Distributed detection for landslide prediction using wireless sensor network

In this paper we propose a wireless sensor network (WSN) architecture for landslide prediction in the rocky mountain regions of the Konkan Railways. The overall scheme has two major components: (i) Wear leveled, fault tolerant energy efficient routing protocol and (ii) distributed decision on the occurrence or non-occurrence of landslide. In this paper our focus is on statistical modeling of the landslide strain data and subsequent distributed decision algorithm. We simulate pressure variation on rock samples in a testbed at IIT-Bombay and measure the corresponding strain. This strain data is modeled using variable mean Gaussian process (VMGP). We examine different distributed decision algorithms and find that the Distributed Scalar Based Detection (DSBD) gives as good results as the centralized detection (CD) scheme with respect to probability of missed detection, probability of false alarm, with lesser energy consumption at nodes. Receiver Operating Characteristic (ROC) curves are presented to compare the relative performance of different schemes. Simulation of CD scheme was also done on a simple mica2 testbed.