基于PDA思想的分布式多传感器融合方法

Distributed fusion of multisensor data based on probabilistic data fusion

在传感器观测噪声不一致或有异常数据存在的条件下,分布式数据融合因没有剔除严重偏离真实值的传感器估计值,从而影响下一步的融合估计.对此,利用概率数据互联的思想,设计以融合中心预测值为中心、传感器节点估计值为观测值的预测域,并引入定向概率数据互联,对进入预测域的传感器估计值分配权重.仿真结果表明,利用概率数据互联思想的多传感器有效地实现了数据融合,其融合精度较传统分布式融合有所提高;在异常数据明显的情况下,算法的效果更加显著.

When the abnormal datas of sensor exist, the traditional method can not eliminate abnormal datas. The distributed fusion of multisensor data based on probabilistic data fusion is proposed to overcome this problem. The area of state estimation to sensor is assumed to be the validation region, and the prediction of sensor fusion center is designed to be that of the validition region. Simulation results show that the algorithm based on PDA is superior to the traditional distributed fusion. The approach provides stability of the association event probability and increases the performance of target tracking.