A Walk through Sensor Network Programming Models

Introduction

Today, cheap and smart devices with multiple on-board sensors, networked with wireless links are available from research groups and companies. Due to the small size of these devices (also known as motes) pervasive computing concept is now far from just an imaginary idea. Sensor networks (Akyildiz, et al., 2002a; Yick, et al., 2008) gather data from the physical world and deliver it to a central node or sink where the data can be processed and analyzed. There are many applications of sensor networks, ranging from environmental monitoring to precision agriculture, animal tracking, volcano monitoring, weather forecasting, military situation awareness, structural monitoring, coal mine safety, medical emergency management, and traffic control. A sensor network typically has the following characteristics:

• •

  Homogeneous or heterogeneous structure, comprising a large number of sensor devices, a small processing unit, and a wireless communications mechanism.

• •

  Deployment of cheaper devices, rather than more powerful devices since data density and redundancy are key requirements (Costa, et al., 2007).

• •

  Little or no infrastructure, beyond base station data sinks (Yick, et al., 2008).

• •

  Short, low speed communication range between devices and the use of multi-hop routing to deliver messages. Communications may be asymmetric and unreliable (Woo, et al., 2003).

• •

  Battery-powered devices with a limited energy budget, possibly backed by energy scavenging and solar cells (Estrin, 2001; Yick, et al., 2008).

• •

  Unreliable devices that may stop functioning either due to physical damage or because they have run out of energy.

• •

  Dense, overlapping data sets and in-network processing of data to reduce network contention and energy consumption.

With a fast moving world of building powerful sensors, the processing cost is becoming cheaper every day. Sensor devices are powerful enough to act like small computers autonomously managing their own resources and collaborating with other wireless nodes to provide detailed observations of the physical world. To rapidly build flexible sensing programs, researchers have proposed various programming models. A programming model is simply a development platform which constitutes a compiler for a domain specific language and a set of services. Services introduced by a programming model are the ones that are commonly required by the applications in which the programming model is intended to be used. As an example services which might be provided by a programming model are execution model (thread-based or event-based), routing algorithm and time synchronization. Research community has proposed a number of programming models for the wireless sensor networks. One of the most well known contributions of the research community to the programming models used for wireless sensor networks is the TinyOS system. Using the right programming model can have enormous impact in the development, deployment and maintainability of sensor networks. As for instance, choosing a programming model that is event-based requires a different set of expertise and imposes different type of restrictions compared to a thread-based programming model.