Multiple Hypothesis Correlation in Track-to-Track Fusion Management

Abstract

Track to track fusion systems require a capability to perform track matching across the reporting sensors. In conditions where significant ambiguity exists, for example due to closely spaced objects, a simple single frame assignment algorithm can produce poor results. For measurement-to-track fusion this has long been recognized and sophisticated multiple hypothesis, multiple frame, data association methods considerably improve tracking performance in these challenging scenarios. The most successful of the multiple frame methods are multiple hypothesis tracking (MHT) and multiple frame assignments (MFA), which is formulated as a multidimensional assignment problem. The performance advantage of the multiple frame methods over the single frame methods follows from the ability to hold difficult decisions in abeyance until more information is available and the opportunity to change past decisions to improve current decisions. In this chapter, the multiple source track correlation and fusion problem is formulated as a multidimensional assignment problem. The computation of cost coefficients for the multiple frame correlation assignments is based on a novel batch MAP estimation approach. Based on the multidimensional assignments we introduce a novel multiple hypothesis track correlation approach that allows one to make robust track management decisions over multiple frames of data. The use of the proposed multiple hypothesis, multiple frame correlation system, is expected to improve the fusion system performance in scenarios where significant track assignment ambiguity exists. In the same way that multiple frame processing has shown improvements in the tracking performance in measurement-to-track fusion applications, we expect to achieve improvements in the track-to-track fusion problem.