A Feasibility Check for Geographical Cluster Based Routing under Inaccurate Node Localization in Wireless Sensor Networks

Localized geographic single path routing along a wireless network graph requires exact location information about the network nodes to assure message delivery guarantees. Node localization in practice however is not exact. Errors ranging from several centimeters up to several meters are usual. How to perform localized routing in practice when such errors are prevalent? In this work we look at a promising routing variant which does not completely overcome this practical problem but which mitigates it. The concept does away with trying to find node positions as precise as possible but allows inaccuracies from the very beginning. It partitions the plane by a regular mesh of hexagons. The only information which is of interest is in which cell of that partitioning a node is located in. Using this node embedding, a virtual geographic overlay graph can then be constructed. To find the node positions we apply three variants of multidimensional scaling, two of them being a node localization approach which has been well studied in the context of sensor networks and one which we apply here for the first time in that context. Using the location information we get from these localization approaches we embed the nodes into the clusters their location falls into. We define two graph metrics to assess the quality of the overlay graph obtained by the embedding. Applying these two metrics in a simulation study, we show that cluster based routing is an eligible approach to support localized geographic routing when location errors are prevalent.