Memory-Attenuated Least Square Filtering and Its Application

In order to deal with the problem in which the conventional Kalman filtering may be instable or divergent when noise statistics is unknown, a new adaptive filtering is presented, which is defined as memory-attenuated least square filtering (MALSF). The error covariance is multiplied by a decay factor to avoid the divergence and an adaptive estimation for decay factor is developed, and a recursive algorithm based on least square filtering is presented. The descriptions of the noise statistics are not required. This algorithm is simple and has the adaptability. MALSF is applied to INS/DS integrated navigation system. Simulation results show that the proposed algorithm has adaptability and has better estimation accuracy than the conventional Kalman filtering and the least square filtering when noise statistics information is unknown