SENSOR NETWORK OPTIMIZATION BASED ON TARGET APPLICATION
Numerous applications rely on data obtained from a wireless sensor network where application performance is of utmost importance. However, energy usage is also important, as sensors in the network typically run on batteries. Constantly gathering all the data from all the sensors can be cost prohibitive and often unnecessary. Researchers have found that using a large number of sensors, in some cases, did not always lead to better application performance. Many times the applications performed better using a subset of the available sensors. In this dissertation, the main objective is to develop methods for learning to reconﬁgure the wireless sensor network to maximize the target application performance. We propose a novel heuristic algorithm, Sensor Network Conﬁguration Learning (SNCL), that dynamically reconﬁgures the sensor network to maximize the performance of the target application. We evaluate SNCL on synthetic and real datasets and show that our method is better than other popular methods. We show our method, using a variety of synthetic data experiments, has polynomial (low-order quadratic) scalability in the worst case. We prove that in some classes of scenarios our method is guaranteed to converge to the optimal solution. The benefits that result from using SNCL stem from our multi-objective optimization approach utilizing a composite reward function that can take into account both application performance and network performance.