Interpretable Aeroelastic Models for Control at Insect Scale

When I cut off states, I can better understand what's happening in the system. A dotted black line represents reality, our full-scale model that we are comparing our data to. However, a different maneuver is hidden in the background. The deformation is easier to see in this maneuver, with a dark red line capturing low-rank behavior but not bending or wake vorticity.

The rank four model is not capturing all the necessary elements, as shown by the blue line. The rank six model is looking okay for coefficient of lift, but it is important to include deformation in the model to identify missing elements. Moving to rank eight, we are not perfectly tracking our full order model but are doing a good enough job for feedback control to handle remaining errors.

Interpreting the coefficients from the model is crucial. The test maneuver of pitching up, holding, and pitching down helps us observe acceleration and quasi-steady behavior. Different test maneuvers may be used in various situations. The rank nine model is tracking the data well.

The coefficient of lift and curvature are shown in the model, with good accuracy in rank eight and nine. It is important to consider the interpretation of the data and how it can be used in different scenarios.