This paper discusses the effectiveness of neural networks for modeling dynamic systems as follows. The neural network consists of the input, state and output layers. The state layer comprises unit delay feedback connections in order to enable the expression of dynamic systems. In the first approach, the neural network is used to implement the external description model for a class of dynamic systems. It is shown that the neural network gives better estimation results and compares favorably with a linear regression model. In the second approach, as internal description models, linear and bilinear models are obtained by using the previously estimated results. That is, these models can be calculated by using the first and second approximations of sigmoid functions at threshold, according to Taylor's formula. Further, reduced-order realizations for each model are executed by using internally balanced realization methods, by which the singular values of reachability and observability matrices are given equally. Finally, some numerical examples are presented to illustrate the proposed methods.
A method using neural-networks to compensate for errors in the position and orientation of robot manipulators is proposed. The errors of robot manipulators due to miss in setting or modeling are measured and learned by a neural network. The reference input to robot controllers is modified by using the neural network. The compensation algorithm is discussed in the paper. A method to add the second neural network to the first one is also proposed in order to obtain the accuracy of the compensation without the much increase of the learning time in the back-propagation process. The proposed compensation method is investigated by computer simulation for a two-link and a six-link manipulator, and the method is compared with the one using linear interpolation. It is verified that the errors in the robot manipulators are significantly reduced by using the proposed compensation method. The method is also applied to a two-link SCARA type D.D. robot and its effectiveness is verified by experiments.
Identification problem of furnace heat patterns by using a neural network is described. The neural network used in this paper can identify a nonlinear system whose structure is very large and complex. For the application to practical identification problems, it is necessary to construct a neural network which is very accurate and effective for practical use. In this paper, by using a prediction error for new data which are not used for learning, a suitable neural network structure and a suitable number of learning are found out, and a neural network which is very effective for identification of furnace heat patterns is constructed. The prediction results obtained by using the neural network is compared with the results obtained by using GMDH (Group Method of Data Handling) models. It is shown that the neural network in this paper gives better prediction results as compared with GMDH models.
In this paper, an outdoor type visual guidance system by using an image recognition method is proposed for an outdoor AGV to travel stably. The outdoor type visual guidance system uses a ITV camera in order to obtain reflective images of three typed guiding marks adhered to the traveling line surface which is illuminated by a natrium light. The camera with the high speed shutter device is able to obtain a visual image in 1ms. Factors which affect detecting image characteristics are examined through the computer simulation. Traveling information can be detected within from 20ms to 50ms recognition time. In outdoor traveling experiment, the developed outdoor AGV was verified to travel smoothly on a straight line at 60m/min with as little as±25mm amplitude deviation which is less than the guiding marks width and with 10sec. period hunting motion.
In this paper, a new solution concept for linear programming problem with an interval objective function is proposed based on the worst achievement rate. First, given coefficients of the objective function, the achievement rate for each feasible solution is defined. Using the achievement rate, the interval linear programming problem is formulated as an optimization problem of the worst achivement rate among the objective coefficients within given intervals. The optimization problem of the worst achievement rate is reduced to a min-max or max-min problem subject to separate constraints. It is shown that the solution optimizing the worst achievement rate has a desirable property. A solution algorithm based on the relaxation procedure and the simplex method is proposed to solve the optimization problem of the worst achievement rate. In order to illustrate the proposed solution algorithm, a numerical example is given.