Underwater Source Localization Using an Artificial Lateral Line System With Pressure and Flow Velocity Sensor Fusion

摘要

Mimicking fish lateral line for hydrodynamic sensing, various artificial lateral line systems have been developed for underwater robotics. However, previous studies of artificial lateral line systems for underwater source localization were based on a single sensing modality of either pressure or flow velocity. Inspired by the functions of flow and pressure receptors of fish, in this article, we developed an artificial lateral line system integrated with pressure sensors and flow velocity sensors. A dual-sensor fusion modality was proposed to locate near-field dipole source by measuring hydrodynamic signals. A multilayer perceptron neural network was constructed to process the pressure and flow velocity signals and to predict near-field dipole coordinates in two and three dimensions. In a 2-D plane, compared with a single pressure- or velocity-sensing modality, the dual-sensor fusion modality reduced the mean localization error by 30%. In a 3-D space, the mean localization error based on the dual-sensor fusion modality was approximately 0.1 body lengths, when considering a source located within one body length from the artificial lateral line. In addition, we studied the influence of the number of sensors on the localization accuracy via analysis of variance. Our experimental results indicated that the bio-inspired dual-sensor fusion modality is beneficial for improving the performance of artificial lateral line systems for underwater source localization.