End-To-End Performance Evaluation of Sensor Fusion and Bias Estimation for Multi-Sensor Hand-Off

A method for evaluation of the end-to-end performance of a hierarchical sensor fusion system with joint track correlation and fusion and bias estimation is presented along with simulation results that illustrate the method. In a track fusion system, the first level of the fusion system performs joint track correlation and fusion with bias estimation for tracks from a subset of the sensors and those fused tracks are handed off to a second-level for joint track correlation and fusion with bias estimation for tracks from the remaining sensors. The simulated scenario includes two sensors at Level 1 and one sensor at Level 2 for the fusion of a total of three sensors. The joint track correlation, fusion, and bias estimation is achieved with Murty's K-best Hypotheses algorithm. Joint track fusion and bias estimation is performed separately for all of the correlation hypotheses, and the hypotheses are scored and re- ranked to find the best. The established metrics of pattern accuracy, pattern consistency, pattern containment, and the probability of correct correlation are used to characterize performance versus track density. A joint track and bias containment metric over a cluster of tracks, pattern complex containment, is developed and proposed for assessing the end-to-end performance after handover to the sensors at Level 2. Monte Carlo simulations were performed to demonstrate and compare the pattern metrics and probability of correct correlation after hand-over for various aspect angles between sensors, various number of targets in the pattern, and various levels of sensor bias. The results show that the pattern complex containment metric matches the chi square threshold probabilities when track-to-track correlation is near perfect, and therefore, it should be a good indicator of successful sensor hand-over performance.