Improved variational inference for tracking in clutter

We apply the expectation propagation (EP) algorithm to temporally track targets using sensors that produce spurious clutter detections, and may sometimes fail to detect the true target. The variational inference framework underlying EP allows the tracker to be easily adapted to varying measurement models. We develop variants of EP based on single Gaussian and Gaussian mixture approximations of posterior target location distributions, which offer a tradeoff between accuracy and computational complexity. Experiments show improved tracking accuracy and uncertainty estimation relative to widely used baseline tracking algorithms.