Underwater Wireless Sensor Networks Using Distributed Fusion of Optimally Quantized Estimations from Local Trackers

Underwater wireless sensor networks (UWSNs) have come to rely heavily on localization technology due to their usefulness in a wide variety of contexts. Multi-sensor underwater surveillance has been an important area of study for both military and civilian uses. UWSNs employ measurement quantization to send data from individual sensors to the fusion Centre in order to accomplish centralized tracking/fusion despite bandwidth constraints. Yet, due to its non-linear nature, azimuth information is difficult to quantify at the measurement level. This work presents a solution by distributing the tracking and quantizing the local estimates in order to increase available bandwidth and decrease the computational burden. Yet, underwater acoustic communications have a very low bandwidth because of the temporal and geographical diversity of sound channels. Data quantization helps decrease the length of information sent from local sensors to the fusion Centre, which can help alleviate network congestion. Although quantization might lessen the burden on a network's budget, the information loss that ensues from the process also leads to subpar tracking performance. This work presents a target tracking system based on optimum quantization to address this issue. Namely, it reduces the extra covariance introduced by quantization in low-bit quantized measurements, hence enhancing their tracking efficiency. The simulation results show that our method outperforms the standard uniform quantization-based target tracking technique and that the increase in data length has a negligible impact on its performance. The proposed technique has been tested in simulation and shown to effectively reduce localization errors and reduce the burden of data transfer in massive 3D submerged sensor nodes.