Early Warning System for Landslide Predictions using Wireless Sensor Networks

Every year landslides cause heavy loss of life and property particularly in areas that are prone to heavy monsoon. Hence, there is a major need to develop technology that can provide early warning. In this talk we will describe the architecture and preliminary results of using Wireless Sensor Network (WSN) to predict landslides. Primary advantages of using WSN for landslide prediction are: (i) dense data, (ii) overall increased accuracy and robustness because of the dense deployment of sensors,(iii) real-time monitoring and prediction of events and (iv)inexpensive with respect to ease of deployment. In the rocky western (Konkan) coast of India landslides are mainly caused by the increase in strain due to rain water logging in rocks joints, fissures, cavities, causing rocks to decompose and slide down the slope. Existing solutions are expensive and not that accurate.
These methods involve use of expensive equipment like extensometers, tilt sensors, and displacement sensors. In all these, one needs to drill deep holes (approximately 30 meters) to install these sensors. Cost and deployment difficulty preclude dense deployment. We propose a method wherein each node consists of an /inexpensive strain gauge/, a wireless senor node (mote), a mini data acquisition card, and a small size battery. The strain gauge is mounted on the surface of the rock (using few centimeter indentations in the rock) and measures fine rock movement due to built up pressure. We have developed a new algorithm called /CAMP/ / (Clustering and Multi-hop Protocol)/for distributed clustering and multi-hop routing. We also present different distributed hypothesis testing algorithms for deciding the occurrence or non-occurrence of landslides.

Location: Hughes Aircraft Electrical Engineering Center (EEB) - 248

Audiences: Everyone Is Invited

Contact: Shane Goodoff