Comprehensive Development Support for Wireless Sensor Networks

Wireless sensor networks (WSNs) are heterogeneous networks that comprise tiny, resource-constraint sensor nodes, gateways and backend systems. Embedded into the environment, sensor nodes measure ambient parameters such as temperature, motion, etc. and gateways provide the integration with traditional networks while backend systems process and visualize received data. Application development for WSNs is complex as it unites the challenges of distributed applications and embedded programming. In addition, heterogeneity, unpredictable environmental influences and the size of the networks further complicate this situation. To master these issues, developers typically perform multiple, distinct steps: simulations support testing and optimizing applications in a controllable environment and visualizations on remote computers serve as surrogates for the display-less simulated or real sensor nodes. Finally, developers port the application to the sensor nodes, gateways and backend systems. In this work, we propose a novel development framework that integrates and improves the formerly separate steps of simulation, visualization and application development for heterogeneous WSNs. It consists of the four components Shawn, SpyGlass, Fabric and microFibre. Shawn is a novel simulation tool for the design and optimization of applications prior to their real-world deployment. The level of detail provided by traditional tools that mimic the real world as closely as possible comes at a price and it is often not required. Instead, Shawn uses abstract and exchangeable models allowing users to focus on their research goal while operating at orders of magnitudes higher speeds. SpyGlass is a modular, extensible and platform-independent visualization environment. Unlike existing tools, it provides the infrastructure for arbitrary visualizations independent from the actual WSN platform. Fabric provides the link between simulation, visualization and heterogeneous WSN devices. It is a novel system for the generation of application- and data type-specific middleware for heterogeneous target platforms. Developers provide a data type definition augmented with annotations. Based on this input, multiple modules conjointly generate optimized middleware instances. microFibre, a new (de-)serialization scheme implemented as such a module, shows how considerable bandwidth savings are achieved without requiring manual optimization.
These four components are contained and valuable in themselves. Combined, they constitute a sound development framework that supports developers at all stages from an initial idea to a final WSN application. We present experimental measurements and a case study showing how this framework facilitated the realization of the real-world WSN project MarathonNet.