Robust optimization for uncertain data generation rate in wireless sensor networks

Wireless Sensor Networks (WSNs) stand out with their applicability in many different areas such as military, health, environment, disaster management, habitat observation along with recent technological innovations. On the other hand, Wireless Body Area Networks (WBANs), is a special type of WSNs that has been recently adopted and is widely used in important health applications such as remote monitoring of patients' health conditions. As a large number of sensor nodes are scattered randomly to the area to be observed, replacing or recharging the batteries of the sensors may not be meaningful or possible in practice. That's why it becomes very important to ensure the energy-efficient use of the network because the main expectation from a network is to have it stay functional as long as possible. Another problem encountered in WSN applications is stated as traffic uncertainty in the literature. In some cases, sensor nodes generate data in an event-driven manner. This means that a sensor generates data only if an event occurs in its sensing region. Otherwise, the sensors save energy by going into a sleep state. For this reason, the amount of data to be transmitted is not known precisely in advance, which leads to traffic uncertainty. In the literature, uncetainty in data production rate is generally neglected. However, that deteriorates the performance of the network. Body sensor nodes within WBAN are generally configured to perform event-driven measurements. Neglecting the uncertainty in data traffic at the network design stage can lead to situations that may adversely affect the health of patients. In order to address network lifetime and uncertain data generation rate problems in WSN technology, we first present an energy-efficient WBAN design model where the data generation rate for each sensor is assumed to be known precisely. Then, under the prediction that the amount of data measured by sensors in unit time may differ, we introduce a robust energy-efficient WSN design model, where the set of applicable production rates is expressed with a polyhedral set. Initially, we have a semi-infinite programming model for which a tight formulation is obtained using duality. For the experimental tests, we use random instances generated based on a 1.75 m male body. We analyze the effect of uncertainty in data generation rates on WBAN performance with respect to different metrics . Even though the network lifetimes with the deterministic model are better than the ones with the robust model, it has been observed that the network designs in the deterministic model get worse than the robust model when faced with a challenging applicable scenario other than the expected one.