The physical layer of a WSN is nothing new: radio-frequency transmission at unlicensed bands. LOS is not required. The data-link layer monitors the channels and transmits frames only when the channel is idle. The network layer and transport layer require more discussion. Like ad hoc networks, the routing paths between each two nodes cannot be determined and configured prior to deployment because there is no predefined fixed infrastructure in WSNs. Sensor nodes have to discover multihop routes to relevant nodes themselves. This is often done via routing data dissemination, in which packets that contain the transmitter and the distance to the root are flooded in the network. A sensor node, upon receiving such packets, will be able to find a “ parent ” who is closer to the root; hence, a distribution tree can be generated. Data collection from sensor nodes can be routed back to the root following the distribution tree.
Task or query dissemination throughout a sensor network is data-centric in association with data aggregation, a routing scheme known as directed diffusion. Sensor nodes are not addressed uniformly using numeric identifications; instead, the addressing and naming schemes are correlated with the application. They are identified by “ attribute – value ” pairs in their data. A task in the form of some attribute inquiry is sent out from some nodes in the hope of obtaining relevant data from other nodes, and then all participating nodes form a routing gradient toward the originators. In the case where a WSN is used as a platform of the sensory database, the applications and underlying routing schemes must support declarative queries, thereby making the detail of in-network query processing and optimization transparent to the user. Power consumption is another crucial factor when it comes to in-network aggregation support of query processing. Sophisticated power-aware query processing and packet routing schemes have been devised to reduce the overall power consumption of a WSN.
Sensory data delivery can be performed in several ways. Sensor node can actively report readings periodically to its parent or only report when an event occurs. The delivery procedure can also be initiated by a user issuing a command that is diffused across the network. Depending on the design objectives, a WSN may apply different data delivery models to different sensor nodes. For example, some high-level roots in the distribution tree may employ a request-and-response mechanism for queries, whereas some low-level sensor nodes may simply report data continuously.
Compared with mobile ad hoc network, network communication over WSNs imposes additional constraints other than node mobility and power consumption. Sensors node are more prone to failure, and their computational capability and memory capacity are greatly limited. When designing a protocol stack of a WSN, these constraints have to be taken into account. Specifically, because complete raw data forwarding is not necessary in many circumstances, data aggregation may be conducted at various levels of the distribution tree to reduce the amount of data being transferred upward to the gateway while still providing sufficient information to other nodes. Furthermore, data aggregation can be combined with applications of the WSN to further improve the efficiency of data collection and dissemination schemes. This reflects one of the most important characteristics of WSNs: cross-layer design. The well-known sensor operating system is TinyOS, which is an open-source, event-based embedded operating system developed at the University of California, Berkeley. TinyOS provides a set of components for networking, memory management, and power management, as well as data acquisition and query processing tools. The programming language supported by TinyOS is nesC, a C-like language for embedded network system development.
Source of Information : Elsevier Wireless Networking Complete 2010