|
| 講演者 |
杉江俊治 教授 京都大学 情報学研究科 システム科学専攻 |
|
|---|---|
| 講演題目 | 「射影型学習に基づく連続時間システムの同定」 |
| 概要 |
|---|
| 制御対象の精密なモデルを得ることは制御系設計において最も重要な問題の一つである． 従来の同定手法の多くは，離散時間モデルを同定するものがほとんどであるが， サンプリングされた入出力信号から連続時間モデルを直接に同定する直接法が近年注目されてきている． これは，現実のシステムは元来連続時間系であるものがほとんどであり，制御系設計の際には連続時間系として捉える方が便利なことが多いためである． このような直接法の一つとして，著者らは近年，射影型反復学習制御を用いる手法を提案してきている． 本セミナーでは，連続時間系の同定の問題点を示すとともに，提案手法の原理，性質およびその有効性について，数値例を用いながら述べる． |
| 問い合せ先 |
|---|
|
藤田政之 機械制御システム専攻 電話/FAX: (03)5734-3813 E-mail: fujita@ctrl.titech.ac.jp |
|
| 講演者 |
Professor Dr.-Ing. Bernd Tibken Faculty of Electrical and Information Engineering University of Wuppertal, Germany |
|
|---|---|
| 講演題目 | 「From Lyapunov Stability Theory to New Computational Algorithms」 |
| 概要 |
|---|
| In this talk the central results from Lyapunov Stability Theory are reviewed and new computational methods for a constructive approach will be presented. In the past the test of a system for asymptotic stability has very often been carried out by selecting an appropriate quadratic Lyapunov function and the use of sometimes rough estimates in order to ensure the negativity of the derivative of the Lyapunov function along the trajectories of the nonlinear dynamical system under consideration. These methods will be reviewed and some limitations of this approach will be shown for simple examples. In the following the talk will focus on the computational problems involved, e.g., the definiteness test for functions and the estimation of the region of attraction of an asymptotically stable stationary point of the nonlinear system. I will present also an introduction into modern developments involving semidefinite programming (SDP) approaches and new decompositions of polynomials as sums of squares (SOS). The talk will end with a conclusion and an outlook. |
| 問い合せ先 |
|---|
|
藤田政之 機械制御システム専攻 電話/FAX: (03)5734-3813 E-mail: fujita@ctrl.titech.ac.jp |
|
| 講演者 |
Professor Dr.-Ing. Oliver Sawodny Institute for System Dynamics University of Stuttgart, Germany |
|
|---|---|
| 講演題目 | 「Analysis and Control Synthesis for Nonlinear and Distributed Parameter Systems by Flatness-based Methods」 |
| 概要 |
|---|
| In the talk at three examples, which has been treated at the institute during the last years, about modelling and simulation of systems with dominant nonlinearities or distributed parameters were presented. As suitable control approaches flatness based methods are discussed. As an example for systems with significant nonlinearities pneumatic actuators are introduced. The dynamic modelling of the valve flow and the pressure in the chambers of the cylinder is explained in detail. Secondly, a heating process in glas industry is introduced as an example for a distributed parameter system. Via the describing partial differential equations optimized control strategies for the plant are derived in order to achieve minimized energy consumption and to shorten the set-up time in case of production changes. Last example is the active oscillation damping control for a fire rescue turntable ladder. New approaches for the dynamic model via partial differential equations are presented. Based on a multi body approach the system for the control is discussed. |
| 問い合せ先 |
|---|
|
藤田政之 機械制御システム専攻 電話/FAX: (03)5734-3813 E-mail: fujita@ctrl.titech.ac.jp |
「東工大機械系21世紀COE講演会」
|
| 講演者 |
Professor Mark W. Spong Department of General Engineering University of Illinois at Urbana-Champaign, Urbana, IL |
|
|---|---|
| 講演題目 | 「Coordination of Multi-Agent Systems」 |
| 概要 |
|---|
| The problem of coordination of multi-agent systems arises in numerous applications, both in natural and in man-made systems. Examples found in nature include the schooling of fish, swarming of insects, synchronously flashing fireflies, contraction of heart muscles, and firing of neurons in the brain. Applications in engineering arise in superconducting Josephson junction arrays, design of oscillator circuits, and coordination of multiple robots and sensor networks. In this talk we will present some new ideas on coordination and control of multiple-robot networks. These problems are challenging not only because of the nonlinear dynamics of the individual agents but also because of the constraints of communication among the agents. For example, the agents may only be able to communicate with their nearest neighbors and the communication may be over an unreliable medium, such as a wireless network or the Internet, which introduces the problems of random delays and packet losses. We will discuss the role of passivity as a unifying paradigm to design controllers for these systems. As applications of our results we will discuss the general problem of synchronization of multiple Lagrangian systems and multi-robot teleoperation with time delay. |
| 講演者 |
Professor Francesco Bullo Department of Mechanical Engineering University of California at Santa Barbara, Santa Barbara, CA |
|
|---|---|
| 講演題目 | 「Motion Coordination For Multi-Agent」 |
| 概要 |
|---|
| Motion coordination is an extraordinary phenomenon in biological systems and a powerful tool in man-made systems. Although individual agents have no global knowledge of the system, complex behaviors emerge from local interactions. The subject of this talk is the design of motion-enabled sensor networks, i.e., networks where nodal motions are purposefully induced in order to perform useful tasks. Example scenarios include how to deploy sensor nodes in a region of interest and how to exploit mobility in target tracking or boundary estimation. The algorithms combine distributed feedback, information processing and geometric structures. The key technical challenge, and the subject of ongoing research, is to design adaptive behaviors that tolerate asynchronicity and communication/process failures. |
| 問い合せ先 |
|---|
|
藤田政之 機械制御システム専攻 電話/FAX: (03)5734-3813 E-mail: fujita@ctrl.titech.ac.jp 三平満司 機械制御システム専攻 電話/FAX : （03）5734-2552 E-mail : sampei@ctrl.titech.ac.jp |
「東工大機械系21世紀COE講演会」
|
| 講演者 |
Professor Dr.-Ing. Dr.-Ing.h.c. Gunther Schmidt Interactive Systems and Control Group Technical University of Munich, Munich, Germany |
|
|---|---|
| 講演題目 | 「Artificial Cognitive Functionalities in Vision-guided Humanoid Walking」 |
| 概要 |
|---|
| The integration of basic cognitive functionalities into the operation of real-time, closed-loop systems seems to be one of the major trends in control. Potential applications are found in vehicle and defence systems or robotics. With respect to biological models we distinguish perceptive, mental und executive cognitive functionalities. It is well-known that there also exists a close relationship between cognition and notions such as intelligence und autonomy. Based on recent results from theoretical and experimental research work in the field of intelligent humanoid robot walking this talk will demonstrate how some degree of locomotion autonomy can be achieved on the basis of cognitive functionalities, such as task-dependent visual perception and gaze control, object classification and decision making as well as step sequence planning and adaptation. |
| 問い合せ先 |
|---|
|
藤田政之 機械制御システム専攻 電話/FAX: (03)5734-3813 E-mail: fujita@ctrl.titech.ac.jp |