A Covariance Matching based Adaptive MCKF Algorithm

For the problem that the accuracy of the traditional Kalman filtering algorithm is decreased when the system model contains unknown noise parameters, an adaptive Monte Carlo Kalman filtering algorithm is proposed to estimate the unknown noise parameters online and in real time. The core idea of this algorithm is to approximate the theoretical value with the average of the innovation covariance sequence in time, and then obtain the estimation of unknown noise parameters. In the algorithm, we first estimate the covariance of measurement noise by using covariance matching method, and linearize the measurement function. Then we estimate the process noise covariance by the orthogonal property of the residual sequence and the innovation sequence. At last, the adaptive MCKF method is applied to track a random sinusoidal signal, the result shows that this algorithm with unknown noise parameters in system model, can better estimate the unknown parameters, and the adaptive filtering algorithm has higher accuracy and stronger robustness.