CAJ | 학술논문

为解决粒子滤波应用到IMM算法时计算量过大的问题,融合交互式多模型和粒子滤波,提出了一种采用多速率方法的交互式多模型粒子滤波(multirate interacting multiple model particle filter,MRIMMPF)算法.该算法采用多模型结构来跟踪任意机动的目标;使用一种3模型、one-third速率/全速率跟踪算法,一个one-third速率模型处理非机动或微弱机动,2个全速率模型用于机动模式,以处理非线性、非高斯问题.仿真结果表明,MRIMMPF算法在性能上并不低于交互多模型粒子滤波(IMMPF)算法,但是计算量明显减小.
위해결입자려파응용도IMM산법시계산량과대적문제,융합교호식다모형화입자려파,제출료일충채용다속솔방법적교호식다모형입자려파(multirate interacting multiple model particle filter,MRIMMPF)산법.해산법채용다모형결구래근종임의궤동적목표;사용일충3모형、one-third속솔/전속솔근종산법,일개one-third속솔모형처리비궤동혹미약궤동,2개전속솔모형용우궤동모식,이처리비선성、비고사문제.방진결과표명,MRIMMPF산법재성능상병불저우교호다모형입자려파(IMMPF)산법,단시계산량명현감소.
By combining the interacting multiple model with the particle filter, a new interacting multiple model particle filter (multi-rate interacting multiple model particle filter, i.e. MRIMMPF) algorithm is proposed, in which a multi-rate technique is adopted. In the algorithm the multiple models are used for tracking arbitrary maneuvering target. To deal with the nonlinear and non-Gaussian problems, a particle filter is adopted in each model. But when the particle filter is applied in the IMM algorithm, an expensive computation appears in the whole process. In order to solve this kind of disadvantage, the multi-rate technique is adopted, by doing so, the computation is obviously reduced. The simulation results show that the MRIMMPF algorithm works as well as the IMMPF algorithm in performance.