Multi adaptive Kalman filtering with particle swarm optimization

The key problem of adaptive navigation is to determine the adaptive factors, in order to control the outlying effects of dynamic model errors. The optimal adaptive factors, however, are difficult to be obtained. On the base of multi adaptive robust Kalman filtering, a new kind of multi adaptive robust filtering, which uses particle swarm optimization to determine the factors, is proposed. The adaptive factors optimized by particle swarm optimization have higher reliability than those from current methods. First, multi adaptive factors are computed according to difference of the predicted state and calculated one; then particle swarm optimization is employed to look for more accurate factors if the reasonable fitting function is chosen. An actual dynamic GPS data set is employed to test the new adaptive filtering procedure. It is shown that multi adaptive robust filtering with particle swarm optimization can control the influence of outliers more efficiently, and improve the accuracy of navigation.