Flexible artificial lateral line based on luminous flux for underwater velocity vector estimation.

When fish swim, a specific 'water flow field' forms around their bodies. The lateral line system can provide real-time feedback on flow field variations through the stimulation of hair cells by water currents. Therefore, this paper developed a flexible artificial lateral line (ALL) sensor unit based on the principle of luminous flux that measures flow velocity vector. The sensor employs a dual-layer inverted cup-shaped rocker design. Water flow impacts the rocker, compressing the flexible silicone spring and converting flow velocity changes into variations in luminous flux received by photosensitive units in multiple directions, thereby achieving local flow velocity vector sensing. To address traditional modeling challenges posed by large deformations, nonlinear mechanics, and coupling characteristics in flexible materials, a deep neural network-based flow velocity perception algorithm named CLANN is proposed. This algorithm not only facilitates calibration of flow velocity vectors but also enables multi-sensor data fusion for more accurate flow velocity prediction. Finally, the proposed ALL sensor unit was integrated onto an underwater robotic platform. Under attitude disturbance conditions, the sensor was fused with an inertial measurement unit to achieve multi-sensor fusion estimation of the robot's velocity vector. Results indicate that the measured velocity vector exhibits a mean absolute error of 0.048 m/s in magnitude and 16.49° in direction, with a linearity coefficient ([Formula: see text]) of 0.896. Furthermore, the robot can estimate its own trajectory under different motion states with an error of 0.284 m.