Measuring Low-Power Wireless

We measure the temporal and spatial behavior of modern low-power wireless radios. Current work focuses on 802.15.4, due to its increasing importance in sensornets, home automation, and the 6lowpan community. Measuring many platforms and environments can separate hardware-specific and enviromental effects from more general ones. The results enable better theoretical formulations, improve protocols, and lead to better and more accurate simulation tools.

MNet Architecture

The MNet architecture explores how to architecturally improve sensor networking. The core principle of the architecture is to minimize the energy cost of diagnosing network behavior, transforming the typical "black box" embedded sensornet to a well-understood and transparent system that is easy to optimize, manage, and deploy.

Operating System/Language Co-Design

TinyOS is the de-facto standard OS for embedded sensor networks. The ability to quickly build large and complex systems raises novel challenges in isolating independent subsystems from one another in the absence of virtualization. TinyOS's component and concurrency models provide greater program structure for run-time checks and compile-time verification. Co-designing the language and OS in parallel allows us to not only add language primitives to help analysis, but also to design the OS so that it is more checkable.

Wireless Implanted Networks

We are currently working with neuroscientists at UCSF to explore the systems software challenges that implanted medical devices impose. Long-term implants must last for years on as small a battery as possible but must also be able to react very quickly to environmental changes. The short-term goal is to enable a whole new class of scientific experiments. The long-term goal is to design an operating system for these devices.

Storage-Based Applications

Processors, radios, and batteries are not improving greatly in cost or energy efficiency, but non-volatile memory is. The ability to cheaply store large amounts of data locally has the possibilty of changing sensor networks from communication-centric to storage-centric devices. As embedded sensor nodes are not directly accessed by end-users, however, they require very different storage abstractions. We are exploring storage-centric networks through application deployments (field geology and Stanford's Green Dorm project) as well as low-level storage abstractions (the TINX index).