Supporting lightweight adaptations in context-aware wireless sensor networks

Context-aware environments are being populated with Wireless Sensor Networks (WSNs), observing sensory context elements, and adapting their behavior accordingly. Although adaptation has been known as a common approach for addressing context-awareness, the resource-scarceness of WSNs raises the requirements for lightweight adaptations. The related work in the field of updating WSN applications mostly focuses on i) developing techniques to distribute a monolithic program to a set of nodes or ii) reprogramming the whole sensor nodes, which have been seen as impractical and inefficient solutions for a large number of sensors deployed in inaccessible regions. In this paper, we propose a new software development paradigm, which revisits the way WSN applications are designed in order to optimize the adaptation process. Our approach promotes lightweight adaptation by proposing a component model reconfiguring modules at the behavior-level instead of component-level. We evaluate this model by analyzing a sample reconfigurable application atop Contiki---a popular operating system for sensor nodes. The preliminary analysis shows that our adaptation approach is efficient in terms of energy consumption, memory usage, and reconfiguration complexity.