An Interlaced Extended Kalman Filter for sensor networks localisation

Abstract

Sensor networks have become a widely used technology for applications ranging from military surveillance to industrial fault detection. So far, the evolution in micro-electronics has made it possible to build networks of inexpensive nodes characterised by modest computation and storage capability as well as limited battery life. In such a context, having an accurate knowledge about nodes position is fundamental to achieve almost any task. Several techniques to deal with the localisation problem have been proposed in literature: most of them rely on a centralised approach, whereas others work in a distributed fashion. However, a number of approaches do require a prior knowledge of particular nodes, i.e. anchors, whereas others can face the problem without relying on this information. In this paper, a new approach based on an Interlaced Extended Kalman Filter (IEKF) is proposed: the algorithm, working in a distributed fashion, provides an accurate estimation of node poses with a reduced computational complexity. Moreover, no prior knowledge for any nodes is required to produce an estimation in a relative coordinate system. Exhaustive experiments, carried on MICAz nodes, are shown to prove the effectiveness of the proposed IEKF.