Research And Implementation Of Resource Management On WSN Nodes

Wireless Sensor Network is comprised of a large number of sensor nodes and the wireless network is organized by wireless communication and self-organization. It combines the computer technology, wireless communication technology and sensor technology together and become very popular in recent studies. Now WSN is wildly used in many areas, including military, environmental monitoring, forest fires prevention and so on.Sensor node is the basic unit of WSN, and it is very cheap and simple compared to other embedded system, so it can be deployed in various environment. But the simplicity of sensor nodes also means the hardware resource is very limited in sensor node. The capability of calculation of the processor is limited and also the memory and the power. To effectively use the limited resource in sensor nodes has become a challenging problem in the study of sensor OS.The applications running on a sensor node are complex and they have different requirements on using the hardware resource of a sensor node, so sensor OS need provide some ways of abstraction of hardware to ease the use of the hardware resource, and sensor OS should also keep flexibility to a certain degree to cater for the different requirements of the applications. Moreover, the memory in a sensor node is very limited so it needs effective management with small memory overhead. The last consideration is about the energy of a sensor node. A sensor node is always powered by a battery, so OS need to effectively use the power to lengthen the work duration of a sensor node.The dissertation provides methods to manage the resource of sensor nodes. The solution proposed by this dissertation is based on event-driven programming mode, which is extensively used in current sensor OS. We provide a unified abstraction model for hardware with unified interface, and also maintain the service queue for the access to the specified peripheral device. This dissertation provides a way to combine user-defined memory pool and the system memory pool together to manage the memory and shows good results. At last this dissertation provides a way to manage the power state of the peripheral device and the processor, which can reduce the overhead of deciding the correct power state for specified hardware.