Wang Jie, He Jiacheng, Peng Bei, Wang Gang
School of Mechanical and Electrical Engineering, University of Electronic Science and Technology of China, PR China.
School of Information and Communication Engineering, University of Electronic Science and Technology of China, Chengdu 611731, PR China.
ISA Trans. 2024 Dec;155:148-163. doi: 10.1016/j.isatra.2024.09.015. Epub 2024 Sep 16.
The traditional interacting multiple model Kalman filtering algorithm (IMM-KF) can deal with the maneuvering target problem under Gaussian noise by soft switching among possible motion models. In practice, its performance is likely to degrade when handling non-Gaussian noise. We introduce the Gaussian mixture model (GMM) into the IMM-KF, and the GMM is utilized to model the non-Gaussian measurement noise as a mixture of multiple Gaussian probability densities with a certain probability. Then, a GIMM framework is proposed that enables accurate switching and fusion among multiple possible motion and noise models. And combined with Kalman filtering (KF), a GIMM-KF algorithm is proposed that enables accurate state estimation of maneuvering targets under non-Gaussian noise conditions. Subsequently, the provided simulations and experiments validate that the GIMM-KF algorithm outperforms existing methods in terms of accuracy, stability and robustness.
传统的交互式多模型卡尔曼滤波算法（IMM-KF）可以通过在可能的运动模型之间进行软切换来处理高斯噪声下的机动目标问题。在实际应用中，当处理非高斯噪声时，其性能可能会下降。我们将高斯混合模型（GMM）引入到IMM-KF中，利用GMM将非高斯测量噪声建模为多个具有一定概率的高斯概率密度的混合。然后，提出了一种GIMM框架，该框架能够在多个可能的运动和噪声模型之间进行精确的切换和融合。并且结合卡尔曼滤波（KF），提出了一种GIMM-KF算法，该算法能够在非高斯噪声条件下对机动目标进行精确的状态估计。随后，所提供的仿真和实验验证了GIMM-KF算法在准确性、稳定性和鲁棒性方面优于现有方法。
