Intelligibility and control for context-aware Internet of Things applications. (R-8150)

The Internet of Things (IoT) is growing rapidly. By the end of 2020, it will consist out of more than 50 to 100 billion physical devices, each capable of gathering context about our surroundings and affording new kinds of interactions. On this scale, advanced algorithms, such as machine learning techniques, are required to process and interpret the data that is gathered. Since devices that have sensors are ubiquitous nowadays, context-aware applications have become a reality. However, the behavior of context-aware application is rarely clear to the user, making them hard use and to control. Context-aware IoT applications are even more difficult to use since the computing power and data is distributed in physical space, and because of the degree of dynamism and ephemerality (IoT devices that leave or join the network). My aim is to explore and create approaches, techniques and tools that allow end-users to make context-aware IoT applications intelligible (easier to understand), control and predict, and therefore more usable. I first identify what users need and want to know of machine learning techniques that are used to interpret context data. Next, I investigate how this information can be optimally communicated to the user for both in situ output using the IoT devices themselves, and for external output using augmented reality. Finally, I define a process to generate meta-user interfaces: interfaces to control the behavior of sets of IoT devices.

Keywords:HUMAN-COMPUTER INTERACTION

Disciplines:Applied mathematics in specific fields, Computer architecture and networks, Distributed computing, Information sciences, Information systems, Programming languages, Scientific computing, Theoretical computer science, Visual computing