Islam Hussein, Applied Defense Solutions, Matthew P. Wilkins, Applied Defense Solutions, Christopher W. T. Roscoe, Applied Defense Solutions, Weston Faber, Texas A&M University, Suman Chakravorty, Texas A&M University, Paul. W. Schumacher, AFRL
Keywords: Space Situational Awareness, High Performance Computing Applications for Space Surveillance
Abstract:
Finite Set Statistics (FISST) is a rigorous Bayesian multi-hypothesis management tool for the joint detection, classification and tracking of multi-sensor, multi-object systems. Implicit within the approach are solutions to the data association and target label-tracking problems. The full FISST filtering equations, however, are intractable. While FISST-based methods such as the PHD and CPHD filters are tractable, they require heavy moment approximations to the full FISST equations that result in a significant loss of information contained in the collected data. In this paper, we review Smart Sampling Markov Chain Monte Carlo (SSMCMC) that enables FISST to be tractable while avoiding moment approximations. We study the effect of tuning key SSMCMC parameters on tracking quality and computation time. The study is performed on a representative space object catalog with varying numbers of RSOs. The solution is implemented in the Scala computing language at the Maui High Performance Computing Center (MHPCC) facility.
Date of Conference: September 20-23, 2016
Track: Poster