A Inverted-Wishart Distribution Approach for the Noise Covariance Matrix Estimation Applied to Target Tracking Problems

Vol. 1, 2019. - 108102
Oral
Favoritar este trabalho
Como citar esse trabalho?
Resumo

Improving the precision of an instrument is a hard task. Generally, this can be accomplished using stochastic filters, e.g., Kalman Filter (KF). Normally, these filters have to be tuned by an arbitrary definition of its noise covariance matrices. In this context, the objective of this paper is estimate dynamically the measurement noise covariance matrix, by means of the Bayes Theorem (BT), in a Kalman Filter algorithm, using the Inverted-Wishart distribution (IW-R Kalman Filter). The proposed method is compared with three other classic stochastic filters. In this comparison the IW-R Kalman Filter has achieved the best RMSE (Root Mean Square Error), outperforming the other filters. Therefore, one can conclude that dynamic estimation of noise covariances could improve the performance of the traditional stochastic filters.

Instituições
  • 1 Universidade Tecnológica Federal do Paraná
Eixo Temático
  • Identificação de Sistemas
Palavras-chave
Target tracking
Kalman filter
Adaptive Kalman filter
Inverted-Wishart distribution