This is a RF-signal based localization scheme which works even in case of channel error. The core design of this novel approach is to have the entire localization space divided into different regions by constructing the perpendicular bisectors between each pair of reference nodes, the ones whose locations are known. These regions are called vertices, edges and faces. The authors introduce the concept of Location Sequence which is the combination of distance ranks from each reference node to the constructed regions. The length of this Location Sequence is based on the number of reference nodes in the localization space and these sequences are processed based on the statistical metrics: Spearman’s rank order correlation coefficient and Kendall’s Tau. The Kendall’s Tau metric is shown to have less localization error when compared to the other. The location of the unknown node is estimated based on the RSS measurement from the regions and constructing its own Location Sequence. The centroid of the region with the nearest matching Location Sequence of the unknown node is the identified location of the unknown node. The SBL shows improvement in localization error in comparison to the other localization approaches.
Although this protocol seems to work well enough when only 10 nodes are present. I find it hard to believe that this would be able to be scaled to a much larger network because the computations alone would drain much of the energy and make the network useless for collecting data.
ReplyDeleteThis paper based on RF-Signal based localization concept.Entire space divided into regions and perpendicular lines drawn between nodes.To find out the location of unknown node the author introduces the concept of location sequence and location sequence table.But this method works when many of nodes known it's position and when network is not densely populated.For example when 10 nodes are there in the region out of which 9 nodes known there position then this method works.In practical mobile networks has thousands of nodes then calculation,time latency is high and this method is not energy efficient needs lots of memory to store location sequence table.
ReplyDeleteI felt this paper as confusing and complicated. One thing which is not clear is that they have more reference nodes than unknown nodes but in general I think we should have less reference nodes and more unknown nodes to evaluate the protocols performance.They did not mention about how energy efficient their scheme is? In my opinion the proposed model is cost effective and utilizes more memory as there are lot of calculations needed to be done by unknown nodes to estimate their position. Overall this paper is interesting.
ReplyDeleteI concur with what others say about the scalability of the approach. Due to the computationally intensive operations involved in the algorithm, its hard to believe that it can be adapted to scenarios where there are fewer nodes which know their location. Moreover, the paper does not measure the memory footprint of the algorithm, due to the location tables used, which is a very important to do especially in sensor networks where memory availability is constrained
ReplyDelete