FISST has been created in part to address the issues in probabilistic inference that the "cookbook Bayesian" viewpoint encourages us to ignore or even be unaware of. These issues include: 1. dealing with poorly characterized sensor likelihoods 2. dealing with "ambiguous" data 3. constructing likelihoods for "ambiguous" data 4. constructing true MT likelihoods and Markov transition densities 5. dealing with the "curse of dimensionality" in MT problems 6. providing a single, fully probabilistic, systematic, and unified foundation for MS-MT detection, tracking, ID, data fusion, sensor management, performance estimation, and threat estimation and prediction, while 7. accomplishing all of this within the framework of a direct, relatively simple generalization of standard statistics and undergraduate calculus During the last two years FISST has emerged from the realm of basic research to a range of practical engineering research applications. The purpose of this lecture has been to summarize the FISST approach and its use in such applications. The main challenges ahead are to increase the calculability of MT filtering, in general, beyond the Gaussian approximation.
展开▼