摘要
针对移动机器人多传感器单个或组合故障的情况,提出一种基于EMA_UKF(expected mode augmentation-unscented Kalman filter)方法,用于解决传统固定结构交互多模型算法(FSMM)因模型数量多而造成实时性较差,以及扩展卡尔曼滤波(EKF)计算复杂且精度不高的问题。EMA_UKF方法将期望模型扩张算法(EMA)与无味卡尔曼滤波方法(UKF)相结合,首先利用模型集合自适应来确定期望模型;然后用期望模型扩张初始模型集,通过UKF滤波得到接近真实模型状态的估计结果,判断传感器故障类型。最后,通过与传统的FSMM方法的实验对比,表明该方法能够有效地判断出移动机器人单个或组合传感器故障类型,并且明显地提高了诊断精度。
For mobile robot multi-sensor single or combination of fault conditions, a kind of method based on EMA_UKF （expected mode augmentation-unscented kalman filter） was put forward. In order to solve the tradition-al fixed structure interacting multiple model algorithm （ FSMM） caused by model number poor real-time perform-ance and extended Kalman filter （EKF） solve the problem of the nonlinear system accuracy is not high, EMA_UKF combined the expected-mode augmented （EMA） method with unscented Kalman filter （UKF） method, Firstly, u-sing adaptive model set to determine the expectation model. Then, using expectation model expand initial model set, the real model state estimation is obtained by UKF filter, judging the sensor fault type. Finally, through exper-iments compared with traditional FSMM methods, the results show that the method can accurately judge the mobile robot single or combined sensor fault type, and the diagnostic accuracy has been obviously improved.
出处
《科学技术与工程》
北大核心
2017年第5期230-236,共7页
Science Technology and Engineering
基金
国家自然科学基金(61165005)
江西省教育厅科技项目(GJJ50542)资助
