量测提升卡尔曼滤波

Kalman Filter Based on Measurement Lifting Strategy

滤波器设计是系统辨识和状态估计的重要基础.卡尔曼滤波通过状态预测和量测更新的实现框架,在最小方差准则下实现对目标状态的最优估计,但在单传感器量测环境中其滤波精度易受量测噪声随机性的影响.本文提出一种基于量测提升策略的卡尔曼滤波算法实现框架,新方法依据当前时刻量测和量测噪声先验统计信息构建虚拟量测,并通过对虚拟量测采样以及融合提升系统量测信息可靠性,进而改善状态估计精度.同时,针对算法在工程应用中实时性、准确性以及鲁棒性等需求,设计了分布式加权融合和集中式一致性融合的两种实现结构.理论分析和仿真实验结果验证了算法的可行性和有效性.

Filter design is the signification foundation for system identification and state estimation. Based on the real- ization construction of state prediction and measurement update, Kalman filter can obtain the optimal estimation of state esti- mated under the linear minimum variance criterion, but the filtering precision is vulnerable to the random characteristics in single sensor condition. A novel realization structure of Kalman filter based on measurement lifting strategy is proposed in the paper. At first, virtual measurement is constructed on the basis of latest measurement and the prior statistical information of measurement noise modeling. Then, virtual measurements are reasonably sampled and fusion to modify the measurement reliability, and the estimation precision is improved. In addition, aiming to the algorithm requirements including real-time, precise and robustness in engineering application, the distributed weight fusion structure and the centralized consistency fu- sion are designed respectively. Finally, the theoretical analysis and experimental results show the feasibility and efficiency of algorithm proposed.