鲁棒高斯和集合卡尔曼滤波及其在纯角度跟踪中的应用

Robust Gaussian-sum ensemble Kalman filter and its application in bearings-only tracking

中文关键词  纯角度跟踪  异常值  非高斯噪声  集合卡尔曼滤波  高斯和

英文关键词  bearings-only tracking  outliers  non-Gaussian noise  ensemble Kalman filter  Gaussian-sum

中文摘要

      针对纯角度目标跟踪中量测信息易受异常值和非高斯噪声干扰的问题, 提出了一种新的非线性滤波算法– 鲁棒高斯和集合卡尔曼滤波(robust Gaussian-sum ensemble Kalman filter, RGSEnKF)算法. 首先, 采用Huber技术重塑 集合卡尔曼滤波的量测更新过程, 能够有效地处理量测中的异常值. 随后, 将改进的集合卡尔曼滤波在高斯和框架 下进行扩展, 得到RGSEnKF算法, 可以进一步解决受非高斯噪声干扰的非线性系统的状态估计问题. 此外, 新算法 中包含距离参数化初始化策略和高斯分量融合策略. 前者是为了减小纯角度跟踪中距离信息不可观测的影响, 而后 者可以避免高斯分量数目随时间不断增长. 大量仿真结果验证了新算法的有效性和鲁棒性.

英文摘要

      In order to deal with the situation that measurements are easily contaminated by outliers and non-Gaussian noise, a new nonlinear filtering algorithm called the robust Gaussian-sum ensemble Kalman filter (RGSEnKF) is proposed for the bearings-only tracking problem. Firstly, the measurement update process of the ensemble Kalman filter is reformulated by using Huber technique so that outliers can be dealt with efficiently. Further, the improved ensemble Kalman filter is extended within a Gaussian-sum framework, the result is RGSEnKF algorithm which can handle the state estimation problem of nonlinear system corrupted by non-Gaussian noise. Moreover, the new algorithm includes a range-parameterized initialization strategy and a Gaussian merging strategy. The former strategy can reduce the effect of unobservability of range in bearings-only tracking and the latter can prevent the number of Gaussian components from increasing over time. Lots of simulation results validate the effectiveness and robustness of the new algorithm.