Research On Application Of H_∞ Filtering Theory In Multi-sensor Information Fusion State Estimation

With the rapid development of modern signal processing technology, computer technology, network communication technology, artificial intelligence technology, parallel computing by software and hardware technology, and the constantly appear of new type sensor, multi-sensor information fusion will become an important technical means in military and civilian high-tech systems. Multi-sensor information fusion has extensive research fields and branches, in which information fusion state estimation is very important one. The problems such as multi-sensor target tracking, multi-sensor signal filtering and deconvolution etc are classified to information fusion state estimation. The existing multi-sensor information fusion state estimation methods are based on classical Kalman filtering, whose algorithm is on basis of H₂ estimation rules. It requires accurate system model and the exact known statistical characteristics for external disturbance signal. Kalman filtering has deficiencies when fusion system model has uncertainty, time-delay or nonlinear condition, etc. While H_∞filtering theory is an important branch of the modern robust control, it is a filtering technology developed for uncertain system model or uncertain external disturbance. Introducing the H_∞filtering ideas into multi-sensor information fusion state estimation has the important theoretical significance and practical application value.This dissertation systematically reviews the development history, research status and relevant classical algorithms in multi-sensor information fusion state estimation technology, deeply analyzes the advantages and disadvantages of the algorithms, and sums up the existing fusion estimation problems, introduces the basic concepts and methods of H_∞filtering, sums up the H_∞filter design methods and ideas for stochastic continuous and discrete linear system, compares the realization ways between Kalman filter and H_∞filter, and discusses the main similarities and differences on the idea of two kinds filters. It mainly studies on how to combine the H_∞filtering theory with multi-sensor information fusion state estimation technology, takes full advantage of H_∞filtering theory and methods to solve the problems of system parameters having uncertainty, fusion system state having time-delay, fusion process having stability degree constraints, and system definition having nonlinear constraints, etc, and tries to build the framework of H_∞filtering theory used in multi-sensor information fusion state estimation.The main results of this dissertation are:1. It proposes design methods of centralized and distributed H_∞fusion filter for one class of stochastic uncertain multi-sensor fusion system. The dissertation introduces arithmetic model for stochastic uncertain multi-sensor fusion system, and obtains an existence theory for fusion filter of this kind of stochastic uncertain multi-sensor fusion system according to discrete time bounded real lemma, schur's complement and linear matrix inequities solving techniques. On basis of this theory, it gives the design methods for centralized and distributes H_∞fusion filter. By an example of uncertain multi-sensor targets tracking fusion problem, the dissertation designs the H_∞fusion filter, compares the fusing performance and results with Kalman fusion filer, and verifies the correctness and effectiveness of the achieved theory and methods.2. It proposes design methods of H_∞fusion filter for time-delay multi-sensor fusion system with state dependent noise to solve the problem of states abstracted from actual system model which exist time-delay or stochastic perturbation that make them difficult to be fused and estimated. The dissertation derives an arithmetic model description of this kind of multi-sensor fusion system, defines the performance index for H_∞fusion filer, obtains a stability theorem of the H_∞filer by Lyapunov function method, on basis of which to satisfy the fusion filtering performance index, makes use of LMI solving technique, and obtains the H_∞fusion filer. In order to verify the results' correctness, the dissertation gives an example of state estimation for time-delay fusion system with state dependent noise, and designs the H_∞fusion filter. The simulation results show good performance of the state estimation.3. It proposes design methods of H_∞fusion filer with stability degree constraint for multi-sensor information fusion system with state dependent noise. For fusion system model whose parameters have stochastic perturbation, besides paying attention on performances of estimation accuracy or constraint index etc, we should also consider the tracking convergence speed under the condition of guaranteeing the designed fusion filter's stable. On basis of the H_∞fusion filer design framework, the dissertation further considers the fusion filer's design problem with stability constraints, proposes the system's mathematical model, makes use of continue time system bounded real lemma and system attenuation speed control lemma, and obtains the existence theorem of the H_∞fusion filer. Finally, we achieve the fusion filter's design method on baisis of the theorem. The simulation example shows that the stability constraint is truly at its job.4. It proposes design method of H_∞fusion filer on basis of linear matrix inequities transforming and solving technique for one class of multi-sensor information fusion system with nonlinear condition constraints. For linear multi-senor fusion system state estimation, the existing theories, methods and research frameworks have been matured. To generalize the methods of linear fusion system state estimation to nonlinear system has important practical significance and theoratical research value. Summarized the problems existed in nonlinear fusion estimation and their ideas to solve them, the dissertation combines the design methods which had been discussed for different kinds of H_∞fusion filer, makes use of nonlinear system stability lemma and linear matrix inequities transforming and solving technique, and obtains the existence theorem of the fusion filter for one class of nonlinear fusion system with state dependent noise. Base on this theorem, the dissertation further achieves the fusion filter's design methods and steps. The simulation results show the effectiveness of the obtained conclusions and methods.In sum, the dissertation considers the problems in multi-sensor information fusion state estimation such as system parameters having uncertainty, fusion system state having time-delay, etc, and obtains some pointed theorems and fusion filter design methods which could be considered as the furthering complement and development for multi-sensor information fusion state estimation. Finally, the problems to be solved related to H_∞filtering theory applied in multi-sensor information fusion and further research topics are summarized, and the prospect of the developing tendency is analyzed as well.