Expected-Mode Augmentation Algorithms for Variable-Structure Multiple-Model Estimation

Survey of maneuvering target tracking. part v: multiple-model methods

“…As shown in [207] and [212] theoretically, such an augmentation improves accuracy of MM estimation, which is supported by the demonstrations given in [186], [207], [212], and [297] for a variety of applications. Since the estimated mode is added constantly to the working set M k of models, the optimal fusion rule of two MM estimators of Section VIC is instrumental here.…”

“…Most VSMM algorithms developed so far, including those of [224], [219], [215], [195], [207], [212], and [346], use this optimal fusion rule.…”

“…resulting in the expected-mode augmentation [207,212], 2) the estimatem ML kjk = argmax m f(z k j s k = m) that is the model with the largest likelihood, resulting in the ML model augmentation [297], and 3) MAP estimatem MAP k = argmax m Pfs k = m j z k g, which is the model with the largest posterior probability. A promising alternative is to augment the model set also by the predicted modes, such aŝ…”

“…Since the estimated mode is added constantly to the working set M k of models, the optimal fusion rule of two MM estimators of Section VIC is instrumental here. More generally, the model set M k can be augmented by two or more models using average modes over a (likely) subset of M k [207,212] or models with the largest likelihoods or probabilities. Augmentations based on MMSE, ML, and MAP estimates have distinct characteristics.…”

“…In each of these designs, the probability mass and the location of every model are determined. The simplest possible diamond-set design (with one at the center and six on the first layer) was implemented in [211] and [207] for an example of maneuvering target tracking using MM algorithms.…”