On mobile and wearable devices, many sensors provide connections between the real world and the digital realm, enabling computing access to many forms of raw data, including audio, video, inertial motion sensing, and touchscreen control. Through continuous sensing, mobile device services play a large role in personal computing, assistive technology, and many other fields. However deriving expressive input from high data rate sensors currently suffers from fundamental limitations to the efficiency and performance of handling densely pipelined streams of raw data. This is largely because sensor configuration and control in modern mobile systems induces significant system latency, preventing immediate control of sensor operation and limiting algorithmic expressiveness.
This project pursues the transformation of operating system and sensor architecture to provide system support for reactive sensor operation: using a low-level process to observe and react to context, immediately executing developer-defined sensor configuration, sampling, and processing before providing data access at the application layer. This paradigm provides developers with expressive access to sensor operation while maintaining energy-efficiency and high-performance in the mobile system. Thus, the project discerns systems principles to model and design reactive programming abstractions for sensor processing and control through operating system services, system libraries, daemon processes, and sensor system architecture modifications.
