Filtering Technique With Applications To Terminal Guidance Of Hypersonic Flight Vehicles With Large Maneuvering Ability

The developing trend of the world for the past several decades can generally bedescribed as “overall peace marred by local wars”. In light of this very importantnotation, missile weapons have increasingly become the core focus as a new mean ofattack in the battlefield. Fast attack missile of the next generation such as the TBM(Theatre Ballistic Missile) can execute large maneuvering at super high speed,posing great challenges to the existing interception guidance systems; thus, anin-depth study of some key technique of terminal guidance laws for interception ofmissile weapons will be very helpful for the People’s Liberation Army to takecontrol of the strategic high ground in future wars, and of great significance for thedefense of the territorial integrity and the security of people’s lives and property.Almost all advanced guidance laws used for interception require target motioninformation. While in many cases this information cannot be measured directly, itmust be obtained through online estimation from noise polluted measurements,which requires the support of filtering technique. Filtering technique has always beena hot issue in the domain of signal processing. For the general non-linear andnon-Gaussian systems, there are no recursive explicit solutions for optimal filtering.However, under many circumstances, many kinetic models of missile-targetengagement in terminal guidance problems of hypersonic aircraft are characterizedby non-linear and non-Gaussian in nature; for such problems, it is a more meaningfulpractice to develop suboptimal filtering methods.Based on the background of terminal guidance of aircraft like TBM of the nextgeneration, the modeling and filtering problems of discrete systems in the presenceof unknown information are studied，and for the ineffectiveness of certaintyequivalence theorem in the interception of hypersonic speed target with largemaneuvering ability, a possible solution is given and multiple simulations are carriedout on improved software toolkit based on Matlab. Specifically, the followingdomains are included in the research:First, the Gaussian approximation filtering technique for general non-linear andnon-Gaussian stochastic discrete system is studied based on optimal Bayesianestimation theory. After the derivation of the optimal filter for linear discrete Gaussian system–the KF (Kalman Filter), suboptimal Gaussian approximations ofnon-linear stochastic systems are derived based on the analogy of KF. According todifferent types of system noise, suboptimal non-linear filters with additive andhidden noise are derived respectively, namely the EKF (Extended Kalman Filter) andUKF (Unscented Kalman Filter) algorithms; To improve the numerical robustness ofUKF for systems with high dimension, an auxiliary function method is introducedwhich recalculated the sample points of UKF; A class of hybrid system with manyworking modes is investigated, and the principle and realization method of twoclasses of multiple model algorithms are studied based on Bayesian estimationtheories for hybrid systems. The algorithms provide a fine solution for stateestimation problems for the hybrid systems.Then, the tracking problems for hypersonic aircraft with large maneuvering abilityare discussed. Shaping filters are introduced to imitate the unknown non-Gaussianvariables in systems characterized by non-linearity in nature. Central differencetheory is used to approximate non-linear functions, which lead to the derivation ofcentral difference Kalman Filter (CDKF for short) for non-linear stochastic systemswith additive and hidden noise, respectively. An iterative update step is introduced inthe second step of CDKF, giving an optimal filter named ICDKF under MAP(maximum a posteriori) sense. Two typical non-linear estimation problems from thedomain of terminal guidance in the interception of hypersonic speed demonstrate thatthe proposed ICDKF algorithm effectively improves the estimation accuracy,compared to the mainstream filter of EKF and UKF, thus demonstrate its superiority.And then, the necessity of designing non-linear filters in accordance with thedifferent type of noise entering the system is confirmed through a simulationexample, and then a novel structure of the design of non-linear filters is proposed. Inorder to meet the requirements of the project, a improved filtering and guidancetoolbox using such design method is developed which provides a way to performconsistency checking for some non-linear filters based on suboptimal Gaussianapproximations, and makes it easy to modify the simulation parameters and replacefiltering and guidance algorithms at the beginning of the simulation, and processlarge amounts of statistical data after the end of the simulation. In order to givescientific compare between two non-linear Gaussian filters, according to the specialstructure of the Gaussian filter, an improved initialization method for general non-linear system and a hypothesis testing comparative method based on hypothesistesting theory are proposed, which guarantee the scientific performance comparisonof two non-linear filters.The effectiveness of the proposed methods are validatedthrough the novel filtering and guidance toolbox metioned above.Finally, for the target to be intercepted as TBM of the next generation which hassupersonic speed and can execute large maneuvering how to effectively interceptsuch target is studied and an integration design method of filtering/guidance is given.Integrating CDKF into multiple model algorithms (Multiple Model AdaptiveEstimation, MMAE), a multiple model central difference algorithm (MMAE/CDKF)is used to deal with filtering task in the presence of unknown signal and gives targetacceleration estimation. Considering that the large computation burden of MMAE/CDKF can make it difficult for the implementation of the algorithm on anengineering computer, for target acceleration with the structure type bang-bang, thetechnique of aggregation and pruning are introduced to make simple the MMAE/CDKF algorithm. A fast multiple model central differential algorithm (fast MMAE/CDKF) is developed which retained much of the accuracy as MMAE/CDKF, whilegreatly reduced the amount of computation. Extensive simulation experimentsconfirm the effectiveness of the proposed two algorithms. In order to make better useof the improvement of estimation of acceleration and line-of-sight angel due to theemployment the two proposed algorithms, and solve better the problem of failure ofcertainty equivalence theorem in the interception of hypersonic target with largemaneuvering ability, a switching differential game guidance law named DGL/N isproposed based on basic logic reasoning. Such method takes fully into account theinfluence of the estimation accuracy of the guidance filter and the limited mobility ofthe interceptor, thus gives better homing performance compared to the conventionalguidance law for terminal interception.In summary, several novel design and evaluation methods of terminal guidancelaws and guidance filters for the terminal guidance problems in the interception ofhypersonic speed target with large maneuvering ability are studied in this dissertation,thus lay some theoretical foundation and give some possible methods for thedevelopment of the next-generation missile interception system.