Into the wild: taming uncertainty in perpetual mobile networks

Catalyzed by advances in low-power electronics and energy harvesting, we can now build mobile systems that operate perpetually, sensing and streaming data to scientists. Unfortunately, current mobile software systems are not well suited to the combined challenges of node mobility, unpredictable connectivity, and variable energy. Under such conditions, an efficient and robust system must adapt to its changing network and energy environment. Writing adaptive code and understanding the impact of local adjustments on global network performance present significant challenges. This talk describes research aimed at significantly improving the programmability and performance of perpetual mobile systems, in the context of two ongoing collaborative wildlife studies, focused on endangered turtles and invasive mongooses. At the core of this effort is the Eon programming language and runtime system which significantly simplifies programming energy-aware code by separating adaptation from program logic. Additionally, extensions to Eon's local adaptation features provide support for efficient, balanced, and fair data collection in energy-constrained mobile networks.