Gehlen, JoshuaJoshuaGehlenGovaers, FelixFelixGovaers2025-10-092025-10-092025https://publica.fraunhofer.de/handle/publica/49721010.23919/FUSION65864.2025.111241152-s2.0-105015869916In this paper, a novel tensor operator is introduced that directly solves the prediction step of the Bayes recursion for multi-dimensional discretized probability densities in Canonical Polyadic Decomposition form. The proposed operator combines computational efficiency with the possibility for non-linear and non-Gaussian scenarios. Based on the Continuous White Noise Velocity model, it enables a broad range of application in target tracking.enfalseFokker-Planck equationnon-linear state estimationtarget trackingTensor decompositionconference paper