基于IMM-PF的分布式估计融合算法  

Distributed fusion algorithm based on IMM-PF

作　　者：彭志专[1] 冯金富[1] 钟咏兵[1] 伍友利[1] 梁晓龙[1] 

机构地区：[1]空军工程大学工程学院,西安710038

出　　处：《控制与决策》2008年第7期837-840,共4页Control and Decision

基　　金：国家863计划项目(2006AA701307);国家自然科学基金项目(60674031)

摘　　要：针对基于扩展卡尔曼滤波的估计融合算法存在线性化误差,且受高斯噪声假设限制的问题,提出一种基于交互式多模型粒子滤波(IMM-PF)的分布式多传感器估计融合算法.各传感器节点采用IMM-PF算法,以便在非线性、非高斯条件下稳健地跟踪机动目标;融合中心则采用基于粒子滤波(PF)的分布式融合方法进行全局估计融合.该算法适用于非线性、非高斯条件下的多传感器状态估计.仿真结果表明,该算法能够提高多传感器系统状态估计的精度.Distributed fusion algorithm for nonlinear/non-Gaussian situations is addressed. The usual fusion approach is based on extended Kalman filter （EKF） which often leads to poor convergence and erratic filter behavior in highly nonlinear systems. Particle filtering （PF） is quite convenient in nonlinear/non-Gaussian filteIing problems. Based on PF,a distributed fusion algorithm is developed, which can be used in nonlinear/non-Gaussian applications. At the sensor level,observer maintains its own estimate with an interacting multiple model particle filter （IMM-PF）,and at the fusion level,the full state estimate is processed by using nonlinear fusion rule. Simulation results show that the proposed method can significantly improve the state estimation precision of the multisensor system.

关 键 词：分布式融合 粒子滤波 交互式多模型 非线性/非高斯 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]