Traditionally, models of physical processes are derived from first principles and/or expert knowledge from researchers in the field. The same holds in the field of control design, where often simple explicative control designs are preferred for their transparency. However, the predictions made by first principles models, and the control performance of simple controllers, is often in-adequate. Rather than throwing these models or controllers out, we developed a model structural adaptation method (MSAM) to improve such systems. This method proposes small changes to the structure of a model and adapts its changes using a novel gradient-based iterative method. Using this method we were able to adapt the accepted model of vortex-induced vibration (a fluid-structure interaction phenomenon) from literature to vastly improve its predictive capacity with small changes to its structure [1]. Similarly, we were able to adapt simple control designs to accommodate non-linear environments [2].

  1. La Cava, W. G., Sahare, K., & Danai, K. (2017). Restructuring Controllers to Accommodate Plant Nonlinearities. Journal of Dynamic Systems, Measurement, and Control, 139(8), 081004. link

  2. La Cava, W. G., & Danai, K. (2016). Gradient-based adaptation of continuous dynamic model structures. International Journal of Systems Science, 47(1), 249-263. link