Some Extensive Researches Of Set-membership Estimation Theory And Its Applications

Although the set-membership estimation theory has a long history and there are plenty of achievements in both the theory and its applications, there is still much room to be improved. Since there is still a lot of work left to be done, this dissertation carried out some extensive researches include analyzing the existed typical set-membership filter,proposing a new set-membership filter and a new set-membership based fault detection and isolation method. Our work could be mainly divided into three parts.In the first part, firstly the stability of traditional Optimal Bounding Ellipsoid algorithm and is discussed and some convergence properties of it is given. Because of the disability of traditional Optimal Bounding Ellipsoid set-membership filter in case of incomplete observations, a new set-membership filter with incomplete observations is proposed to deal with the problem. By using the bounded assumptions, the relationship between bound of estimation error and data dropout rates is obtained. If the data dropout rates is less than a maximal value, then the set-membership filter will be asymptotically stable. The result of simulation proved the effectiveness of our method, if the incomplete observation assumptions are satisfied, the estimation error of our method could be controlled in a reasonable range and the estimated value is stable enough.In the second part, the application example of our new filter is introduced. Firstly the wearable sensor based indoor localization system is described, based on human kinematics, the relative indoor localization subsystem and analyzed the shortcoming of accumulative error in application is modeled. Our set-membership filter is applied to fuse the relative localization data and absolute localization data which is provided by RFID tags. It is found that the problem of accumulative error was perfectly solved, and the accuracy of our result was significantly improved.In the third part, a fault detection and isolation problem in linear switched systems is studied, based on the set-membership identification theory, a new ellipsoid feasible parameter set based method is proposed to deal with this problem. This method is applied into the robotic assembly of electrical connectors, the experiment results show that our method is very suitable for linear switched model. Compared with the traditional residual based method, the proposed method has the advantage of high detection rate, no false detection, fast detection speed and good robustness.