Distributed underwater glider network with consensus Kalman filter for environmental field estimation

Abstract

A distributed coordinated dynamic sensor network for optimal environmental field estimation is proposed and tested on simulated and real data. The architecture is used to distribute the estimation of 3D timevarying fields on a network of underwater gliders in which communication constraints are one of the main limiting factors for achieving complete network automation and de-centralization. The dynamic network is integrated with a network of relay nodes that the agents can reach at the surface through satellite or radio links. Local field statistics are estimated by a Kalman filter-based algorithm by agents and the information in the network is shared by using the asynchronous consensus algorithm. The system is tested on realistic scenarios that are simulated by true oceanographic forecast models. A heterogeneous network of underwater and surface wave gliders is used to estimate 3D sea water temperature fields. Simulation results show that the network achieves a consensus with an average root mean square error at steady state that is below 0.5 °C.