多传感器异步采样信息融合估计

【作者】 肖春玉；

【导师】 孙书利；

【作者基本信息】 黑龙江大学 ， 运筹学与控制论， 2011， 硕士

【摘要】 随着多传感器数据融合技术在军事和民用等领域中的广泛应用，国内外学者对它的研究也在迅速升温然而目前研究较多的是同步数据融合问题，即假设各传感器同步对目标进行测量，并且同步传送数据到融合中心而在实际中经常遇到的却是异步融合问题，如所用的各种传感器具有不同的采样速率，固有延迟和通信延迟，都会产生异步多传感器数据融合问题因此研究多传感器异步采样信息融合技术具有更重要的理论意义和应用价值本文主要对多传感器异步采样信息融合估计的算法进行了研究，包括具有整数倍采样率的多传感器系统的信息融合滤波器的设计，具有有理数倍采样率的多传感器系统的信息融合滤波器的设计，以及具有四种速率的多速率多传感器系统的信息融合滤波器的设计对具有整数倍采样率的线性离散随机系统，基于已知建立在最细尺度上的状态方程，通过对状态方程和观测方程进行分块，将模型转化为建立在各个尺度上的状态空间模型对于新的具有单采样率的多传感器系统，推导出了任意两个传感器子系统之间的状态滤波误差互协方差阵的计算公式最后基于线性最小方差最优加权融合算法给出了状态融合滤波器对具有有理数倍采样率的线性离散随机系统，通过模型变换将原来的异步采样系统转化为了具有单采样率的多传感器系统，从而得到局部Kalman滤波器针对多传感器系统，推导了任意两个传感器子系统之间的估计误差互协方差阵最后基于线性最小方差最优加权融合算法给出了状态融合滤波器对具有多速率的线性离散随机系统，其中涉及到四种速率：状态更新率，观测采样率，估计更新率和估计输出率假定系统模型中的状态更新率与观测采样率各不相同对于单传感器，估计更新率与观测采样率相同，估计输出率为任意的正整数首先通过模型变换将原多速率系统转化为单速率线性离散随机系统，随后给出了局部单传感器的滤波器最后依据分布式加权融合算法，给出了两种加权融合状态滤波器更多还原

【Abstract】 With the multi-sensor data fusion technology widely used in military and civil-ian fields, it has been received intensive attention by international and domesticresearchers. However, the problem of synchronization data fusion has been gainedmore investigation, which requires that all sensors to measure a target synchronouslyand all measurements data are communicated to the fusion center synchronously.But in practice, we often encountered with the problem of the multi-rate asyn-chronous data fusion. For example, the sensors may have di?erent sampling peri-ods, di?erent inherent delays and di?erent communication delays, which will leadto asynchronous data fusion. So it is more important in theoretical meaning andapplying values to study the multi-rate asynchronous information fusion estimationproblems. In this paper, we study the information fusion estimation algorithms forthe multi-sensor asynchronous sampling systems, including the systems with inte-ger times sampling rates, with rational times sampling rates and with four multiplesampling rates.For the linear discrete-time stochastic system with integer times sampling rates,based on known state equation modeled in the finest scale, the original system istransformed into a new multi-sensor system based on every scale by augmentationand partition of the state and measurement equations. For the new stochasticsystems with single integer times sampling rate, the cross-covariance matrices offiltering errors between arbitrary two sensors are derived. At last, the multi-sensoroptimal information fusion state filters are presented based on the optimal fusionestimation algorithm in the linear minimum variance sense.For the linear discrete-time stochastic system with rational times samplingrates, it is formalized into a synchronous sampling system with single samplingrate by the model transformation. Then the local Kalman filter of each sensors isobtained. Meanwhile, the cross-covariance matrices of estimation errors betweenarbitrary sensors are derived. At last, the multi-sensor optimal information fusionstate filters are presented based on the optimal fusion estimation algorithm in thelinear minimum variance sense.For a multi-rate linear discrete stochastic system, there exist four rates: the state updating rate,the measurement sampling rates, the estimate updating ratesand the estimate output rate. Suppose that the state updating rate is di?erent fromthe measurement sampling rates. For the single sensor the estimate updating ratesare equal to the measurement sampling rates, the estimate output rate is an arbi-trary positive integer. First, through model transform, the multi-rate asynchronoussampling system is formalized into a synchronous sampling system. Then the statefilters of local sensors are given. At last, the multi-sensor optimal information fusionstate filters are presented based on the optimal fusion estimation algorithm in thelinear minimum variance sense.更多还原