Modern sensor networks utilize sets of small, self-configuring nodes to acquire data about an environment, process it and disseminate the results to other entities as needed or requested. The system must enable efficient and flexible data acquisition and processing and support timely and reliable dissemination of the information. This is a significant challenge because the nodes that make up these networks are battery-powered and severely resource-constrained.

We first study sampling strategies for sensor networks. We concentrate on sample losses that occur in multi-hop clustered architectures for energy-efficient communications. We analyze the case in which an object is moving through this network and is detected by nodes that are close to its path. The temporal sampling of signals emitted by this object are well-approximated as a nonhomogeneous Poisson process. The errors that occur in reconstruction from these samples are characterized.

We then propose an analytical model of and performance evaluation for the energy consumption of data-centric routing protocols. The analytical results allow us to determine which data dissemination approach - high-power broadcast or multi-hop, low-power unicasts - should be used in a given situation. The cases of error-free and error-prone communications are both characterized.

The performance of wireless LAN and sensor networks is then studied experimentally in the eStadium “Living Laboratory”. We report on the design, instrumentation and characterization of this heterogeneous wireless network testbed. We also propose an application framework that provides safety surveillance and emergency response via integration and inter-operability of sensor networks and wireless LANs.