Parallel implementation of the multiple hypothesis tracking algorithm

Algorithms of interest to the DOD are being developed which will be used in time-critical applications which support surveillance and tracking detect an unknown number of targets. These multitarget tracking algorithms typically demand large computational resources, since they require the identification and tracking of several hundred targets in real-time. It is expected that parallel implementation of multitarget tracking algorithms will provide the desired real-time tracking capability. The multitarget tracking algorithm of particular interest in this investigation is the Multiple Hypothesis Tracking (MHT) algorithm, the basic theory of which a description is provided in the report. It will be demonstrated that parellel implementation and performance of computer algorithms, to a large extent, are dependent on various aspects of the particular architecture of the parallel machine. In this regard, the architecture of the machine used for the MHT algorithm implementation will be described, as well as the implementation itself. The intent of this report will be to investigate speedup of the multiple hypothesis multitarget tracking algorithm employing the Paragon XP/S supercomputer as its implementation platform. As a result of this investigation, future areas for investigation are suggested.