Assessment of Hierarchical Multi-Sensor Multi-Target Track Fusion in the Presence of Large Sensor Biases

The evaluation and assessment of a hierarchical sensor fusion system employing joint track fusion and bias estimation under the presence of large sensor biases is presented along with simulation results that illustrate the performance. In the proposed track fusion system, the first level of the system performs track to track correlation and joint fusion and bias estimation with tracks from a subset of sensors. The debiased fused tracks are in turn handed off to a second-level for track to track correlation and joint fusion and bias estimation with tracks from the remaining sensors. Track to track correlation is achieved using Murty's K-best Hypotheses algorithm with joint track fusion and bias estimation performed for each correlation hypotheses. Established and well known metrics of correlation correctness, root mean squared error, and normalized estimation error squared are used to characterize estimator and correlator performance versus track density. The hand-over of tracks from level 1 to level 2 fusion (i.e., AB to ABC) is assessed with pattern metrics computed over a cluster of tracks, which include pattern accuracy, pattern consistency, and pattern complex containment. In previously published results, [8], a degradation of system performance occurred with large sensor biases. In this paper, we specifically address the assessment of the performance of joint fusion and bias estimation in the presence of large sensor biases with cross correlations between the estimated tracks and biases from one level to the next. Monte Carlo simulations were performed to demonstrate and compare the tracking metrics, pattern metrics, and correlation correctness after hand-over in the presence of a large sensor bias (i.e., a bias covariance that is as large as the sensor's random error covariance).