On the Design and Analysis of Transport Protocols over Wireless Sensor Networks

Sensor networks are typically data driven where the whole network cooperates in communicating data from source sensors to sinks (typical repository/server). One of the main characteristics of a typical sensor node is the limited power supply it has (Kahn et al., 1999). Usually, it is battery operated which might last for some months to a year (depending on the type of application and other application specifications). Sensing nodes typically exhibit limited capabilities in terms of processing, communication, and especially, power (Pottie et al., 2000). Different application would have different constraints and priorities on how their sensor network must behave. Thus, energy conservation is of prime consideration in sensor network protocols in order to maximize the network's operational lifetime. Rather than sending individual data items from sensors to sinks, it is more energy efficient to send aggregated data. The net effect of this aggregation is, by transmitting less data units, considerable energy savings can be achieved which is the main idea behind in-network (Madden et al., 2002) aggregation and further distributed processing of the data. Since enabling communication between sensors and sinks is the major role of sensor networks, many research works [Gopalsamy et al., 2002] have investigated energy-aware data delivery. However, sensor networks experience wireless errors and congestion more severely than other wireless networks because of the low capability to recover from losses and the high node-density. Therefore, robustness is also important to energy conservation since unreliable data delivery, which increases the probability of data retransmission under high loss rates, results in the consumption of a large amount of energy. Although the problem has been addressed by previous works [Heinzelman et al., 1999 & Ye et al., 2003] in the context of wireless ad-hoc networks, such approaches cannot be directly applied to the sensor environment. Because of the distinctive characteristics of multipoint-to-point communication vs. point-to-multipoint communication, the data delivery problem in sensor networks can be seen as consisting of two problems: downstream and upstream data delivery. Therefore, we address these problems as two separate ones. Firstly, a sink-tosensors energy-aware data delivery scheme is proposed to solve the downstream problem while considering robustness simultaneously. Secondly, a sensors-to-sink energy-aware data delivery scheme is proposed to address the upstream problem. 1