A particle swarm optimization based satisfactory bias-allowable filtering

This paper investigates a low-order filter which admits the estimation output has as lower system bias as possible for maneuvering object tracking problem. Firstly, the unknown maneuver of the target is decomposed into non-stochastic and stochastic parts. Secondly, by assuming the intensity and duration of the unknown non-stochastic part are finite, the transfer function from the unknown maneuvering of the target to the estimation system bias is deduced by classical control theory. Then, the relationship between dynamic bias coefficient and filter gain matrix is analyzed. Thirdly, the satisfactory bias-allowable filter is proposed by particle swarm optimization(PSO) method for the given order system, which can guarantee the dynamic bias coefficient as lower as possible under the constraints of regional pole index and error variance upper bound index. Differing from H∞ filter, the proposed filter considers the dynamic index of the estimation error, which makes it possible of using lower system model to track maneuvering object and can let the estimation output satisfy the satisfactory system error and stochastic error index requirements. Finally, the effectiveness of the proposed bias-allowable filter is illustrated by a simulation example.