Coordination Services for Wireless Sensor Networks
Sameer Sundresh, Gul Agha, Kirill A. Mechitov, Wooyoung Kim, YoungMin Kwon · 2004
Wireless sensor-actuator networks can provide the ability to continuously monitor the integrity of structures in real-time, detect damage at an early stage, and provide robustness in the case of catastrophic failures with a fraction of cost associated with today’s wired networks. However, sensor-actuator networks require a new paradigm of computing—one which explicitly addresses less capable hardware, unreliable communication with limited bandwidth, and severe energy constraints. We are developing algorithms and software tools which will facilitate monitoring and protection of civil structures using such networks. Specifically, our research addresses problems of communication and coordination in this environment. A key idea in our approach is to push the computation to the embedded nodes—thus providing greater robustness through decentralization and probabilistic correctness guarantees through statistical reasoning. For example, a high frequency measurement in a dense sensor array results in an overwhelmingly large amount of data. By aggregating data in the network we eliminate redundant information at the sensor nodes in situ, thus reducing transmitted and stored data and conserving power. Sensor readings need to be correlated in time and space and actuation needs to be synchronized to provide effective monitoring and protection of civil structures. To facilitate such coordination through self-organizing networked sensors, we are developing a range of middleware services such as time synchronization and node localization. Finally, we describe a scalable simulator that we are developing to assist application development on sensor networks. The simulator models three key aspects of a sensor network: the sensor nodes, the data network, and the environment with which the sensors interact. In the future we plan to interface the simulator with actual sensor nodes so that a variety of scenarios can be easily replayed for testing purposes. 1