This thesis is concerned with the adaptive control theory and its applications to the autonomous flight control system of unmanned aerial vehicle. First we propose a adaptive control system based on multiple module architecture and new learning algorithm based on Lyapunov design methods that is applicable in practical problems are proposed. We examine the performance of the proposed method both in simulations and experiments. It is shown that multiple modules are successfully trained and specialized for different domains in the state space in a cooperative way. Furthermore, the control system which consists of several online modules is applied to the autonomous flight control system of aero-robot, and we evaluate our method by flight experiments. Second, the output regulation problem for linear time-invariant systems with unknown parameters is considered. Based on the Lyapunov stability theory, a stabilizing adaptive controller is derived. It is shown that an adaptive controller can be designed using the solution of the parameter dependent Riccati equation if the derivative of the solution is sufficiently small. Then sufficient conditions for the output regulation problem with full information to be solvable are established. Furthermore, the condition on the solution of the Riccati equation imposed above is relaxed. Finally, adaptive output regulation for nonlinear systems described by multiple linear models with unknown parameters is considered. We design a local stabilizing controller for affine nonlinear system using the solution of the state dependent Riccati equation and local output regulation is established using a state dependent regulator equation. Then locally stabilizing adaptive state-feedback controllers for nonlinear systems described by multiple linear models with unknown parameters are designed based on the Lyapunov stability theory. Local adaptive output regulation is also established using a state dependent regulator equation. We extend our method to output feedback control. The adaptive laws are derived from Lyapunov stability analysis which guarantees that observer error and parameter estimation error are bounded provided that the state and the control are bounded. Simulation results are given to illustrate the theory. / Kyoto University (京都大学) / 0048 / 新制・課程博士 / 博士(工学) / 甲第13819号 / 工博第2923号 / 新制||工||1432(附属図書館) / 26035 / UT51-2008-C735 / 京都大学大学院工学研究科航空宇宙工学専攻 / (主査)教授 市川 朗, 教授 椹木 哲夫, 教授 松久 寛 / 学位規則第4条第1項該当

科学研究費補助金 研究種目:基盤研究(A)(2) 課題番号:08555101 研究代表者:早川 義一 研究期間:1996-1997年度