並列タイトル等サイバーフィジカルシステムのための状態推定理論に関する研究
一般注記This thesis discusses estimation theories for cyber physical systems. Some social concepts for the next generation based on IoT technology have been discussed. In a context of control theory, cyber physical systems are regard as a new system for the next generation. In this the- sis, estimation of quadrotor UAV and distributed estimation over delayed sensor networks are studied as a main problem of cyber physical systems. In Chapter 3, we propose a maximum-likelihood-estimation method for a quadrotor UAV given the existence of sensor delays. The state equation of the UAV is nonlinear, and thus, we propose an approximated method that consists of two steps. The first step estimates the past state based on the delayed output through an extended Kalman filter. The second step involves calculating an estimate of the present state by simulating the original system from the past to the present. It is proven that the proposed method provides an approximated maximum-likelihood- estimation. The e ff ectiveness of the estimator is verified by performing experiments. In Chapter 4, a distributed delay-compensated observer for wireless sensor network with de- lay is proposed. Each node of the sensor network aggregates data from the other nodes and sends the aggregated data to the neighbor nodes. In this communication, each node also compensates communication delays among the neighbor nodes. Therefore, all node can synchronize sensor measurements by using the scalable and local communication in real-time. All nodes estimate the state variables of a system simultaneously. The observer in each node is similar to the ex- iting delay-compensated observer with multi-sensor delays. Convergence rates of the proposed observer can be arbitrarily designed regardless of communication delays. The e ff ectiveness of the proposed method is verified by numerical simulation. In Chapter 5, we consider the design problem of an unknown-input observer for distributed network systems under existence of communication delays. In the proposed method, each node estimates all states and calculates inputs from own estimate. It is assumed that the controller of each node is given by an observer-based controller. In the calculation of each node, input values of the other nodes cannot be utilized. Therefore, each node calculates alternative inputs instead of the unknown input of the other nodes. The alternative inputs are generated by own estimate based on the feedback controller of the other nodes given by the assumption. Each node utilizes these values in the calculation of the estimation and the delay compensation instead of the unknown inputs. The stability of the estimation error of the proposed observer is proven by a Lyapunov-Krasovskii functional. The stability condition is given by a linear matrix inequality (LMI). Finally, a result of a numerical simulation is shown to verify the e ff ectiveness of the proposed method.
(主査) 教授 山下 裕, 教授 金子 俊一, 教授 近野 敦, 准教授 小林 孝一
情報科学研究科(システム情報科学専攻)
コレクション(個別)国立国会図書館デジタルコレクション > デジタル化資料 > 博士論文
受理日(W3CDTF)2019-06-03T12:31:31+09:00
連携機関・データベース国立国会図書館 : 国立国会図書館デジタルコレクション