Unified Bayes multitarget fusion of ambiguous data sources

The fact that evidence can take highly disparate forms has been a major stumbling block in multisource-multitarget data fusion. Evidence can have at least three forms: unambiguous data (easily amenable to probabilistic analysis); ambiguously-generated data (difficult to characterize probabilistically); and ambiguous data (difficult to even model mathematically). We summarize a unified, systematic, and fully probabilistic methodology for fusing all three data types with the aim of detecting, tracking, and identifying multiple targets. The basic tool is the generalized likelihood function, which hedges against the inherent uncertainties associated with ambiguous and ambiguously-generated data.