A look at the PMHT

We combine concepts from numerous papers to provide a derivation and description of a generalized Probabilistic Multi-Hypothesis Tracker (PMHT) that can track multiple targets in a cluttered environment, utilizing multiple sensors and feature measurements, if available. We also provide a simplified method of performing the maximization step of the algorithm when multiple sensors are used, a consistent covariance approximation of the algorithm when using multiple sensors, and we explore the use of deterministic annealing to improve performance and discuss implementation difficulties. Additionally, we derive the complexity of the PMHT and the JPDAF to better understand the advantages of each algorithm.