Browse Theses

Wireless sensor networks (WSNs) have recently emerged as a promising platform for many non-traditional applications, such as wildfire monitoring and battlefield surveillance. Due to bug fixes, feature enhancements and demand changes, the code running on deployed wireless sensors often needs to be updated, which is done through energy-consuming wireless communication. Since the energy supply of battery-powered sensors is limited, the network lifetime is reduced if more energy is consumed for software update, especially at the early stage of a WSNs life when bug fixes and feature enhancements are frequent, or in WSNs that support multiple applications, and frequently demand a subset of sensors to fetch and run different applications.

In this dissertation, I propose an energy-efficient software update management framework for WSNs. The diff-based software update process can be divided into three phases: new binary generation, diff-patch generation, and patch distribution. I identify the energy-saving opportunities in each phase and develop a set of novel schemes to achieve overall energy efficiency.

In the phase of generating new binary after source code changes, I design an update-conscious compilation approach to improve the code similarity between the new and old binaries. In the phase of generating update patch, I adopt simple primitives in the literature and develop a set of advanced primitives. I then study the energy-efficient patch distribution in WSNs and develop a multicast-based code distribution protocol to effectively disseminate the patch to individual sensors.

In summary, this dissertation successfully addresses an important problem in WSNs. Update-conscious compilation is the first work that compiles the code with the goal of improving code similarity, and proves to be effective. The other components in the proposed framework also advance the state of the art. The proposed software update management framework benefits all WSN users, as software update is indispensable in WSNs. The techniques developed in this framework can also be adapted to other platforms such as the smart phone network.