Extending the Reliability of Wireless Sensor Networks through Informed Periodic Redeployment

Abstract

This paper investigates the reliability of wireless sensor networks, deployed over a square area, in regards to two aspects: network connectivity and node failures. Analyzing the phenomenon known as the border effects on the connectivity of such networks, we derive exact expressions for the expected effective connectivity degree of border nodes. We show that the relative average number of neighbors for nodes in the borders is independent of the node transmission range and of the overall network node density. Assuming a network composed of N uniformly distributed nodes over a square area of side L, our simulation experiments demonstrate that the connectivity of the overall network is dominated by the average node degree in the corner borders of the square network area. Using this result, and considering sensor node failure rates, we derive analytical expressions for the mean time to disconnect (MTTD) and the mean number of sensors remaining (MNSR) upon disconnection for a given network. For precise reliability estimates we also calculate the sensor redeployment period ΔT and the number of sensors per redeployment ΔN, that should be effected in order to keep the network continuously connected with probability higher than 99%. We then run additional simulations for a network subject to sensors failures to obtain experimental MTTD and MNSR values which we found to be very close to the analytically derived ones. These experiments also ratified that periodic sensor redeployments characterized by the pair (ΔN, ΔT), resulting from our analysis, can continuously extend the reliability of wireless sensor networks.