Distributed estimation of the pelagic scattering layer using multiple buoyancy controlled underwater vehicles

This paper introduces an adaptive sampling strategy for a multi-vehicle sensor network to explore an underwater biological system known as the pelagic scattering layer, which is a region in the water column with a high density of marine organisms that reflect sound. Ever-changing ocean flow presents a challenge to multi-vehicle coordination in this process. The presence of an ocean flow field may disrupt inter-vehicle spacing so that the group loses communication, strays from a desired formation, or ends up with reduced area of coverage, especially for robotic drifting vehicles whose motion is largely influenced by the ambient flow field. However, those vehicles may also take advantage of the vertical variation of the flow to form a desired cohesive configuration since they can tune their vertical position via depth control. A motion-control algorithm, therefore, is a key element of the adaptive sampling strategy. The paper derives a decentralized coordination algorithm to stabilize a cohesive formation in a two-dimensional flow field with an initial unknown vertical distribution. The algorithm works with a distributed extended Kalman filter that generates a local estimate of the flow from pairwise range measurements between vehicles. Another component of the sampling strategy is the modeling of organism density dynamics using estimation of the density with local optical measurements, such as from onboard cameras. The density evolution is estimated using an ensemble Kalman filter and the results are fed into the motion-coordination algorithm. Numerical simulations illustrate the efficacy of this strategy and motivate ongoing and future efforts to extend the result to a three-dimensional geophysical flow. •A decentralized depth control algorithm guides multiple drifting vehicles to varying depths, ensuring cohesive, collision-free formation in shear flow.•Buoyancy control extends to unknown shear flow using extended Kalman filter for parameter estimation, enabling formation.•Ensemble Kalman filter estimates scattering layer density from depth predictions using operator and local measurements.