Jerk模型用于高速机动飞行器跟踪的性能分析

Analysis of Performance of Jerk Models in Tracking High Speed Maneuvering Vehicles

针对高速机动飞行器的跟踪应用,对SJ(Singer Jerk)、CSJ(当前统计Jerk)、MCSJ(修正当前统计Jerk)和αJ等4种主要Jerk模型进行了综述。比较了它们的建模假设、设计参数以及实际应用时的限制因素。基于UKF(无迹卡尔曼滤波)算法,分别在单台和2台雷达的测量条件下,在3种滤波参数集设置下,利用各模型对近空高速滑翔飞行器进行了跟踪仿真。飞行器真实飞行轨迹根据简化动力学方程通过四阶龙格库塔数值积分方法得到。仿真结果表明:4种模型都具有很好的初值鲁棒性;在目标弱机动段落,4种模型具有较好的参数鲁棒性;但在强机动段落,SJ、CSJ、MCSJ等3种模型的参数鲁棒性很差,滤波误差发生严重突跳,αJ模型的参数鲁棒性最好。针对高速机动飞行器的跟踪应用,αJ模型的综合性能最优,而MCSJ模型最差。

A review is given on four Jerk models for tracking HSMV （High Speed Maneuvering Vehicles）, namely, SJ （Singer Jerk）, CSJ （Current Statistic Jerk）, MCSJ （Modified Current Statistic Jerk） and aJ. Their modeling hy- potheses, designed parameters and constraints in real applications are compared. Based on UKF （Unscented Kalman Filter） algorithm, tracking simulation is performed for near space high speed gliding vehicles using the four models under single-radar and dual-radar measurement conditions and with three filtering parameter sets. The real trajectory is generated by solving simplified dynamics formula using Runge-Kutta method. Simulation results show that all four jerk models have good robustness in filter initial value and good robustness for model parameters in weak ma- neuvering course. In strong maneuvering course, SJ, CSJ and MCSJ models have poor robustness and the estimate errors increase sharply while aJ has good robustness. To track HSMV, aJ has the best comprehensive performance while MCSJ is the worst one.