Bioinspired Sensing of Undulatory Flow Fields Generated by Leg Kicks in Swimming

TL;DR

This work tackles sensing the undulatory flow fields produced by human leg kicks using a bioinspired artificial lateral line (ALL). It introduces a three-module sensing framework that fuses time-domain (1DCNN-BiLSTM) and time-frequency (STFT-2DCNN) features via an attention-based fusion mechanism to improve kick-pattern recognition and kicking-leg localization. Experiments with a 1:5 scale leg model demonstrate high kick-pattern identification accuracy (≈95.7%) and robust localization performance (RMSE reductions of over 20–60% versus baselines), with performance strengthening at higher kick frequencies due to increased flow velocity and SNR. The approach shows promise for underwater human-robot interaction and cooperative tasks, enabling leader-follower formations and assistive capabilities for swimmers in complex aquatic environments. Future work will extend to 3D real-world swimming scenarios and sensor fusion with IMUs to further enhance robustness and applicability.

Abstract

The artificial lateral line (ALL) is a bioinspired flow sensing system for underwater robots, comprising of distributed flow sensors. The ALL has been successfully applied to detect the undulatory flow fields generated by body undulation and tail-flapping of bioinspired robotic fish. However, its feasibility and performance in sensing the undulatory flow fields produced by human leg kicks during swimming has not been systematically tested and studied. This paper presents a novel sensing framework to investigate the undulatory flow field generated by swimmer's leg kicks, leveraging bioinspired ALL sensing. To evaluate the feasibility of using the ALL system for sensing the undulatory flow fields generated by swimmer leg kicks, this paper designs an experimental platform integrating an ALL system and a lab-fabricated human leg model. To enhance the accuracy of flow sensing, this paper proposes a feature extraction method that dynamically fuses time-domain and time-frequency characteristics. Specifically, time-domain features are extracted using one-dimensional convolutional neural networks and bidirectional long short-term memory networks (1DCNN-BiLSTM), while time-frequency features are extracted using short-term Fourier transform and two-dimensional convolutional neural networks (STFT-2DCNN). These features are then dynamically fused based on attention mechanisms to achieve accurate sensing of the undulatory flow field. Furthermore, extensive experiments are conducted to test various scenarios inspired by human swimming, such as leg kick pattern recognition and kicking leg localization, achieving satisfactory results.

Figures (13)

• Figure 1: Illustration of the design of the scaled leg model mimicking human swimmers. The leg model is 3D printed, and actuated using two servo motors, simulating the kicking motion of swimmers wearing fins.
• Figure 2: Schematic diagram of the ALL-based flow sensing for the undulatory flow field generated by leg kicks. The ALL system samples flow pressure distribution of the downstream wake of swimmer's leg kicks for sensing tasks such as kick pattern recognition and kicking leg localization. (a) Top view illustrating the dimensions of the leg model and ALL system, and (b) side view illustrating the flutter kick pattern.
• Figure 3: The proposed network architecture of the sensing framework for undulatory flow generated by swimmer's kicks. This framework consists of three primary modules, including the time-domain feature extraction module, the time-frequency domain feature extraction module, and the attention-based feature fusion module.
• Figure 4: Illustration of the STFT spectrums of the distributed pressure measurements sampled by the ALL system in the undulatory flow field. The $x$-axis represents time, while the $y$-axis represents frequency. Each column shows the frequency spectrum for longitudinal displacements $L_y$ of 20 mm or 40 mm under the same leg kick pattern ${s}_i \in \boldsymbol{S}$, and each row shows the spectrum for different leg kick patterns ${s}_i$ at the same $L_y$.
• Figure 5: Illustration of distributed pressure measurements collected by three pressure sensors ($P_1, P_2, P_3$) in the ALL system. Subfigures 5(a), 5(b), and 5(c) show the data from sensors $P_1$, $P_2$, and $P_3$, respectively. Different colored curves represent various leg kick patterns ${s}_i \in \boldsymbol{S}$: purple dashed line for ${s}_1$, red dashed line for ${s}_2$, blue dashed line for ${s}_3$, yellow line for ${s}_4$, green line for ${s}_5$ and orange line for ${s}_6$. Each figure contains a top subfigure showing pressure data at a longitudinal displacement $L_y$ of 20 mm and a bottom subfigure showing data at $L_y$ of 40 mm.