Tones for Real: Managing Multipath in Underwater Acoustic Wakeup

The principles of sensor networks, with inexpensive low-power, wireless, in-situ sensing, are today going underwater with acoustic communication. With large acoustic delays, new modems use techniques such as low-power wakeup tones to reduce energy otherwise wasted on idle listening, and recent applications and MAC protocols integrate tones in their operation. While all wireless data-networks suffer from receiver-multipath, we show that tone-echos cause self-multipath for tone-based protocols. We address this interference with Self-Reflection Tone Learning (SRTL), a novel approach which uses Bayesian techniques to discriminate echos from noise or communication from other nodes. We present experiments from controlled environments to show that SRTL's knowledge corresponds to physical-world predictions, and that it can tolerate random noise up to at least two false triggers per sample.