On Distributed Sensor Fusion in Batteryless Intermittent Networks

Distributed and collaborative computation has never been considered before in networks of batteryless sensors. This can bring many advantages for applications (e.g. longer transmission ranges, lower network costs), however introducing new research challenges. In this paper, we focus on the well-known distributed sensor fusion but in an intermittently-powered batteryless sensor network. The goal is to estimate a parameter collaboratively by considering individual sensor measurements. We show that, even though the nodes stop operation with high probability due to random power failures and they neither communicate with their neighbors nor perform computation most of the time, the simplest implementation of the fully-distributed sensor fusion based on average consensus improves the overall estimation quality of the network considerably. In the light of this, we anticipate that if harvested energy is used efficiently so that nodes have more opportunity to receive and send packets, existing fully-distributed protocols can be implemented with tiny modifications in networks of batteryless sensors.