A process distribution approach for multi sensor data fusion systems based on geographical dataspace partitioning

In this work we present a new process distribution approach for sensor data fusion (SDF) systems, called FuseDis (fusion distribution using tessellated space). It is based on a hybrid partitioning approach, aimed at managing computational burden and achieving scalability. First, a functional decomposition divides SDF functionality into taskgroups, vectorizing operations. Second, a partitioning of the dataspace, based on geographic attributes of the data, is applied to parallelize the processing. A tessellation applied to the plot-space implicitly defines for each tile the set of candidate-tracks yielding useful correlations with the plots in a tile. Some tracks may occur as correlation-candidates for multiple tiles. Conflicts caused by correlations of such tracks with plots in different tiles, are solved by combining correlations of the involved tracks and plots into independent association problems.