Decentralised Data Fusion with Dynamic Topologies - A Graphical Model Approach

Abstract

This paper considers the problem of dynamic communication topologies for decentralised data fusion systems. The task of decentralised data fusion is considered as a specific instance of the general distributed inference probem in which there is a single common state of interest which is (partially) observed by a number of mobile sensor platforms. Physical communication links between mobile platforms form a connectivity graph which changes over the lifetime of the system. The challenge is to maintain a correct estimate of the common state. Our general approach is to formulate the decentralised data fusion problem using graphical model techniques. Inference is performed using the distributed Junction Tree architecture developed by Paskin and Guestrin. To address the problem of dynamic network reconfiguration, the suitability of three well-known algorithms (Hugin, Shafer-Shenoy, Covariance Intersect) is investigated. The algorithms are evaluated using a simulated tracking problem for a static and dynamic target.