Table 2.

Statistical comparison of lift force with and without flow sensing

Case	Standard deviation of lift (mN)	Mean accumulated reward
No control	$305\pm 20$	$\mathrm{N}/A$
I (Load)	$191\pm 20$	$-1696\pm 492$
II (Pressure)	$199\pm 33$	$-1860\pm 596$
III (Both)	$176\pm 11$	$-1716\pm 452$

These values were averaged over five agents trained for each of the respective cases, and were calculated over a 4000 time step horizon, which corresponds to approximately 1 minute of testing or four-times the length of a training episode. Uncertainty shown is equal to the standard deviation in the presented value across five separate training sessions. Supplementary Fig. 2 shows examples of the load signal over a 60 s interval for the three different observations.