THE MEDIATION DEVICE (MD)

The time-average power consumption of a network node can be written as

where

• P = time-average power consumed, W

• α = duty cycle of operation

• P_o = power consumed while in active operation, W

• P_s = power consumed while in standby mode, W

To reduce the time-average power consumption, both the operation power and the standby power should be reduced. For a given technology and set of applications supported by the network, however, a practical limitation exists for both the operation power and the standby power. For most applications, the operation power is much larger than the standby power:

Under this assumption, we can see that by reducing the duty cycle, low power consumption levels and associated long battery lifetimes can be reached. For example, for a node with 10-mW operation power and 10-μW standby power, if the duty cycle is 0.1 percent, then the time-average power drain is about 19.99 μW. If the node is supplied by a 750 mAh AAA battery, linearly regulated to 1 V, it will have a battery life of more than 37,000 hours, or more than 4 years.

In the simplest case, and for pedagogical purposes, the MD may be assumed to be mains powered, with its receiver continuously active. It records (or, equivalently, calculates from an internal timebase) the relative time differences (offsets) between the received beacons of each node in range. Because the data throughput of wireless sensor networks is expected to be low, the majority of the beacons the MD receives will be query beacons.

Assuming node A needs to send a message to node B, a message transfer would proceed as follows:

The MD can be a dedicated device, as just described; however, ensuring the placement of an MD within range of all network nodes only implies the type of system administration and planning that self-organizing wireless sensor networks attempt to avoid. Further, requiring the MD to receive constantly is not compatible with the concept of a low-power network. An improvement on this situation can be realized if the MD function is distributed throughout all nodes in the network. In this scheme, each node in the network temporarily operates as an MD at a time chosen by an independent random process (i.e., a node begins operation as an MD at a random time, independent of other nodes). To ensure all beacons within range are received, each node receives for a period of time equal to one beacon period, plus the length of a beacon transmission, and then makes transmissions required of an MD to other network nodes when the nodes are receiving. After this procedure, the node returns to normal (i.e., non-MD) operation. Because each device functions as an MD only rarely, the average communication duty cycle of each device can remain very low, and the overall network to remain a low-power, low-cost asynchronous network.

Assuming node A needs to send a message to node B, a message transfer under the distributed MD protocol would proceed as follows:

The MD protocols described so far do not consider channel access per se; they rely on the ALOHA strategy (due to the low data throughput/channel capacity ratio) to minimize packet collisions within the network. This would be quite suitable if the network was a licensed service and no other services were present in the channel.

Implementation of wireless sensor networks is expected to be mainly (if not exclusively) in unlicensed radio bands, however. The selection of a suitable channel access technique for a wireless device operating in an unlicensed radio band, such as an Industrial, Scientific, and Medical (ISM) band, is a complicated one, especially if coexistence with other services, such as WLANs, is required. In an unlicensed band, the usual assumption made in licensed bands that any signal present must be from the same network, or at most a different network in the same service, does not apply; any signals, from sources as varied as microwave ovens and WLANs, may be present. In such interference-limited cases, if a Carrier Sense Multiple Access with Collision Avoidance (CSMA-CA) strategy is employed, based solely on detected energy in the channel, the "fairness doctrine" for channel access often fails. Data throughput may suffer as the device repeatedly backs off in the presence of interference from other services. Some of these services will not employ CSMA-CA in return to ensure channel access at a later time (e.g., microwave ovens). In the MD protocols, channel status assessments are not made prior to beacon transmissions because beacon timing is of critical importance to network operation. Because the duty cycle is so low, this causes little interference to other coexisting services. One alternative to this approach would be to employ a simple channel energy detection CSMA-CA algorithm with the backoff quanta set to be a beacon period (i.e., devices back off only in multiples of the beacon period).

For channel status assessment in unlicensed bands, it is generally necessary not just to detect energy in the channel prior to transmission, but also to identify (as far as possible) the signal present. This must be done quickly, for the channel status (i.e., presence of interference) may change rapidly; in addition, channel status assessment costs precious battery life. Algorithms meeting these requirements are an area for future research.