Font Size: a A A

Distributed Fusion Filter With Unknown Input System And Application

Posted on:2009-04-07Degree:MasterType:Thesis
Country:ChinaCandidate:J H BaiFull Text:PDF
GTID:2208360245960124Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
The state estimation problems for the systems with unknown inputs, disturbances and biases wide exist in control, communications, signal processing, and fault diagnosis. In multi-sensor environment, different sensors may be affected by different disturbances. The study of state and input estimation problems for the systems with unkown inputs has the very important significance not only in the theory but also in the engineering practice. In this paper, we investigate the distributed information fusion state estimation algorithms for the systems with unknown inputs, including the design of information fusion estimators for systems with stochastic biases, with unknown inputs, and the self-tuning information fusion filter for systems with unknown stochastic bias and unknown noise statistical information.For discrete-time linear systems with stochastic biases, the system is transferred to a special case of multi-model and multi-sensor system by the augmented approach of dynamic systems. Based on the optimal weighted fusion estimation algorithms in the linear minimum variance sense, the distributed optimal fusion Kalman state filter and system bias filter are obtained, respectively. When the noise statistical information is unknown, a distributed identification algorithm is given by using correlation functions. Further, the distributed self-tuning information fusion Kalman filters with a two-stage fusion structure for system state and bias are presented.A linear unbiased minimum variance state filter that does not depend on unknown inputs is developed for discrete-time stochastic linear systems with unknown inputs, where there is not any prior information for the unknown inputs. When there are multiple sensors, the cross-covariance matrix of filtering errors between any two local estimators is derived. Further, the distributed optimal weighted information fusion state filter is given for discrete-time linear systems with unknown inputs of system or/and sensors based on the optimal weighted fusion algorithm in the linear minimum variance sense.For multi-sensor systems, where some sensors have not biases, some have stochastic biases and others have unknown inputs of no any prior information, the cross-covariance matrix of local estimation errors between any two sensors is derived. Also, a fusion state filter in the linear minimum variance sense is given. Applying fault detection technologies to multi-sensor systems, we present a distributed information fusion estimation algorithm with fault detection and give the corresponding distributed fusion structure.
Keywords/Search Tags:unknown input, stochastic bias, distributed information fusion estimation, self-tuning fusion estimation, cross-covariance matrix
PDF Full Text Request
Related items