Wake interaction provides hydrodynamic gain and flow-aided navigation in fish schools. The lateral spacing Ly and phase angle Φ relative to upstream wake are two important states for downstream swimmers. In this paper, the lateral wake tracking and phase matching of two diagonal flapping swimmers are studied through experiments. Bio-inspired differential pressure (DP) sensing on the downstream swimmer is adopted to capture the wake interaction features. Two DP sensing strategies, the symmetrical differential pressure (SDP) and leading edge differential pressure (LDP), are employed to capture the wake interaction features. SDP measures the pressure difference of two symmetrical ports on the two sides of the downstream swimmer, and LDP measures the pressure difference of leading edge port against the two side ports. One-dimensional convolutional neural networks (1D CNN) with a parallel structure are trained to decode wake interaction states (Ly and Φ) based on DP signals. The 1D CNN model is trained and tested offline and is used to estimate the wake interaction states online. Three demonstrations of online lateral wake tracking and phase matching control are carried out. Compared with SDP, LDP predicts Ly and Φ more accurately. It is found that the downstream wakes are more compact after control, which is consistent with high propulsive efficiency mode.
Read full abstract