Artificial Lateral Line Sensor for Robotic Fish Speed Measurement Based on Surface Flow Field Detection and Turbulence Noise Suppression

Compared with traditional underwater vehicles, robotic fish have been receiving increasing attention in recent years due to their excellent maneuverability. However, the characteristics of fishlike undulatory motions and complex underwater working environment have posed significant challenges to robotic fish speed measurement, limiting their autonomy. To overcome these challenges, an artificial lateral line sensor (ALLS) was developed, drawing inspiration from the tactile system of fish. It captured the real-time speed of robotic fish through assessing the deformation of the stressed component under laminar flow impact. To mitigate turbulence disturbances near the ALLS, three flow control components, fairing, flow conditioner, and flow collector, were proposed to attenuate turbulence noise under the viscous effect. Furthermore, a physics-informed calibration method was presented to establish the nonlinear model of ALLS. Specifically, a physical model embedding algorithm based on data resampling was used to mitigate the risk of overfitting by the multilayer perceptron, considering the influence of turbulence disturbance and fishlike undulatory noise. Compared with the classical calibration method based on physical model fitting, the calibration method proposed in this paper reduced the error by 36.0%. Our ALLS’s final mean absolute error was 0.016 m/s with a linearity (R
$^{2}$
) of 0.956. The experimental results indicated that the significant changes in the motion state of robotic fish reduced the accuracy of ALLS. The fusion with other sensors is expected to enhance the robustness of ALLS in the future.
Note to Practitioners
—The motivation of this paper is to design an artificial lateral line sensor based on surface flow field detection and turbulence noise suppression, providing a small-sized and high-precision solution to the speed measurement problem of bionic robotic fish. Most existing ALLS research focused on developing new types of sensors based on different measurement principles, without suppressing the noise caused by fishlike motions, and most experiments were conducted in environments with excessive controls rather than free-swimming robotic fish. To this end, we developed an ALLS based on deformation measurement and proposed three flow control components to make the measured flow more stable. Furthermore, a physics-informed overfitting suppression method was used for the calibration task of the ALLS. A series of simulations and experiments demonstrated that the proposed turbulence noise suppression and calibration method were practical and effective. Hopefully, our methods can provide theoretical and technical guidance to marine engineers for underwater vehicle speed measurement and flow sensing. The recommended flow control component is applicable for conditioning surface fluids in pneumatic control systems. Furthermore, the proposed biomimetic tactile sensor is poised to inspire tactile-based human-machine interaction methods.