly the overall computation time (up to 40%).