[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

In this study, a monitoring system associated with a real-time machining simulator have been developed to detect machining trouble. Usually, monitoring signal is processed to compare with some reference value called a threshold. However, the thresholds depend on cutting conditions, and they are decided through various cutting experiments which requires extensive amount of time and efforts. In order to decide a suitable threshold dynamically during machining operation, a real-time machining simulator is integrated in a monitoring system.

View full abstract

Intelligent control of machining is introduced. Now, big jet airplanes are developed in Japan. Its purpose is to keep the ability of designers of airplanes in Japan. Works in U.S. had to develop new planes and spacecrafts in 1960 decade. They used new strong materials, however, effective cutting conditions of them were unknown. Adaptive control to determine automatically the optimum cutting speed and feedrate of machining tools was researched for such materials. The technology would become useful in Japan soon. The history of such technologies is introduced here.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

We propose an incremental learning by using self-organizing neural structure with dynamic and spatial changing weights for concept formation. An essential core of this self-organization is based on an extended Hebbian rule for the spatial changing and a self-organizing learning with incomplete information for the dynamic changing. A concept formation problem requires the neural network to acquire the complete information structure of a concept using a complete observation of the concept. The connection structure of self-organizing network can store with the information structure by using the two rules. The Hebbian rule can create a necessary connection corresponding to the blind complete information. On the other hand, self-organization can delete unnecessary connections. Finally incremental learning ability of the proposed neural network is proven on the concept formation problem under some conditions.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

We describe a construction of robust multi-robot system. Reinforcement learning that uses Bayesian discrimination method for segmenting the continuous state and action spaces simultaneously is applied to a robot for behavior acquisition. In order to adapt to environmental change, the state space has to be covered widely by effective rules with diversity. We perceive generated rules in learning process as effective rules. However, holding all generated rules isn’t efficient, the indicator of effective rule is needed. We propose the indicator that presents con-figuration state of rule and improve adaptability of system by considering configuration state of internal-model.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

Lip Reading is known as one of the word recognition method in high noise environments. We propose a hardware realization of lip reading system by using FPGA. By hardware realization of a lip reading system, high-speed and compact stand-alone realization is able to be expected. In this paper, configured for real time processing is expected and operation of each part in the system is conferred. Experimental results of lip reading are also included and are compared with ordinal system developed as software program to show the effectiveness of the proposed system.

View full abstract

Measurement of human static and dynamic visual acuity and other visual function is very important for traffic safety education. However, conventional test machine of human visual function has the problem that measurement time is long and the measurement value of kinetic eyesight is inaccurate. A new definition method of kinetic eyesight was proposed in this work. Moreover, we have developed a portable test machine of human visual function with multi-option, and its performance also has evaluated.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

Longitudinal power transmission system (LPS) being common in developing countries, they frequently encounter instabilities due to loss of power and voltage stability during summer months, particularly during afternoon periods diversity when they are subjected to heavy loading, the load density being non-uniform, the reactive demand being also high. During normal steady state operation, the security being low, these longitudinal lines often pressed to wheel heavy real power to industrial load centres. At times, these lines are used to transport higher magnitude of power beyond normal operating limit fixed terminal reactive support for some period of time, mostly some of the lines are required to feed power to an adjoining sub-grid which might be in deficit of desired real power source. This paper presents an approach using QV, PV or nose curves for determining proximity to collapse so that operators can take timely preventive measures to avoid losing the system. In this paper two methods are used. One is predictor and parameterization method and another is corrector method.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

We are developing human reasoning engine based on the fuzzy reasoning methods. This human reasoning engine will be explained in this paper. The following five reasoning algorithms and example of application are contained in this human reasoning engine. (1) Mamdani's fuzzy reasoning method. (2) Functional fuzzy reasoning method. (3) Simplified fuzzy method. (4)Product-Sum type fuzzy method. (5) Distance type fuzzy reasoning method.

View full abstract

[in Japanese]

View full abstract

This paper proposes a design method of two-degree-of-freedom(2DOF) robust servo system for control of a disk file head-positioning system. Proposed 2DOF robust servo system can attain a target value response and robust stability independently each other. In addition, in this paper the periodic disturbance caused by from disk rotation is considered. The periodic disturbance causes state estimation error and instability finally. In order to reduce the adverse effects of the periodic disturbance and improve the closed loop performance the proposed 2DOF and repetitive control are applied.

View full abstract

[in Japanese]

View full abstract

Robust control system should be designed based upon tolerances on time-response deterioration, since time responses are the very phenomena in the actual world. The step response shape, among others, is the most suitable for evaluation of system dynamic characteristics. According to the partial model matching developed by one of the present authors, control systems can be designed paying attention to the shape of the step responses. Presented here is its extension to design of robust control systems based on the shape of step responses. The design is reduced into a nonlinear programming problem, which can be easily solved with any well established computer program. The validity of the method will be shown through design of SISO continuous-time and sampled-data, and MIMO decoupling I-PD control systems.

View full abstract

[in Japanese]

View full abstract

In this paper, a Q-learning system using a prior information of plants is proposed, and a serial double inverted pendulum is swung up and stabilized by using the proposed method. The swing up control process consists of the following phase : swing up the first pendulum, swing up the second pendulum with stabilizing the first one, and stabilizing the two pendulums. 3 conventional controllers are used as prior information of the plant. A Q-learning system with adaptive Q-tables is used to obtain initial values that achieve successful result for each conventional controller. Effectiveness of the proposed method is shown by simulations and experiments.

View full abstract

In simulations, when a mobile robot acts, an action yields the corresponding state transition in deterministic environments. Whenever the robot selects the same action at a state, the position and the orientation of the robot are assumed to be identical after transitions. However, in a real environment, the same action at the same state can lead the robot to different states. The problem called the state-action deviation problem. To conquer this problem, a vision-based reinforcement learning system using Hough transformation is proposed. In the proposed method, state transitions of mobile robot are conducted considering its position and orientation. The effectiveness of the proposed method is shown by experiments in a real environment.

View full abstract

Reinforcement learning is an effective technique for making obtaining desirable actions under unknown environments without supervisors. However, multi-agent reinforcement learning using conventional methods requires many learning iterations to achieve admissible result. Therefore, in this paper, a multi-agent reinforcement learning system using CMAC is proposed. By using the CMAC, a learning effect can spread over similar state-action pairs. Thus the number of learning iterations can be reduced. Computer simulations are conducted for hunter problems to show effectiveness of the proposed method, and results of the proposed method are compared to those of conventional method.

View full abstract

Reinforcement learning is one of effective controller for autonomous robots. Because it does not need priori knowledge, and policies to complete given tasks are obtained automatically by repeating trial and error. However a large number of trials are required to realize complex tasks. So the task that can be obtained using the real robot is restricted to simple ones.We consider distributed autonomous agents, and propose a method to obtain an environment model that is independent of the task. And we show that the environment model is constructed quickly by sharing the experience of each agent, even when each agent has own independent task. And we demonstrate the leaning time is decreased by utilizing the model.

View full abstract

[in Japanese]

View full abstract

This paper reports the educational effects in the process of constructing “ the straw leaning tower” and “the jumping machine ”, both of which need some specialized knowledge of the mechanical engineering. Although building up the practical objects is necessary for bringing up the students’ creativity, the first year trial shows that it often fails without enough preparation and that constructing the ideas with the creative thinking way and the precise plan to realize them are indispensable. In the second and third year the practice of FD remarkably improves the both creativity of the straw leaning tower and jumping machine.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract