| As hardware technology pushes sensor performance near its fundamental limit, further improvement in sensor technology relies on the application of advanced signal processing techniques. The use of sensor arrays is one such example. As a further extension of the concept of sensor array, a sensor network to have greater potential for performance improvement due to its versatility to be deployed in a wide range of applications and its ability to monitor a large physical area. To tap into the full potential of sensor networks, we have to deal with the communication problem. The best long term solution will be based on an adhoc network architecture because of its ability to operate without infrastructural support. However, the stringent constraint on energy resources has become the most significant challenge in sensor network design.; In this dissertation, we anticipate a set of application level tasks and design energy efficient algorithms to fulfill their communication needs. One of the application level tasks is the multi-hop communication between a large number of sensors and an information gathering entity called the USER. We envision that the bulk of the long-range multihop traffic consists of short packets carrying target detection reports or network status queries and replies. Based on this assumption, we present a table-driven, multi-path network structure and a routing algorithm, based on the concept of a sequential stochastic assignment, that has demonstrated, through simulation experiments, significant performance improvement over the classic minimum metric routing algorithm.; Another major application level task in sensor network operation is the formation of an adaptive local network to perform various cooperative signal processing functions on detected targets, which is essential for target identification and tracking purposes. This task involved the selection of a central processing node and the construction of routes from each sensor to the central node. The size, location, and topology of the local network will depend on the signal strength, mobility and location of the target. Such networks typically have a short operational life time, just long enough for the necessary data to be gathered and uploaded to the central node. In this dissertation, we present an energy efficient and highly scalable election algorithm, called Single-Winner Election (SWE), to select the central node and build energy efficient paths from each sensor to the central node. |
