Improved Track Consistency for Improved Multiple Sensor Tracking (MDA04-121)

The objective of this proposed SBIR Phase II project is to demonstrate (using Monte Carlo simulations) the practicality of new algorithms to substantially improve track consistency in network centric sensor data fusion. Track consistency is the property that the computed error covariance matrix of a track realistically represents the uncertainty of the actual errors of the target track. Tracker computed error covariance matrices are used in data (and track) association processing and in the track filter. Consequently, degraded track consistency causes data misassociations (assignment of the wrong measurement or local track to a network track) that will substantially degrade fusion tracking performance. Performance of subsequent processing functions, such as for discrimination, sensor resource management, and weapons resource management depend not only on the accuracy of each track but may use the covariance matrix to indicate how accurate each track is. Hence, an inconsistent track covariance can mislead the subsequent processing and the warfighter about how good each track is. Our Phase I effort proved the feasibility for new processing methods to correct for the common causes of degraded track consistency in fusion tracks, such as, misassociations (miscorrelations), residual sensor biases, and tracker design model parameter errors.