Systems challenges for medical sensor networks

We identify and address a set of challenges for building medical sensor networks. The first challenge is designing a network architecture for medical sensor networks. We propose CodeBlue architecture to support mobile nodes, device discovery, dynamic queries and multiple sensor types. The key contribution of the CodeBlue architecture is the adoption of publish/subscribe model to allow flexibility and efficiency for sensor data delivery. The second challenge is performing publish/subscribe routing on low power sensors. By implementing an ad-hoc multicast layer for CodeBlue, we identify performance issues that need to be addressed to achieve high packet delivery ratio. TinyADMR, our implementation of ADMR multicast routing protocol, demonstrates that selecting high quality links outperforms using hop counts as the routing metric. We also study scalability limitations of ADMR caused by memory and bandwidth constraints. The third challenge concerns monitoring deployed medical sensor networks. Due to resource limitations, it is not practical to log the network behavior on the sensor nodes. However, it is necessary to reconstruct the network dynamics for evaluation and debugging purposes. We propose a passive monitoring infrastructure, LiveNet, that provides entirely passive logging and analysis framework for reconstructing the behavior of deployed medical sensor networks.