An Evaluation of Several Fusion Algorithms for Multi-sensor Tracking System

摘要

In multi-sensor environment, two kinds of methodologies have been used for fusion, which are measurement-level fusion and track-level fusion. Generally, the measurement-level fusion is optimal but computationally inefficient, and the track-level method is more efficient but suboptimal. In most cases, the later seems more attractive. However, in track-level fusion algorithms, estimated state vector from each sensor is not independent because of the common process noises. Based on the facts, many fusion algorithms, such as weighted covariance, information matrix and covariance intersection, were derived to combine the local estimates. In this paper, a performance evaluation is executed to study the performance of various track-to-track fusion algorithms from aspects of fusion accuracy, feedback and process noises. Tacking results of three fusion algorithms are compared with each other using Monte Carlo simulation.