Acoustic Data Processing Using the Decentralized Square Root Information Filter

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.