Improved unscented kalman filter for bounded state estimation

The unscented kalman filter is widely used in many application fields. In some occasions, the state variable is bounded to some feasible region and this makes the standard unscented kalman filter impossible to be utilized directly. Seldom researches have been made to solve this problem before. Two methods, the projection based method and the sigma points boundary shrinkage based method, are proposed to improve the standard unscented kalman filter to make it suitable for the bounded state estimation question. Experimental results show that the proposed methods are effective.