System-level Support for Macroprogramming of Networked Sensing Applications

Abstract

Wireless sensor networks (WSNs) provide low-cost, embedded sense-and-respond capability, and are therefore an integral part of the vision of pervasive computing. Most research on WSNs to date has focused on the development of efficient protocols for infrastructure establishment. Application development for WSNs is still very daunting for the non-expert. This paper proposes hierarchical layers of abstractions to categorize the functionality of various WSN protocols from a programming perspective. We then address the issue of systemlevel support for programming models in sensor networks, and describe the design of the DART runtime system, which supports a macroprogramming model called the Abstract Task Graph. DART prototypes for two different target platforms have been implemented. The modular structure of the runtime is inspired by our proposed layers of programming abstraction. The design of DART simplifies software synthesis of ATaG programs, and enables near plug-and-play integration of different protocols and services at the lower layers with minimal impact on the higher layers.