Title: Time-series estimation using transition networks on realistic, sparse flow data

Abstract: Inspired by biological swimming and flying that utilises distributed haptic sensors to control complex flows, a novel data-driven approach for time-series estimation is proposed. We exploit sparse, real-time pressure sensor input, in combination with pre-trained transition networks, so as to estimate aerodynamic loads in unsteady, highly-separated flows. The transition networks contain the aerodynamic states of the system as nodes along with the underlying dynamics as edges. The method is tested on realistic (i.e. noisy) experimental data from the canonical flow around an accelerating elliptical plate ($Re=194\,000$) at various fixed angles of attack. Aerodynamic loads are then estimated for angles of attack cases that were not included in the training data. The network-based approach provides accurate estimates with relative errors typically averaging less than $10\%$, while their performance deteriorates if the estimated cases are significantly different from the training data. Based on the present findings, transition networks provide a powerful and versatile data-driven approach for real-time signal estimation, opening new avenues for applications to complex and realistic flows.