Recent technological advances in wireless communications make it possible for low cost, low complexity sensor networks to monitor and detect environmental and tactical events. Sensor devices are typically equipped with a low power communication transceiver and a limited processor to facilitate signal processing. One class of widely deployed applications is event-driven: the objective in these applications is to detect and report specific events that occur in a sensor field. This type of application offers minimal traffic load and spends most of its time in a sensing state. Because of energy limitations, techniques for extending sensor lifetimes constitute a major issue in the sensor network research area. The most popular method is to arrange for sensors to spend most of their time in a "sleep state," i.e., a state in which sensing and radio communication devices are turned off. The problem then becomes one of designing protocols determining which sensors are asleep and when. This thesis describes and analyzes a scalable, easily implemented, self-organizing such protocol that yields an energy conserving intruder-detection sensor system. The novelty of our design rests mainly in our application of concepts from cellular automata theory. The system self-assembles periodic wake-sensor barriers (waves) that sweep the sensor field; it is highly effective even in the case of frequent communication failures, sensor failures, large obstacles, and when intruders know sensor locations and the sleep-wake protocols being used. Background and a brief introduction to cellular automata are provided in Chapters 1 and 2. Synchronous sleep-wake protocols are presented in detail in Chapter 3, which is followed in Chapter 4 by the results of extensive experiments. These experiments show how to refine the basic protocols of Chapter 2 in order to enhance the energy-conservation and intruder-detection functions of sensor systems. Chapter 5 shifts to the design of asynchronous sleep-wake protocols, their performance analysis, and their comparative advantages over synchronous protocols.

Our final contribution addresses an intriguing problem that arises in the minimalist sensor paradigm of networks where objects (or targets) are to be sensed and counted in a given field: sensors have limited processing capability and the sensing of targets is error prone (primarily because of environmental features) and simplistic, i.e., it simply counts the number of visible targets within its sensing range. The problem is to take the counts delivered by individual sensors and come up with a total count over the entire field. The obvious difficulty is in handling the over-counting created by intersecting sensing areas; a target is included in the count of every sensor within range of it. To determine the over-counting, we study the correlation patterns in the collection of counts supplied by the sensors. Cumulants comprise the statistical tool used for this purpose; these computations and the classical inclusion-exclusion formula give us the desired total count. Interestingly, the sensing error or "noise" is critical to the method; without the added information created by noise, the method fails.