電気学会論文誌C(電子・情報・システム部門誌)
Online ISSN : 1348-8155
Print ISSN : 0385-4221
ISSN-L : 0385-4221
131 巻, 5 号
選択された号の論文の23件中1~23を表示しています
特集:メタヒューリスティクスとその応用
特集論文
<システム・計測・制御>
<知能,ロボティクス>
<音声画像処理・認識>
  • 中山 史朗, 白川 真一, 矢田 紀子, 長尾 智晴
    2011 年 131 巻 5 号 p. 958-965
    発行日: 2011/05/01
    公開日: 2011/05/01
    ジャーナル フリー
    Automatic construction method for image classification algorithms has been required. Genetic Image Network for Image Classification (GIN-IC) is the automatic construction method for image classification algorithms which include image transformation component using evolutionary computation, and its effectiveness has already been proven. In our study, we try to improve the performance of GIN-IC with AdaBoost algorithm using GIN-IC as weak classifiers to complement with each other. We apply our proposed method to three types of image classification problems, and show the results in this paper. In our method, discrimination rates for training images and test images improved in the experiments compared with the previous method GIN-IC.
<ソフトコンピューティング・学習>
  • 水野 晴規, 岡本 卓, 小圷 成一, 平田 廣則
    2011 年 131 巻 5 号 p. 966-975
    発行日: 2011/05/01
    公開日: 2011/05/01
    ジャーナル フリー
    A complex network design method that finds a desired network structure can become one of strong tools in large-scale system designs. Conventional complex network design methods only tackle static networks, that is, they do not consider the growth of a target network. In this study, we propose a new growing complex network design method. First, let us consider evalution functions which quantitatively express characteristics of desired stuructures using feature quantities. Then, we formulate a growing complex network design problem as a multi-objective optimization problem in order to determine connection targets of a new node using the evaluation functions. Solving the problem, we grow the network, then, we obtain a desired network. We try to generate networks which have desired clustering coefficient and average path concurrently. Through numerical experiments, we confirmed the proposed method is effective as a growing complex network design method.
  • 永吉 雅人, 村尾 元, 玉置 久
    2011 年 131 巻 5 号 p. 976-982
    発行日: 2011/05/01
    公開日: 2011/05/01
    ジャーナル フリー
    Reinforcement Learning (RL) attracts much attention as a technique of realizing computational intelligence such as adaptive and autonomous decentralized systems. In general, however, it is not easy to put RL into practical use. This difficulty includes a problem of designing a suitable action space of an agent, i.e., satisfying two requirements in trade-off: (i) to keep the characteristics (or structure) of an original search space as much as possible in order to seek strategies that lie close to the optimal, and (ii) to reduce the search space as much as possible in order to expedite the learning process.
    In order to design a suitable action space adaptively, we propose switching RL model to mimic a process of an infant's motor development in which gross motor skills develop before fine motor skills. Then, a method for switching controllers is constructed by introducing and referring to the “entropy”. Further, through computational experiments by using robot navigation problems with one and two-dimensional continuous action space, the validity of the proposed method has been confirmed.
  • 鈴木 聡, 満倉 靖恵
    2011 年 131 巻 5 号 p. 983-989
    発行日: 2011/05/01
    公開日: 2011/05/01
    ジャーナル フリー
    The purpose of this paper is to optimize the structure of hierarchical neural network. In this paper, structure optimization is to represent neural network by minimum number of nodes and connections, and is performed by eliminating unnecessary connections from trained neural network by using genetic algorithm. We focus on the neural network which specialized for image recognition problems. The flow of the proposed method is as follows. Firstly, walsh-hadamard transform is applied to images for feature extraction. Secondly, neural network is trained with extracted features based on back-propagation algorithm. After neural network training, unnecessary connections are eliminated from trained neural network by utilizing genetic algorithm. Finally, neural network is retrained to recover the degradation caused by connection elimination. In order to validate the usefulness of the proposed method, face recognition and texture classification examples are used. From the experimental results, it was shown that compact neural network was generated, keeping generalization performance by proposed method.
  • 増田 和明, 栗原 謙三
    2011 年 131 巻 5 号 p. 990-999
    発行日: 2011/05/01
    公開日: 2011/05/01
    ジャーナル フリー
    This paper proposes a constrained global optimization method based on Multi-Objective Particle Swarm Optimization (MOPSO). A constrained optimization problem is transformed into another bi-objective problem which minimizes both the original objective function and the total amount of constraint violations. Then, the global optimum of the former problem is obtained as the Pareto optimal solution of the latter one having no constraint violation. In order to find the particular Pareto optimal solution, the proposed method introduces to MOPSO the following operations such as (a) restricting the number of Pareto optimal solutions obtained at each iteration of MOPSO to urge particles to approach the feasible set of the original constrained problem, (b) choosing the most promising Pareto optimal solution as the global best solution so as to exclude solutions dominated by it, and (c) encouraging to add Pareto optimal solutions if the number of them is too small to recover the diversity of search. Numerical examples verify the effectiveness, efficiency and wide applicability of the proposed method. For some famous engineering design problems, in particular, it can find solutions which are comparative to or better than the previously known best ones.
  • 原 翔悟, 元木 誠, 山崎 洋一, 安田 恵一郎
    2011 年 131 巻 5 号 p. 1000-1008
    発行日: 2011/05/01
    公開日: 2011/05/01
    ジャーナル フリー
    In this paper, we perform experiments that autonomous mobile robot controllers are developed in by the integrative optimization system. Autonomous mobile robot controllers are developed using meta-heuristics, if the behavior of robot is able to evaluate. However, in an actual environment, it is difficult to develop the robot controller using meta-heuristics because meta-heuristics requires large numbers of function evaluations. And large numbers of function evaluations have problems which are to increase experimental time and the maintenance cost. The integrative optimization system is a method to improve these problems. A general integrative optimization system constructs an approximate response surface by the Radial Basis Function network and optimizes to the approximate response surface by the optimization method. Through experiments in a simulator and actual environment, the authors show that it is an effective method to develop the autonomous mobile robot controller by the integrative optimization system.
  • 李 静銀, 玄 光男, 李 瓊球, 李 羲頡
    2011 年 131 巻 5 号 p. 1009-1019
    発行日: 2011/05/01
    公開日: 2011/05/01
    ジャーナル フリー
    This paper deals with the building of the reusable reverse logistics model considering the decision of the backorder or the next arrival of goods. The optimization method to minimize the transportation cost and to minimize the volume of the backorder or the next arrival of goods occurred by the Just in Time delivery of the final delivery stage between the manufacturer and the processing center is proposed. Through the optimization algorithms using the priority-based genetic algorithm and the hybrid genetic algorithm, the sub-optimal delivery routes are determined. Based on the case study of a distilling and sale company in Busan in Korea, the new model of the reusable reverse logistics of empty bottles is built and the effectiveness of the proposed method is verified.
  • 小熊 祐司, 相吉 英太郎
    2011 年 131 巻 5 号 p. 1020-1030
    発行日: 2011/05/01
    公開日: 2011/05/01
    ジャーナル フリー
    Particle Swarm Optimization (PSO), a meta-heuristic global optimization method, has attracted special interest for its simple algorithm and high searching ability. The updating formula of PSO involves coefficients with random numbers as parameters to enhance diversification ability in searching for the global optimum. However, the randomness makes stability of the searching points difficult to be analyzed mathematically, and the users need to adjust the parameter values by trial and error. In this paper, stability of the stochastic dynamics of PSO is analyzed mathematically and exact stability condition taking the randomness into consideration is presented with an index “statistical eigenvalue”, which is a new concept to evaluate the degree of the stability of PSO dynamics. Accuracy and effectiveness of the proposed stability discrimination using the presented index are certified in numerical simulation for simple examples.
<情報システム,エレクトロニック・コマース>
  • 加藤 能史, 濱上 知樹
    2011 年 131 巻 5 号 p. 1031-1037
    発行日: 2011/05/01
    公開日: 2011/05/01
    ジャーナル フリー
    A new genetic algorithm for multi agent based autonomous power distribution network restoration system is proposed. The state of the art of this study is to realize a new genetic algorithm using selective sampling for improving the restoration performance. A proposed method realizes the selective sampling by a virtual accident selecting algorithm that changes probability of selecting virtual accident area. The virtual accident selecting algorithm consists of weight table and area-value list. The weight table represents a difficulty of restoration in each accident area. The area-value list represents a difficulty of restoration in latest generation, and effects on weight table in next generation. This architecture enables the system to change the probability of changing each virtual accident area autonomously from restoration simulation. The simulation results show the proposed method achieves to improve the performance.
特集資料
<システム・計測・制御>
  • 森井 宣人, 相吉 英太郎
    2011 年 131 巻 5 号 p. 1038-1042
    発行日: 2011/05/01
    公開日: 2011/05/01
    ジャーナル フリー
    The ant colony method is one of heuristic methods capable of solving the traveling salesman problem (TSP), in which a good tour is generated by the artificial ant's probabilistic behavior. However, the generated tour length depends on the parameter describing the ant's behavior, and the best parameters corresponding to the problem to be solved is unknown. In this technical note, the evolutional strategy is presented to find the best parameter of the ant colony by using Particle Swarm Optimization (PSO) in the parameter space. Numerical simulations for benchmarks demonstrate effectiveness of the evolutional ant colony method.
論文
<システム・計測・制御>
<知能,ロボティクス>
  • 森岡 一幸, 追永 祐大, 中村 雄一
    2011 年 131 巻 5 号 p. 1050-1058
    発行日: 2011/05/01
    公開日: 2011/05/01
    ジャーナル フリー
    This paper proposes the localization method based on interactive communication between a mobile robot and a networked laser range scanner installed in an intelligent space and achieves human-following control of a mobile robot with the method. Generally, human tracking with cameras or laser range scanners on board the robots has been utilized for control of mobile robots to follow human walking. In addition to human tracking, mobile robots have to perform position estimation simultaneously. There is constraints in measurement for landmark detection or SLAM, because target human walks close to the robot while human following. Then, proposed system consiers to utilize an intelligent environment where sensors are distributed.
    The proposed system exchanges position and heading information estimated in the mobile robot and the networked laser range scanner with each other. The networked laser range scanner searches and detects target human and the robot based on the position information sent from the robot. The robot receives the detection results from the networked laser range scanner. Then, the estimate position is updated and reference velocities for human-following control are calculated with them. Estimation errors with odometry in the robot and unstable tracking of target in the networked laser range scanner are compensated with this system.
    In this paper, communication timing between the robot and the networked laser range scanners while human-following is discussed. Human-following experiments are performed and the results are shown.
  • 村田 知也, 田村 信介, 川井 昌之
    2011 年 131 巻 5 号 p. 1059-1067
    発行日: 2011/05/01
    公開日: 2011/05/01
    ジャーナル フリー
    BFA (Backtrack Free path planning Algorithm) has been implemented for calculating paths of manipulators behaving in 3-dimensional work spaces. In the implementation, a method to avoid collisions between links also has been proposed. This paper also discusses an approach to extending BFA for path plannings of cooperating multi manipulators. In the approach, multiple manipulators are considered as a single composite one with many links. Simulation results demonstrated that BFA enabled the efficient generation of paths both for single and multi manipulators. The algorithm is backtrack free and resolution complete. Computation volume of the algorithm is proportional to the total number of links and does not change with environments where manipulators behave.
<情報システム,エレクトロニック・コマース>
研究開発レター
<電気回路・電子回路>
<情報通信工学>
<音声画像処理・認識>
部門記事
feedback
Top