The need for comprehensive monitoring and analysis of complex environments has brought wireless sensor networks (WSNs) to the forefront of research. To be effective in practice, WSNs need to capture system state, monitor complex properties, and model physical phenomena accurately and robustly, without any biases. Currently, most efforts in this area employ parametric models and other a priori assumptions that postulate known phenomena and system properties. The characteristics of emerging complex applications, however, are often unknown and hard to predict. Thus, optimizations using models based on a priori assumptions often have limited effectiveness.
This research plan develops novel data-driven modeling methods, scalable modular structures, and application optimization algorithms that comprehensively abstract, capture and operate on traces of deployed WSNs. While the research thrusts are intended to provide applicable models, development tools, and system software for WSNs, they simultaneously advance the theory and practice of statistical modeling and optimization algorithms. The models and algorithms are coordinated: the models are built so that they facilitate consequent optimizations, while the optimization algorithms are created in such a way that they take into account the structure and uncertainties of the models. The results are intrinsically interdisciplinary and have impact on a broad range of other fields. This project integrates research and educational activities by engaging students in the research and by incorporating the research into both undergraduate and graduate levels. The research results and educational material are promptly disseminated via academic publications, the Connexions platform, and other WWW sources.
