Unscented particle filter with estimation windows in submarine tracking

In order to estimate the state of uncertain models, a robust filter based on risk sensitive estimator is proposed, which could automatically change the state noise covariance according to the magnitude of the risk function. As a result, sample impoverishment could be mitigated. Another contribution of this paper is to take every sensor measurement into account, when large sample sets are needed to represent the system´s uncertainty, thereby avoiding the risk of losing valuable sensor information during the update of the filter. A simulation example of submarine bearing and frequency tracking is presented, the experiment results show that new algorithm performs better than generic particle filter and unscented particle filter.