To achieve time synchronization in underwater networks, a linear regression is often applied over a set of sending-receiving timestamps, assuming the participating timestamps are consistent. In this paper, we show that collisions, which are not uncommon during signal exchange in underwater environment, would lead to inconsistent timestamps. These inconsistent timestamps are outliers. However, existing synchronization approaches ignore the presence of outliers. To obtain a reliable synchronization, we propose a robust algorithm that identifies and eliminates the outliers by employing Cook's distance before applying linear regression. We justify the proposed algorithm in a mobile underwater environment through extensive simulations.
展开▼
