Building the Internet of Things: Bottom-up Design of a Wireless Sensor Network.

Author:

Gunnar Schaefer

Abstract:

Based on our theoretical know-how in signal processing and communications and in close collaboration with EPFL’s environmental scientists, we are designing and engineering SensorScope , a complete wireless sensor network (WSN) infrastructure, including a powerful and extensible database back-end. The final goal of our design is that a wireless sensor network may be single-handedly deployed, used, and maintained by scientists of other disciplines, without any prior knowledge in wireless networking.

Major areas of research include

  1. the design of a comprehensive, flexible and extensible, yet highly efficient software architecture, as illustrated in Figure 1,
  2. the evaluation of power-efficiency and reliability trade-offs with respect to
    • multi-hop vs. single-hop wireless networking,
    • packet acknowledgments,
    • low-power-listening vs. always-on listening, and
    • opportunistic routing, as well as
  3. real-time network quality and efficiency monitoring.

The sensor network-side of SensorScope is composed of spatially distributed, autonomous, low-power, low-complexity, radio-enabled embedded devices, called motes. Software, running on the motes, is implemented on top of TinyOS , a pseudo operating system, specifically designed for wireless sensor networks. As shown in Figure 1, we have implemented a lean communications stack, which extends the TinyOS abstraction. Below the application layer, motes on the weather stations and on the base station are running identical code. This approach entails the advantages of a traditional network stack, such as reducing development-overhead and error-proneness, whilst maintaining the required flexibility due to its custom design.

The server-side, on the other hand, which is implemented in C, Python, and SQL, is strongly geared towards extensibility. After the acquired data has been filtered and tagged, it is made available in a plain ASCII-format, so that it may be stored in a variety of formats as well as in an arbitrary number of databases. A rudimentary configuration front-end, displaying real-time network quality information, has been developed as well.

 

SensorScope Global Software Architecture

Figure 1: Global Software Architecture of the SensorScope project.

Collaborations:

Guillermo Barrenetxea , François Ingelrest , Martin Vetterli , Karl Aberer

Project Period, Funding Source:

September 2006 – ongoing,

NCCR-MICS