基于交互式多模型算法跟踪临近空间目标

由于目前机动目标模型越来越向模块化、并行计算的方向发展，对目前算法计算效率提出了更高的要求。对于临近空间超声速机动目标一般采用多种机动模型跟踪，单一模型已经很难满足高精度跟踪的需要。因此需要使用基于多种模型进行交叉耦合的交互式多模型（interacting multiple model, IMM）算法，这种算法特点与临近空间目标高速、高机动特性相适应。同时考虑到扩展卡尔曼滤波 (extended Kalman filter, EKF) 算法对强非线性对象滤波效果不好, 无迹卡尔曼滤波(unscented Kalman filter, UKF)算法对于此类问题，可以很好地加以解决。仿真对比试验表明,交互式多模型〖CD*2〗无迹卡尔曼滤波(interacting multiple model unscented Kalman filter， IMM UKF)算法优于单一模型EKF算法。

Tracking for near space target based on IMM algorithm

Due to the current maneuvering target model is developing towards more and more modular and parallel computing, which puts forward higher requirements for calculation efficiency of algorithm. For high supersonic maneuvering targets in near space, it is often tracked with multiform maneuvering target models, since precision of single target maneuvering model cannot satisfy the requirement of tracking. So it is necessary to use the interactive multiple model algorithm for cross coupling based on a variety of models. The characteristic of this algorithm is adapted to that of high speed and high maneuver for near space target. At the same time, considering the filtering result of extended Kalman filter (EKF) algorithm for strong nonlinear targets is bad, unscented Kalman filter (UKF) algorithm can be solved very well for this problem. So through simulation contrast experiment of two kinds of algorithms, it proves that interacting multiple model unscented Kalman filter (IMM UKF) algorithm guarantees the target tracking accuracy is within the allowable range, validity of the algorithm is verified with Matlab simulation results.