The Lab of Things (LoT) may sound like something you’d find in a sci-fi movie, but it is a lot more practical than that: it’s a research platform that makes it easy to deploy interconnected devices in multiple homes, then share your individual research data with other investigators, turning it all into a large-scale study. The LoT thus enhances field studies in such diverse disciplines as healthcare, energy management, and home automation. It not only makes deployment and monitoring easier—it also simplifies the analysis of experimental data and promotes sharing of data, code, and study participants, further lowering the barrier to evaluating ideas in a diverse set of environments where people live, work, or play.
One key to the success of the LoT is the involvement of the academic research community in developing extensions to the LoT infrastructure. These extensions can be in the form of drivers, applications, and cloud components such as analytics.
Shortly after we released the LoT in July of this year, a group of students from University College London (UCL) started poking around the code and got inspired: they’ve developed an analytics engine to scrutinize data collected from experiments and research applications running on the LoT. And this is no slouch of an engine, either. Among other things, it:
- Permits easy integration of different platforms beyond the LoT, including HomeOS and the Internet of Things, allowing them to easily send device data to the engine.
- Provides analytics for data collected from various field studies and for a wide range of use-case scenarios.
- Offers HTTP-based APIs, which makes it connect to an internal/external client—a Windows 8 application, for example.
- Employs a simple interface, enabling users to easily query and analyze data.
The analytical models provided by the UCL Lab of Things Analytics Engine allow the user to evaluate usage patterns of devices, compare data sets, and find anomalies. The engine also has the capability to run custom R scripts, thereby enabling users to employ statistical models beyond those directly implemented in the engine.
—Arjmand Samuel, Senior Research Program Manager, Microsoft Research Connections