Research On Maneuvering Target TrackingBased On Interacting Multiple Modelsalgorithm

As the development of science and technology and the increasing actual needs, target tracking problem attracted widespread attention. Target tracking technology has been widely used in the military field, such as battlefield surveillance, early warning, attack, fire control, and civilian areas such as air traffic control, vessel arrival and personal navigation positioning system, etc. Maneuvering target tracking problem is closely related to people’s daily lives, and it has significant application value, therefore, the issue has become a focus of the researchers at home and abroad.Due to the scientific and technological progress, the maneuverability performance of target has been greatly improved, showing the characteristics of high-speed and high maneuverable, so single model tracking algorithms based on Kalman filter and its improved filter are not suitable for practical applications. Interacting Multiple Model (IMM) algorithm was proposed, and it’s performance is much higher than single model algorithms, so it is widely used in the actual situationHowever, interacting multiple model algorithm still has shortage, for example, the setting of model set, so, it still can not meet the needs in some high accuracy target tracking case.In this thesis, the main contents is maneuvering target tracking algorithm based on interacting multiple model algorithm; the subject is high precision maneuvering target tracking of linear observation environment; the purpose is to propose improved IMM algorithms to meet the high accuracy case of offline or online; the method is analyzing the problems of traditional IMM algorithm, proposing improved algorithms, verifying the algorithms by simulation.This thesis studies two aspects:First, aiming at the problem of tracking accuracy of IMM, Switched IMM(S-IMM) was proposed. S-IMM sets two paralleled model sets for tracking different case of target motion. According to the characteristics of model probability of the two model sets, target state is determined, and then, the optimal result of two model set’output is output. But, the computational burden is large.Second, aiming at the problem of S-IMM, an improved Switched IMM(IS-IMM) algorithm is proposed. IS-IMM also sets two model sets, but at each moment, only one model set is used. By analyzing the characteristics of the current models’ probabilities, target state is determined. If the state is changed, IS-IMM begins to use another model set automatically.In this thesis, the two proposed algorithms are verified by simulation. Simulation results show that S-IMM’s performance is much higher than traditional IMM, but it’s computational burden is large; IS-IMM’s performance is slightly lower than S-IMM, but it’s computational burden is much smaller than S-IMM; IS-IMM’s performance is better than traditional IMM, too.