定位噪声统计特性未知的变分贝叶斯协同目标跟踪

Variational Bayesian cooperative target tracking with unknown localization noise statistics

为增强定位噪声统计特性未知情况下协同目标跟踪的可靠性,提出一种联合估计目标和协同车状态以及定位噪声统计参数的贝叶斯模型。为实现递推估计,设计了在线变分贝叶斯推断算法。仿真结果表明,当定位噪声的统计特性未知且随时间动态变化时,该算法可以有效提高目标跟踪的精度,与单车跟踪相比,协同跟踪误差可以降低18.7%~23.6%,与其他协同算法相比,误差可以降低4.8%~9.7%。

In order to enhance the reliability of cooperative target tracking under unknown localization noise statistics,a Bayesian model is proposed for joint estimation of the states of target vehicle and cooperative vehicle and localization noise statistics parameters.An online variational Bayesian inference algorithm is further developed to realize recursive estimation.The simulation results show that the algorithm can effectively improve the accuracy of target tracking when the statistical characteristics of localization noise is uncertain and changes dynamically.Comparing with non-cooperative algorithm,the tracking error can be reduced by 18.7%~23.6%.In comparison with other cooperative algorithms,the tracking error can be reduced by 4.8%~9.7%.