Integrated detection, estimation, and guidance in pursuit of a maneuvering target

The thesis focuses on efficient solutions of non-cooperative pursuit-evasion games with imperfect information on the sta.te of the system. This problem is important in the context of interception of future maneuverable ballistic missiles. However, the theoretical developments are expected to find application to a broad class of hybrid control and estimation problems in industry. The validity of the results is nevertheless confirmed using a benchmark problem in the area of terminal guidance. A specifie interception scenario between an incoming target with no information and a single interceptor missile with noisy measurements is analyzed in the form of a linear hybrid system subject to additive abrupt changes. The general research is aimed to achieve improved homing accuracy by integrating ideas from detection theory, state estimation theory and guidance. The results achieved can be summarized as follows. (i) Two novel maneuver detectors are developed to diagnoze abrupt changes in a class of hybrid systems (detection and isolation of evasive maneuvers): a new implementation of the GLR detector and the novel adaptive-Jtô GLR detector. (ii) Two novel state estimators for target tracking are derived using the novel maneuver detectors. The state estimators employ parameterized family of functions to described possible evasive maneuvers. (iii) A novel adaptive Bayesian multiple model predictor of the ballistic miss is developed which employs semi-Markov models and ide as from detection theory. (iv) A novel integrated estimation and guidance scheme that significantly improves the homing accuracy is also presented. The integrated scheme employs banks of estimators and guidance laws, a maneuver detector, and an on-line governor; the scheme is adaptive with respect to the uncertainty affecting the probability density function of the filtered state. (v) A novel discretization technique for the family of continuous-time, game theoretic, bang-bang guidance laws is introduced. The performance of the novel algorithms is assessed for the scenario of a pursuit-evasion engagement between a randomly maneuvering ballistic missile and an interceptor. Extensive Monte Carlo simulations are employed to evaluate the main statistical properties of the algorithms. The thesis demonstrates the following. (1) The adaptive-ffo GLR detector delivers a more efficient and reliable diagnosis of evasive maneuvers than the original GLR detector. (2) Modeling of the target behavior by parametric families of functions permits to improve the accuracy of the state estimate. (3) Modeling of the future evasive maneuvers by semi-Markov models and their prediction by a Bayesian multiple model approach improves the homing accuracy of the terminal guidance. (4) The adaptation of the state estimator and the guidance law with respect to the probability density of the filtered state within the integrated scheme provides for further tuning of the terminal guidance scheme. (5) The discretization scheme for the bang-bang guidance laws is important and much simpler in application. Moreover, the homing accuracy achieved by using the discretized law is