| Abstract: |
Wireless Sensor Networks (WSN) have reached a level of maturity, at which they have become a feasible option for monitoring processes of interest, when wired infrastructure is not possible.
Due to remote and inaccessible deployment sites of such networks, they can generally not rely on reliable power sources, but require batteries to supply the energy for the system to perform its
intended task. However, the finite energy store imposed by batteries directly limits the system’s performance and lifetime. Ambient energy harvesting has been shown to be a promising way to
boost the performance and lifetime of WSNs. Unfortunately, enhancing a battery powered device with energy harvesting capabilities will by itself neither provide a lower bound on the expected
sustainable performance level, nor guarantee uninterrupted long-term operation. This thesis addresses the design and runtime management of the power subsystem for solar energy harvesting embedded
systems. We demonstrate that for enabling long-term operation of such systems, a paradigm shift both in the design approach, and the runtime management of the energy is necessary. We provide an
end-to-end power management solution, which consists of (i) a power subsystem capacity planning approach, and (ii) two novel dynamic power management schemes that maximize the minimum achievable
performance level, while ensuring that long-term, i.e., multi-year operation can be sustained. Compared to three State-of-the-Art approaches, our solution maximizes the long-term sustainable
minimum system performance or perform equivalently, but require a smaller solar panel and/or smaller battery. Our theoretical results are supported by simulations using 10 years of solar energy
measurements from various geographical locations. To demonstrate the improvements of proper power subsystem design and management, we further present a case study with a real-world WSN deployment
for geoscientific research in a high-alpine environment. Specifically, this thesis presents the following contributions to the State-of-the-Art: • We present a systematic method for power
subsystem capacity planning, i.e., proper sizing of the solar panel and battery, for solar energy harvesting embedded systems. • We present two novel dynamic power management schemes that
enable the system to maximize the minimum performance level at runtime, while ensuring that uninterrupted operation over time periods on the order of years may be sustained. • We present a
light-weight battery State-of-Charge approximation algorithm that can provide runtime information about the battery fill level without requiring special purpose hardware. Moreover, we show that
this approach can be used to infer the harvested energy. • Through extensive simulation we show that our end-to-end solution achieves significantly better results in terms of minimum
long-term sustainable performance level, duty-cycle stability, and overall energy efficiency when compared to 3 State-of-the-Art approaches. • Finally, using a real-world scientific project,
we demonstrate the improvements in system utility for the end-user application that can be gained with our solution. |
