Search

First, the basic Decentralized Square Root Information Filter (DSRIF) theory was extended in 2 ways. An expression for the likelihood function in terms of DSRIF variables, and a method for distributing the prior and process noise statistics over the set of locally optimal filters were derived. Next, software developed in Phase I research was upgraded to enable multitarget tracking within our distributed filtering environment. Outlier detection/ rejection, track initiation, measurement-to-track and track-to-track association routines were encoded and successfully tested using real MLRS data provided by WSMR. Finally, MTI conducted experimental work wherein 2 video trackers were built and networked with a global processor. Laboratory testing showed that the 2-camera network could track a variety of objects over the entire laboratory space. Moreover, each tracker was able to aid the other in acquiring a common target. Field testing of an individual tracker showed that it could acquire and track objects similar to the SADARM submunition, and against a cloudy background., See also Volume 2, AD-B162 611.

VLSI technology has been developed to the point where special purpose processors may be concatenated to form supercomputers with far greater throughput rates than uniprocessor machines. MTI has developed a parallel form of the conventional Kalman filter that is well suited to being implemented in a multiprocessing environment. Moreover, our Decentralized Square Root Information Filter (DSRIF) has several very unique features which could be incorporated into the design of an integrated test range tracking system with much improved performance over existing methods. Phase I research demonstrated feasibility of the DSRIF as a means for solving the linear least squares estimation problem in decentralized form. Also, an extended form of the DSRIF was derived and successfully used to track real Multiple Rocket Launch System Data provided by the White Sands Missile Range.

Very Large Scale Integration technology has been developed to the point where special purpose processors may be concatenated to form supercomputers with far greater throughput rates than uniprocessor machines. MTI has developed a parallel form of the conventional Kalman filter that is well suited to being implemented in a multiprocessing environment. Moreover, our Decentralized Square Root Information Filter (DSRIF) has several very unique features which could be incorporated into the design of an integrated undersea tracking system with much improved performance over existing methods. Phase I research demonstrated feasibility of the DSRIF as a means for solving the linear least squares estimation problem in decentralized form. Underwater tracking of a high velocity torpedo (undergoing high dynamic maneuvers) was simulated. Also, an extended form of the DSRIF was used to complete the processing of real Multiple Rocket Launch System Data provided by the White Sands Missile Range. In this case, an adaptive form of the extended DSRIF, wherein the process noise levels were made to be a function of the globally optimal estimate error covariance, was successfully used to track the rocket data.