Localization is a key-enabling technology for many applications in underwater wireless sensor networks. Traditional approaches for received signal strength (RSS)-based localization often require uniform distribution for anchor nodes and suffer from poor estimates according to unpredictable and uncontrollable noise conditions. In this paper, we establish an RSS-based localization scheme to determine the location of an unknown normal sensor from a certain measurement set of potential anchor nodes. First, we present a practical path loss model for wireless communication in underwater acoustic environments, where anchor nodes are deployed in a random circumstance. For a given area of interest, the RSS data collection is performed dynamically, where the measurement noises and the correlation among them are taken into account. For a pair of transmitter and receiver, we approximate the geometry distance between them according to a linear regression model. Thus, we can obtain a quick access for the range information, while keeping the error, the communication head and the response time low. We also present a method to correct noises in the distance estimate. Simulation results demonstrate that our localization scheme achieves a better performance for certain scenario settings. The successful localization probability can be up to 90%, where the anchor rate is fixed at 10%.
Keywords: linear regression; localization; received signal strength; relational distance refinement; underwater wireless sensor network.