A computationally efficient multi-model method for uncertain dynamical systems

The multi-model approach stands out as a useful method for identifying systems with uncertainties or unknown parameters. It employs multiple models, each representing a particular value for unknown variables, and integrates their responses at the operational point to control the uncertain/unknown system. Despite considerable efforts in the literature on the multi-model approach, there remains a lack of clear strategies for model selection over all possible models of the system. A key concern is the exponential growth in the number of identification models, rendering the algorithm computationally intensive and unsuitable for real-time applications. This paper introduces a novel approach for selecting a subset of models or optimally positioning them within the parameter uncertainty space formed by different system models, aiming to maximize its coverage and promising significant reductions in computational costs. The effectiveness of the approach is shown in two distinct areas of dynamic systems, namely vehicle handling dynamics known as the bicycle model and power generator dynamics known as swing dynamics.