Key points are not available for this paper at this time.
Abstract Interpolatory reduced-order models (ROMs) rapidly approximate high-dimensional systems by constructing and blending low-order systems at selected interpolation points. This avoids repeated projection of full-order operators and yields orders-of-magnitude speed-ups, making these methods well-suited for real-time prediction, optimization and control. Existing surveys examine individual techniques in isolation and employ advanced differential-geometric or system-theoretic formalisms, limiting accessibility to the broader engineering and computational communities. In contrast, this review proposes a novel taxonomy categorizing all interpolation-based ROM families without cataloguing every variant. For each family, the core mathematical ideas are conveyed through pseudocode and originally integrated into a unified six-stage workflow, accompanied by explicit O(⋅) cost analyses. We then highlight key methodological extensions required for complex, real-world applications focusing on computational mechanics. A combined quantitative and qualitative assessment against six canonical benchmarks and traditional projection–based ROMs provides performance insights. The result is a practical roadmap enabling engineers and scientists to select, adapt and deploy interpolation-based ROMs with clarity and assurance.
Sreekumar et al. (Fri,) studied this question.