Abstract
〈Vol.4 No.2(2011.3)〉
Titles
[Contributed Papers]
▲
■ AGLSDC: A Genetic Local Search Suitable for Parallel Computation
Tottori University・Shuhei KIMURA, Kogakuin University・Takashi NAKAKUKI,
Tottori University・Seiji KIRITA, RIKEN・Mariko OKADA
Because evolutionary algorithms (EAs) generally require many repeated evaluations
of objective functions, it often takes considerable time to solve optimization
problems. Parallel computation is one means to shorten the required computation
time. In earlier works, the authors proposed an EA suitable for coarse-grained
parallel computers, a genetic local search with distance independent diversity
control (GLSDC). Though GLSDC has been applied successfully to several
practical problems, its parallel efficiency abruptly drops off as the number
of CPUs for computation increases. To achieve a higher parallel efficiency,
the authors now propose a new EA, an asynchronous GLSDC (AGLSDC), constructed
by reworking the algorithm of GLSDC. This paper introduces the proposed
method and reports verification of the method through numerical experiments
on several benchmark problems and a practical problem.
▲
■ Stability and Control Performance of Robust Model Reference Adaptive
Control System: An Approach Based on Its Generalized Structure
National Defense Academy・Seitaro TANAHASHI and Keietsu ITAMIYA
This paper proposes a new analytical method for stability and control performance
of a robust model reference adaptive control system (MRACS). In the past
studies, the following analytical policy has been adopted, that is, deriving
tight and quantitative conditions for a stable robust MRACS is not important
since those conditions depend on unknown factors such as the transfer characteristic
of the controlled object. As a result, the most important things about
stability and control performance disappear in algebraic calculations based
on many inequalities. Naturally, it is not easy to answer the following
questions. How is the stabilization of a robust MRACS achieved? What can
we prepare for design of a robust MRACS which has a sufficient stability
margin and a better control performance? The motivation of this paper is
to answer these questions. For the purpose, this paper defines a desired
system for the model reference control, and uses an analytical model parameterized
with the desired controller parameters. Especially in a main loop analysis,
a certain equivalent feedback system is derived, and the stability of a
robust MRACS is explained with an expanded notion of the
L2 stability. According to this method, it is possible to understand stabilization
of a robust MRACS. Furthermore, some useful findings for design of a robust
MRACS can be obtained from the results of analysis and a numerical simulation.
▲
■ Q Value-Based Dynamic Programming with Boltzmann Distribution in Large
Scale Road Network
Waseda University・Shanqing YU, Yelei XU, Shingo MABU
,Manoj Kanta MAINALI, Kaoru SHIMADA and Kotaro HIRASAWA
In this paper, a global optimal traffic assignment strategy, i.e., Q value-based
Dynamic Programming with Boltzmann Distribution is applied to the Kitakyushu
City traffic system. The main idea of the proposed traffic assignment strategy
is to calculate the expected traveling time for each origin-destination
pair and the probability of selecting the next section, then to generate
a considerable number of route candidates for the drivers based on the
calculated probability.
In the simulation, how to select the temperature parameter and the number
of the route candidates is discussed in detail. The comparison between
the proposed method and the shortest path algorithms indicates that the
proposed method could reduce the risk of the traffic congestion occurrence
and save the traveling cost effectively. In addition, the computation time
is given to reveal the feasibility of the proposed method in large scale
networks.
▲
■ Exact Linearization of Two-Input Affine Systems via the Dynamic Extension
Based on the Relative Degree Structure
Tokyo Institute of Technology・Kazuma SEKIGUCHI , Mitsuji SAMPEI,
and Sasebo National College of Technology・Shigeki NAKAURA
This paper discusses the exact linearization problem of two-input affine
systems via the dynamic extension based on the relative degree structure.
A necessary and sufficient condition for two-input systems to be exactly
linearizable via 1-degree dynamic extension is derived, and the input transformations
of two-input systems with dynamic extension are classified into two forms.
Finally, a method to derive the input transformation such that the transformed
system is exactly linearizable via the dynamic extension is proposed. The
proposed method is applied to a mechanical system.
▲
■ Gain-Scheduled Output-Feedback Controllers Using Inexactly Measured
Scheduling Parameters for Linear Parametrically Affine Systems
Japan Aerospace Exploration Agency・Masayuki SATO
This note considers the design problem of Gain-Scheduled (GS) controllers for Linear Parameter-
Varying (LPV) systems whose state-space matrices are supposed to be parametrically
affine with some mild constraints. To be precise, we consider two types
of GS controllers, GS H2 and GS H∞ controllers. The controllers to be
designed are supposed to have parametrically affine state-space matrices.
In sharp contrast to conventional GS controller design methods, the scheduling
parameters are supposed to be inexactly measured. Using parameter-independent
Lyapunov functions, in which some conservatism is admitted, we give sufficient
conditions for designing GS H2 and GS H∞ outputfeedback controllers which
are robust against measurement errors of the measured scheduling parameters.
The proposed methods are formulated in terms of parametrically affine Linear
Matrix Inequalities (LMIs) with single line search parameters The methods
recover the design methods of conventional GS output-feedback controllers,
in which scheduling parameters are supposed to be exactly measured, with
an appropriate choice of the line search parameters. A simple numerical
example is included to illustrate our results.
▲
■ A Bilinear H∞ Controller Using a Bilinear Disturbance Isolation Observer
for a Heat Exchanger Containing Unknown Disturbance Flows
Meiji University・Shin NAKAYAMA and Hiroaki KOBAYASHI
A heat exchanger controlled by manipulating flow rates is widely used and
there often exist unknown disturbance flows in the system. If we approximate
the distributed parameter system to a lumped parameter system by means
of Method of Weighted Residuals (MWR) and apply a controller to depress
the disturbance influence, we must take the following three points into
account: (a) the system has bilinear terms with respect to not only the
input but also the unknown disturbance, (b) we cannot observe the state
vector directly because the components are generalized Fourier coefficients
so they are not physical quantities, and (c) we must rely on estimators
to obtain the internal temperature accurately due to the heat exchanger
structure. In this paper, we consider H∞ control problems for such a bilinear
system and design a bilinear H∞ controller using a bilinear observer.
As a bilinear observer, we use ”a bilinear disturbance isolation observer
(DIO)” which is an extension of a previous work. The performance of this
observer-based H∞ controller is evaluated with several simulations and
experiments.
▲ ■ Diagonalization of Arm Kinematics by the Use of Bi-Articular Muscles
Ritsumeikan University・Takashi MITSUDA, Sadao KAWAMURA
, Mitsunori UEMURA and University of Toyama・Masahiro SEKIMOTO
Mapping from visual information to motor commands is an important issue
in the study of voluntary arm movement. The joint kinematics of a human-like
planar arm, which is generally described as a nonlinear function in a Cartesian
coordinate system, can be approximately linearized in a head-centered polar
coordinate system. This study shows that the mapping from the muscle lengths
to the hand position in a head-centered polar coordinate system can also
be approximately linearized. In addition, this study proposes a diagonalization
of the kinematics that involves adjusting the moment arms of the muscles;
in this diagonalization, the muscle lengths of two bi-articular muscles
are proportional to the distance and direction from the head.
▲
■ Keystroke Dynamics for Individual Identification in Japanese Free Text
Typing
Akashi National College of Technology・Toshiharu SAMURA
and University of Hyogo・Haruhiko NISHIMURA
We develop Web-based typing software in order to collect keystroke data
on a LAN, and investigate several characteristics of keystroke dynamics
in Japanese free text typing. We perform experiments on 112 subjects, representing
three groups according to the number of letters they could type within
five minutes. In this experiment, we extract the feature indices from the
keystroke timing for each single letter and for two-letter combinations
composed of consonant and vowel pairs in Japanese text. Based on an identification
method using a weighted Euclidean distance, personal identification for
the three groups is evaluated and its high performance is confirmed in
proportion to the typing level of the group.
▲
■ Intelligent Medicine Case for Dosing Monitoring: Design and Implementation
University of Tsukuba・Takuo SUZUKI and Yasushi NAKAUCHI
In advanced countries, the number of elderly recipients who take medicines
every day has been increasing due to the aging of the population. Since
they have many risks of inappropriate medication (e.g. overdose), we develop
an intelligent medicine case named iMec in order to assist their caregivers
in monitoring their medication conditions.The medicine case can recognize
whether a recipient picks up medicines from its storage space to check
the adequacy of dosage, and can estimate the present behavioral condition
of the recipient to check the adequacy of dosing timing.
Therefore, it can detect signs of inappropriate medication by comparing
the results with the correct dosage and dosing timing. The present medication
condition confirmed by the medicine case is stored on a server computer
so that caregivers can monitor the information through web interfaces.
Moreover, emergency e-mails are sent to the caregivers when the medicine
case detects a sign of inappropriate medication. We use the Hough transformation
and background subtraction methods for dosage confirmation, and devise
a novel Fuzzy inference method based on association rules for dosing timing
confirmation. As results of experiments and simulations, we confirm that
the medicine case is able to detect 4 kinds of medicines in its storage
space, and is able to select the present behavioral condition from 11 kinds
of candidate behavioral conditions corresponding to common dosing timings.
▲
■ Generation and Adaptation of Transferable Roadway Model for Anticipative
Road Following on Satellite-Roadway-Vehicle Network
Osaka Institute of Technology・Kohji KAMEJIMA
By matching roadway images in an encountered scene with bird’s eye views,
the scope of humans’ perception is expanded to a satellite-roadway-vehicle
network. Based on the geometric consistency of the satellite images with
as-is local terrain, in this paper, a computational mechanism is introduced
for generation and adaptation of a roadway model transferable through the
network. First, the chromatic complexity of the roadway area is represented
as a palette of saliency colors via fractal sampling of the scene image.
Next, the palette is adapted to the associated area in the bird’s eye
view. Finally, the palette is transferred to the bird’s eye view for anticipatively
following the roadway pattern. Experimental results demonstrate that the
transferable model can be applied to the extension of the roadway pattern
prior to physical access.
▲
■ Multiple Cooperative Bilateral Teleoperation with Time-Varying Delay
Kanazawa University・Nam Duc DO, Yusuke YAMASHINA
and Keio University・Toru NAMERIKAWA
This paper deals with a passive-decomposition based control of bilateral
teleoperation between a single master robot and multiple cooperative slave
robots with time varying delay. First, we decompose the dynamics of multiple
slave robots into two decoupled dynamics by using the passive-decomposition:
the shape-system describing dynamics of the cooperative works and the locked-system
representing the overall behavior of the multiple slave robots. Second,
we propose a PD control method for bilateral teleoperation to guarantee
asymptotic stability of the system with time varying delay. Finally, experimental
results show the effectiveness of the proposed teleoperation.
|