Towards the improved robustness of acoustic cooperative localization

Abstract

In this thesis, the relationship between packet loss and the accuracy of an autonomous surface craft's trajectory estimate is explored. In experimental conditions, it is common to lose subframes of acoustic packets, or even entire packets during transmission. Since packets are often used to encode localization information of and range measurements to the sender, lost packets result in poorer navigation of the receiving vehicle. Trajectory estimates, computed using nonlinear least squares optimization, are computed for a variety of client node/server code configurations, and for a series of packet transmission success rates in one of those configurations. Though it is possible for a vehicle to receive some subset of the total number of frames in a packet, partial packets are not usable in current methods. This thesis proposes a method of preparing acoustic packets containing navigation information, so that each frame of a packet is independently useful to a receiver. It is shown in this work that in incorporating navigation information from partially received packets as well, a more accurate trajectory estimate of a non-GPS-aided vehicle is achieved.