One major cause of rapid power depletion in wireless sensor networks is radio communication. Wireless radio communication consumes far more power than processing. A number of techniques have leveraged on this fact to reduce power consumption of nodes. One such technique is data compression, in which the amount data transferred in reduced as a result of the compression process. Eventhough compression algorithms are processor demanding tasks, the fact that wireless communication uses more power and the data must travel between a number of hops before it reaches its destination makes this approach more feasible. Notably, with increase in network size, data packets will need to be transferred among more nodes which results in exponential power savings when applying data compression.
Another technique used for reduction of wireless communication and hence power consumption is in-network processing and filtering of data. Data aggregation techniques have proved beneficial[9,10], as less data travels through the network meaning less wireless communication and therefore less power consumption. However data aggregation introduces some delay in the network as data is held up at aggregator nodes while it awaits arrival of more data in order to perform the aggregation.
Other methods which tackle energy depletion from a different angle are energy aware routing protocols. The aim of energy aware routing protocols are to maximise the network survivability(i.e. extending the time it takes before the network is partitioned into disconnected subnets). These protocols focuses on the overall network's connectivity up-time rather than individual power levels of nodes. So a packet might take a less optimised path to its destination in order to ensure the energy consumption of the network is more equally spread among all nodes. Eventhough more power is used to deliver that specific packet, the creation of disconnected subnets is delayed.
In spite of the mentioned improvements there are still a number of shortcomings in this field; For example in a wireless sensor network, gathered data (sensed physical phenomena) is moved from each node towards one or more central points known as sinks. A sink node either has virtually unlimited resources itself or is connected to a base station machine such as a desktop PC which virtually has unlimited power source, processing power and storage capacity. As all gathered data in the network travel towards the sink, the nodes closer to the sink have to relay messages far more often than other nodes. Considering the fact that wireless communication is a relatively power demanding task for a sensor node, it becomes obvious that nodes closer the sink will finish their power source before other nodes. As a result, network will be divided into islands of disconnected subnetworks. More importantly the base station gets disconnected from the rest of the network therefore data will not be delivered to the user. To deal with depletion of energy on nodes close to sinks due to high communication traffic, mobile sink nodes (known as data MULEs) can be effective. Mules move through the network and collect gathered data from static nodes in the network. This technique distributes the network's power consumption more equally among the nodes in the network by reducing energy spent on relaying messages, and as a result extends the lifetime of a network. However mobility of Mule nodes, adds its own requirements(e.g. mobility power consumption). Nevertheless depending on the nature of an application this can be a feasible approach.
Approaches like the one mentioned above compensate for various shortcomings of static sensor networks by introducing mobility. Mobility in sensor networks exposes new possibilities and solutions in this field which are yet to mature.
Jenson Taylor 2008-01-25