Cooperative navigation for low-bandwidth mobile acoustic networks

Summary

Abstract

This thesis reports on the design and validation of estimation and planning algorithms for underwater vehicle cooperative localization. While attitude and depth are easily instrumented with bounded-error, autonomous underwater vehicles (AUVs) have no internal sensor that directly observes XY position. The global positioning system (GPS) and other radio-based navigation techniques are not available because of the strong attenuation of electromagnetic signals in seawater. The navigation algorithms presented herein fuse local body-frame rate and attitude measurements with range observations between vehicles within a decentralized architecture. The acoustic communication channel is both unreliable and low bandwidth, precluding many state-of-the-art terrestrial cooperative navigation algorithms. We exploit the underlying structure of a post-process centralized estimator in order to derive two real-time decentralized estimation frameworks. First, the origin state method enables a client vehicle to exactly reproduce the corresponding centralized estimate within a server-to-client vehicle network. Second, a graph-based navigation framework produces an approximate reconstruction of the centralized estimate onboard each vehicle. Finally, we present a method to plan a locally optimal server path to localize a client vehicle along a desired nominal trajectory. The planning algorithm introduces a probabilistic channel model into prior Gaussian belief space planning frameworks. In summary, cooperative localization reduces XY position error growth within underwater vehicle networks. Moreover, these methods remove the reliance on static beacon networks, which do not scale to large vehicle networks and limit the range of operations. Each proposed localization algorithm was validated in full-scale AUV field trials. The planning framework was evaluated through numerical simulation.

Bibtex entry

Downloads

1. Errata