Parallelization of a multiple model multitarget tracking algorithm with superlinear speedups

摘要: The interacting multiple model (IMM) estimator has been shown to be very effective when applied air traffic surveillance problem. However, because of the additional filter modules necessary cover possible target maneuvers, IMM also imposes an increasing computational burden. Hence, in effort design a real-time multitarget tracking algorithm that is independent number used state estimator, we propose "coarse-grained" (dynamic) parallelization superior, terms performance, "fine-grained" (static) while not sacrificing accuracy. In addition having potential realizing superlinear speedups, proposed scales larger multiprocessor system and robust, i.e., it adapts diverse scenarios maintaining same level efficiency given any one numerous factors influencing problem size. We develop demonstrate dynamic on shared-memory MIMD for civilian using measurement database based two FAA control radars.