Probability Hypothesis Density Filter Based Design Concept: A Survey for Space Traffic Modeling and Control

The Probability Hypothesis Density (PHD) filter has been recently received a lot of attention by the estimation and data fusion community for its ability to provide a useful solution to the Bayesian filter problem (i.e., implementation issue). Its core foundation to other parallel directions, such as the Sequential Monte Carlo PHD, the Gaussian Mixture PHD and others, offers a viable path to practically implement and realize this promising technology. Potential key solutions offered by a PHD based design paradigm include: (1) non-Gaussian noise and correlated noise process mitigation; (2) replacement of the Extended Kalman Filter (EKF) linearization process to improve numerical instability; and (3) substitution of current convoluted multiple-target multiple-sensors mainstream solutions (e.g., bookkeeping of EKFs on the track file side coupled with a large probability/hypothesis combinatorial computation) with a compact and computational efficient solution framework (i.e., only need one PHD filter which processes a random finite set as a meta target estimate using one meta sensor consisting of multiple sensors being stacked in a “measurement set”). This paper provides a survey on mainstream multiple-target multiple-sensor tracking algorithms and uses those for a baseline comparison to evaluate the potential payoff of the emerging PHD design framework. Potential benefits of a PHD based design framework via the single meta target single meta sensor concept and current emerging missions, such as near Earth object tracking, space traffic modeling and control, and space situational awareness, will also be discussed in the context of closely spaced objects and large target population beyond typical terrestrial domain applications.