Tones for Real: Managing Multipath in Underwater Acoustic Wakeup∗ ISI-TR-659

The principles of sensor networks—low-power, wireless, in-situ sensing with many inexpensive sensors—are only recently penetrating into underwater research. Acoustic communication is best suited for underwater communication, since other methods (optical and radio) attenuate very quickly. Yet acoustic propagation is five orders-ofmagnitude slower than RF, so propagation times stretch to hundreds of milliseconds. A new generation of underwater acoustic modems have added low-power wakeup tones that combat the energy consumption acoustic modems would waste on idle listening. Recently, these tones have been used as an integral part of application layer and MAC protocols. While all wireless data-networks suffer from multipath interference of received data, in this paper, we show that due to large acoustic propagation delay tone echoes cause a unique interference, tone self-multipath, for tonebased protocols. To address this interference we introduce Self-Reflection Tone Learning (SRTL), a novel approach where nodes use Bayesian techniques to discriminate selfreflections from noise and communication from other nodes. We present detailed experiments using an acoustic modem in two physical environments to show that SRTL’s knowledge corresponds to physical-world predictions, that it can cope with reasonable levels of noise, and that it can track a changing multi-path environment. Simulations confirm that these real-world experiments generalize over a wide range of conditions.