Abstract:
A Wireless Sensor Network (WSN) consists of spatially distributed sensor nodes which
monitors environmental conditions such as temperature, humidity, sound or pressure, etc.
Recently there is increasing need to design Wireless Sensor Network systems that support
applications with intensive monitoring and control activities. This application class often
has significant data collection and processing requirements, requiring increased levels of
energy consumption as compared to other WSN applications. Further, many deeply embedded
WSN systems with these data collection and processing requirements are expected
to operate without manual battery recharging for several decades, and therefore require
energy harvesting techniques. For this class of systems, there are currently few e↵ective approaches
that balance careful energy management with high performance communication
and computation requirements.
My dissertation addresses the above problem. Specifically, I propose a set of algorithms
and control methods for energy management in performance-sensitive WSN systems, and
harvesting-aware rate allocation for application utility maximization. First I formally define
the problem of energy harvesting-aware energy management as two optimization problems,
one for individual sensor nodes and another for multi-hop sensor networks. I propose
energy management algorithm to solve both problems optimally and efficiently. These solutions
combine two energy saving techniques, Dynamic Voltage Scaling (DVS), and Dynamic
Modulation Scaling (DMS), alongside with energy harvesting techniques. I then address a
harvesting aware rate allocation problem with the objective of utility maximization. The
problem is solved with an optimal centralized algorithm and a distributed algorithm.
I conducted extensive simulation-based experiments to evaluate the e↵ectiveness of my
proposed algorithms. Specifically I developed simulation software using TOSSIM, the standard
WSN simulator, and EPANET, a public domain, water distribution system modeling
program. This software simulates in high fidelity the computation and communication activities
of WSN nodes, and considers a variety of network setups, energy harvesting profiles
(solar and water), and application scenarios, etc. My algorithms are implemented within
this simulation environment and compared against a series of rival algorithms under various
experimental setups. Extensive simulation results demonstrate the significant advantage of
my algorithms over the rival algorithms.
