中文摘要：

分析了头盔显示器伺服系统动平台参数的不确定性与时变性,推导了连续-离散扩展卡尔曼滤波（CDEKF）与连续-离散平方根无味卡尔曼滤波（CDSR-UKF）的辨识过程,结合头盔显示器伺服系统的动力学模型建立了系统动平台参数的辨识模型,并通过仿真试验对比分析了CDEKF和CDSRUKF的辨识效果。设计了动平台参数的突变试验过程,通过试验对CDSR-UKF的实用性进行了检验。仿真结果表明：CDEKF与CDSR-UKF的标准误差比值范围为1.9~6.3,收敛时间比值范围为1.0~27.7,均方根误差的比值范围为1.4~11.0,后者的计算精度、稳定性和收敛速度均要优于前者,且后者的平均收敛时间约为0.002s,具有较好的在线辨识性能;CDSR-UKF的辨识误差小于10%,对大幅度突变和一般幅度突变参数的辨识收敛时间分别约为0.30s和0.04s,能较好地跟踪参数的变化过程,可满足正常使用情况下的头盔显示器伺服系统动平台参数辨识要求。

英文摘要：

The nondeterminacies and time-varying characteristics of parameters for motion platform of helmet mounted display servo system（HMDSS）were analyzed,the identification processes of continuous-discrete extended Kalman filter（CDEKF） and continuous-discrete square-root unscented Kalman filter（CDSR-UKF）were derived,the parameter identification model of motion platform of HMDSS was presented based on the system dynamics model,and the identification effects of CDEKF and CDSR-UKF were compared by simulation.The mutation experiment of parameters for motion platform was designed and implemented to verify the practicability of CDSR-UKF.Simulation result indicates that the standard error ratios,convergence time ratios and root mean square error ratios of CDEKF to CDSR-UKF are 1.9-6.3,1.0-27.7and 1.4-11.0,which means that CDSR-UKF has higher identification precision,stability and convergence velocity than CDEKF.The average convergence time of CDSR-UKF is about 0.002 s,so CDSRUKF has better capacity of real-time identification.The online estimation error of CDSR-UKF is less than 10%,and the convergence times against large parameter mutation and normal parameter mutation are about 0.30 sand 0.04 srespectively,so CDSR-UKF can well trace changingprocesses of identification parameters and satisfy parameter identification requirements of motion platform of HMDSS in normal usage environment.5tabs,30 figs,26refs.