Joint Estimation of State and Parameter for Tracking System using EM-UKF

Qiang Zhu and Jianxun Li


Unknown parameters, nonlinearity, likelihood function, unscented Kalman filter, expectation-maximization algorithm


Unknown parameters and nonlinearity are two vital issues affecting the performance of tracking system. However, parameters, e.g., system and sensor biases, initial state and characteristic of noises, are not taken into account together in the literature. To solve this, an online state and parameter estimation algorithm is proposed to estimate the system parameters and target state simultaneously in this paper. First we develop the system model and then make use of the expectation-maximization (EM) approach which deals with maximizing the likelihood function of the complete data incorporated with the unscented Kalman filter (UKF). UKF is chosen to alleviate instability caused by nonlinearity generated in the measurement function. Computer simulation shows that the proposed method is effective and reliable, and outperforms the conventional algorithms.

Important Links:

Go Back