Sensor Networks

Wireless sensor networks are formed by a set of small devices that, in addition to its sensing equipment, are furnished with a radio for wireless communication and a small processor. They are expected to sense the environment cheaply and non-intrusively. It has been used in applications varying from collecting data on habitat environment in the Great Duck Island in Maine to performing measurements aimed at Structure Health Monitoring on the Golden Gate Bridge in California. One key advantage of these sensing devices is the reduced need for maintenance. Ideally, the sensing nodes are deployed once and keep sensing the environment without direct human intervention for long periods of time.

Energy is the scarcest resource on the network. Since communications is one of the most energy intensive tasks, sending each measurement to a central station may greatly reduce the lifetime of the network. Statistics can be employed to make the data collection more efficiently. One such way consists in using statistical modeling to summarize the data of each node in the network individually. Once the parameters of model are fit to the data, they can be transmitted instead of the complete series of measurements. Model selection methods are important here as the complexity of the model can be used to meet precision requirements, bandwidth and power availability. Additional energy savings can be made by exploring any spatial structure existing between the data collected by neighboring nodes.

People Involved

• Guilherme Rocha
• Bin Yu