多传感器量测融合算法的性能比较

归纳三类多传感器量测融合算法,即扩维滤波法、伪序贯滤波法和复合量测滤波法。采用协方差分析的方法比较各类算法的滤波精度,证明它们均能在各自给定的条件下实现线性最小均方意义上的最优滤波。仿真实例对各类算法的计算量和灵活性等性能进行比较,结果表明扩维型信息滤波器的计算量最小、灵活性最高,扩维型Kalman滤波器、伪序贯滤波器的计算量较大,而两种复合量测滤波器对各传感器的量测矩阵有一定要求,以致灵活性较差。所得结论对量测融合算法的实际应用具有一定的指导意义。

Performance Comparison of Multisensor Measurement Fusion Algorithms

Currently there exist three multisensor measurement fusion methods, namely, augmented method, pseudo-sequential filtering method, and combined measurement filtering method. Accuracy of these algorithms are compared by a covariance analysis method, and a conclusion is drawn that they can all obtain LMMSE (Linear Minimum Mean-Square Error) estimation under some assumptions. Other performance of these algorithms, such as computation cost and flexibility, is compared by Monte-Carlo simulation. Results show that augmented method based on information filter has the lowest computation cost and highest flexibility, augmented method based Kalman filter and pseudo-sequential filter have higher computation cost, and the two combined measurement filters are less flexible because they demand that sensor measurement matrixes satisfy some additive conditions. These conclusions are valuable in practical engineering applications.