Relaxation of Distributed Data Aggregation for Underwater Acoustic Sensor Networks

PDF

Authors
  1. Rabbat, M.
  2. Coates, M.
Corporate Authors
Defence Research and Development Canada, Atlantic Research Centre, Halifax NS (CAN);McGill Univ Clinic, Montreal QUE (CAN)
Abstract
This aspect of the project is concerned with coordinating the sensors in an underwater acoustic network to collaboratively track an acoustic source or sources. Measurements are taken at each sensor node, and in order to obtain the best accuracy, the measurements should be jointly processed or fused. This requires communication and coordination among the nodes. At the same time, underwater communication is notoriously challenging. Channel conditions change rapidly and high-datarate communications are generally not possible. Consequently, protocols and mechanisms must be used that can adapt to the time-varying and unreliable communication medium. Tracking strategies that rely on transmitting all measurements to a central location are infeasible in this setting, both because of the communication overhead and for robustness concerns. We therefore focus on decentralized methods in which the microprocessor system attached to each sensor execute the tracking algorithms. These local tracking algorithms communicate to share information between neighbouring sensors, distilling the raw measurements into essential summary statistics to reduce the communication overhead. In this project we specifically consider the use of gossip algorithms for processing information over the network because of their robustness to unreliable communication media, and we consider particle filters and random finite set methods for the local target tracking algorithms executed at each node because of their abi
Keywords
data aggregation;distributed computing
Report Number
DRDC-RDDC-2015-C208 — Contract Report
Date of publication
01 Mar 2015
Number of Pages
20
DSTKIM No
CA041884
CANDIS No
803005
Format(s):
Electronic Document(PDF)

Permanent link

Document 1 of 1

Date modified: