Abstract
〈Vol.3 No.1(2010.1)〉
Titles
[Invited Paper]
[Contributed Papers]
▲
■ Control of Dynamic Graphs
Santa Clara University・Aleksandar I. ZECEVIC,
and Dragoslav D. SILJAK
In this paper we develop a general method for the robust stabilization of dynamic graphs. The proposed approach is based on linear matrix inequalities, and can accommodate a broad range of uncertainties and information structure constraints. It is shown that control laws of this type can be applied to graphs in which both the nodes and the edges exhibit time-varying behavior. In all cases, the control design can be formulated as a convex optimization problem.
▲
■ Subionospheric VLF/LF probing of ionospheric perturbations associated
with earthquakes: A possibility of earthquake prediction
The University of Electro-Communications・
Masashi HAYAKAWA, Takumi HORIE,
Fumiya MUTO, Yasushi KASAHARA,
Chubu University・Kenji OHTA,
National Central University・J. Y. LIU,
and The University of Electro-Communications・
Yasuhide HOBARA
The VLF(Very Low Frequency) / LF(Low Frequency) receiving network is established
in Japan, composed of seven observing stations (Moshiri (Hokkaido), Chofu
(UEC, University of Electro-Communications), Tateyama (Chiba), Shimizu,
Kasugai (Nagoya), Maizuru (Kyoto) and Kochi), and a few additional foreign
stations are established in Kamchatka, Taiwan and Indonesia. At each station
we observe simultaneously several VLF/LF transmitter signals (Japanese
JJI (40 kHz, Fukushima), JJI (Ebino, Kyushu)), and foreign VLF transmitters
(NWC (Australia), NPM (Hawaii), NLK (USA)). This Japanese VLF/LF network
is used to study the ionospheric perturbations associated with earthquakes,
and we present two recent results; (1) a statistical result on the correlation
between VLF/LF propagation anomalies and earthquakes and (2) the latest
results during the last six monthson the two particular propagation paths;
JJY-Moshiri and JJY-Taiwan. Some discussions are done on the correlation
of ionospheric perturbations with the earthquakes in the sense of a possibility
of earthquake prediction by means of VLF propagation anomalies.
▲
■ Some Learning Properties of Modular Network SOMs
Kyoyo University・Manabu TAKEDA,
Nara Institute of Science and Technology・
Kazushi IKEDA, and
Kyushu Institute of Technology・Tetsuo FURUKAWA
The Modular Network Self-Organizing Map (mnSOM) is a generalization of the SOM, where each node represents a parametric function such as a multi-layer perceptron or another SOM.
Since there are, in general, fewer given datasets than nodes, some nodes never win in competition and have to update their parameters from the winners in the neighborhood. This is a process that can be regarded as interpolation. This study derives the interpolation curve between winners in simple cases and discusses the distribution of winners based on the neighborhood function.
▲
■ Polynomial Optimization Approach to Stability Analysis of Boundary Controlled
2-Dimensional Fluid Flow
Tokyo Institute of Technology・Yukihiro SAKAI,
Schlumberger, K.K.・Hwayeong YU,
Tokyo Institute of Technology・Kenji KASHIMA,
and Jun-ichi IMURA
In this paper, the stability of incompressible 2D channel flow is investigated.
This problem has many practical applications such as mixing fuel and air
in combustion engine. However, since the flow dynamics is described by
a nonlinear partial differential equation, its stability analysis is a
challenging problem. Most of existing results obtained in an analytical
way are conservative and less flexible. In this paper, we attempt to derive
less conservative and easily checkable criteria for the problem via sum
of squares relaxation technique.
▲
■ Adaptive Design of Role Differentiation by Division of Reward Function in Multi-agent Reinforcement Learning
Ritsumeikan University・Tadahiro TANIGUCHI,
Kyoto University・Kazuma TABUCHI,
and Tetsuo SAWARAGI
There are several problems which discourage an organization from achieving tasks, e.g., partial observation, credit assignment, and concurrent learning in multi-agent reinforcement learning. In many conventional approaches, each agent estimates hidden states, e.g., sensor inputs, positions, and policies of other agents, and reduces the uncertainty in the partially-observable Markov decision process (POMDP), which partially solve the multiagent reinforcement learning problem. In contrast, people reduce uncertainty in human organizations in the real world by autonommously dividing the roles played by each agent. In a framework of reinforcement learning, roles are mainly represented by goals for each agent. This paper presents a method for generating internal rewards from manager agents to worker agents. It also explicitly divides the roles, which enables a POMDP task for each agent to be transformed into a simple MDP task under certain conditions. Several situational experiments are also described and the validity of the proposed method is evaluated.
▲
■ A New Model Predictive Control Approach to DC-DC converter Based on Combinatory Optimization
Chiba University・Abu Zaharin AHMAD, and Kang-Zhi LIU
DC-DC converters by nature present hybrid behavior, which is described by a set of discrete modes associated with continuous dynamics. The control objective is accomplished by switching among the discrete modes. This paper presents a new model predictive control approach to optimize the performance of DC-DC converters. The proposed control algorithm is tailored to DC-DC converters, making full use of the fact of finite number of modes to transform the performance optimization problem into a combinatory optimization task. This approach has made the on-line computation very simple and effective for implementation.
▲ ■ Nonlinear Speed Control Scheme and Its Stability Analysis for SI Engines
Sophia University・Jiangyan ZHANG, Tielong SHEN,
and The University of Rome・Riccardo Marino
For international combustion engines, due to the combustion cyclic nature,
the intake-to-power stroke delay is inherent that causes additional difficulties
in control design and validation phases. In this paper, a nonlinear speed
control scheme is proposed based on the proportional feedback control method.
From the consideration of improving transient performance, a reference
model is introduced to design the feedback controller. Then, the speed
controller is formulated as a designed feedback control law connecting
with a model-based feedforward compensation. The asymptotic convergence
to the desired speed is guaranteed under the presented conditions of the
feedback gains, which include the cases of using a speed-depended gain
function and a constant gain, respectively. For the stability analysis
of the proposed delayed control system, an initial method is presented
via Lyapunov-Krasovskii functional stability theorem. Experimental results
on the transition speed control are shown to demonstrate the control scheme.
▲
■ Intertransaction Class Association Rule Mining based on Genetic Network
Programming and Its Application to Stock Market Prediction
Kobe University・Yuchen YANG, Shingo MABU,
Kaoru SHIMADA, and Kotaro HIRASAWA
Intertransaction association rules have been reported to be useful in many
fields such as stock market prediction, but still there are not so many
efficient methods to dig them out from large data sets. Furthermore, how
to use and measure these more complex rules should be considered carefully.
In this paper, we propose a new intertransaction class association rule
mining method based on Genetic Network Programming (GNP), which has the
ability to overcome some shortages of Apriori-like based intertransaction
association methods. Moreover, a general classifier model for intertransaction
rules is also introduced. In our experiments on the real world application
of stock market prediction, our method shows its efficiency and ability
to obtain good results and can bring more benefits with a suitable classifier
considering larger interval span.
▲
■ Blockage Detection in Pipelines using a Flexible Piezoelectric Film
Sensor
AIST・Nan BU, Naohiro UENO, and Osamu FUKUDA
This paper presents a novel blockage detection method using a flexible piezoelectric film sensor. This sensor is made of oriented aluminum nitride (AlN) thin film, and the total thickness is less than 40 μm. The thin thickness makes this sensor sensitive to stress in the length direction, and in the meanwhile flexibility facilitates installing this sensor to the external surfaces of pipes. With this sensor, pressure pulsation in fluid can be measured non-invasively and a blockage detection method is developed using transfer function analysis. In the experiments, linearity and response characteristics of the AlN film sensor in the length direction was studied. Then experiments of partial blockage detection were conducted using a test pipe system. Validation of the proposed method is confirmed through the experimental results.
|
