Performance evaluation of a mathematical programming based clustering algorithm for a wireless ad hoc network operating in a threat environment

A distributed sensing network consists of several spatially separated sensors, each with possibly different characteristics and not all of them sensing the same set of entities. The sensors are mobile and change location with time. In this work, we evaluate the operational level performance of a column generation (CG) based clustering algorithm, that is developed for locating a given number of clusterheads in a wireless ad hoc sensor network such that maximum information can be gathered from the sensors under hostile conditions. This methodology is also compared with a representative approach (MOBIC - Mobility Based Metric for Clustering) from the clustering algorithms that have been proposed in the literature. Both small (30 nodes) and medium sized (60 nodes) networks are used for comparison purposes.; As a result of the numerical studies, it is concluded that the CG heuristic is superior to the original MOBIC algorithm in terms of sensor coverage. Under 3 of the 64 scenarios, MOBIC provides slightly better coverage. However, the objective function values corresponding to these scenarios indicate that the CG heuristic outperforms MOBIC by 46.06%. According to the packet level analysis performed using OPNET, a smaller percentage of packets sent by the sensors is lost due to the collision of packets transmitted to the same receiver channel of a clusterhead when the results of the CG heuristic are used. The percentage loss values are found to be 1.46% for CG and 2.22% for MOBIC. Finally, the number of packets received by the clusterheads is always higher for the CG heuristic.