In this paper, we do an experimental study of two-degree-of-freedom optimal type-1 servo system based on the LQ (linear quadratic) method using a tank system. First, the experiments for the state feedback system which takes into account of a trade-off between low frequency sensitivity and guaranteed stability margin by using an integral gain are carried out. Next, we extend to the output feedback case and carry out the experiments to show the effectiveness of the LTR (loop transfer recovery) method or the integral gain. As the results of these experiments and its analysis in frequency domain, the influence of an observation noise which appears in the control input, the effect of disturbance rejection and robust stability for a plant perturbation are clarified.
We have developed a neural network (NN) recognition and its hardware for paper currency. We have proposed mask concept to extract characteristics of the currency. Here, we propose a unique mask which has a symmetrical masked area against axes which divide a long side and short one of the currency. We can obtain the same value from both upright image and inverse one of the currency using the axis-symmetrical mask. This means that these values are invariant to upright and inverse of the currency conveyance. Namely, we can recognize twice number of currency kinds compared with the conventional mask though the NN construction is the same. Furthermore, we optimized the axis-symmetrical mask by genetic algorithm. Finally, we show the effectiveness of the optimization by statistical analysis of mask processed values.
In this paper, we consider a suboptimal control problem based on the input-output data through the LFT representation. First we describe the suboptimal control problem in a general setting via the LFT. Then we solve the problem via numerical optimization, where the relation between the design parameters and the cost function plays a key role. For this equation, we discuss how to solve the optimization problem. Finally, we give numerical examples of controller design to demonstrate its effectiveness.
This paper deals with a large-scale scheduling problem for a metal mold assembly process of automatic vending machines, which is based on a job shop process with additional conditions such as precedence constraints among jobs, parallel machines, and introduction of multi-function machines. The objective of the problem is to minimize the sum of the tardiness of each product. The genetic algorithm (GA) is applied to this problem. Particularly, a method for generating initial individuals and a mutation procedure are proposed. It is observed through simulations for a real data that GA is more effective than a heuristic method.
In this paper, an automatic counting method of the passing people through the gate is proposed. In the proposed method, two measurement lines are set on the floor of the inside and outside of the gate at right angle with the moving directions of the passing people. The images of the passing people are obtained in series and two space-time images are generated at each measurement line. By detecting which line the person passes early, the moving directions of the passing people are recognized. Finally, the incoming persons and the outgoing persons could be measured by counting the people data with each direction information. In this paper, we describe the algorithm of counting the passing people using the two space-time images and show some experimental results obtained by using a simple experimental system to verify effectiveness of the proposed method.
This paper considers a synthesis problem for such observer based controllers that are strictly positive real and H2 optimal at the same time. We consider this problem in particular from the viewpoint of the inverse problem, and provide a parametrization of the controller, which shows an interesting coupling structure between the regulator and the observer gains. Among them, we concentrate on an weakly coupled class in this paper to have the LTR structure, which was not provided by the original result. For the class, we give conditions on design parameters for ensuring both properties, which can be independently calculated although the problem is the mixed one.
A method for extracting fundamental frequencies from a mixed tone perfomed by natural instruments is proposed. Acoustical features are assumed in this study : (1) Low-order harmonics have greater powers than other partials, (2) powers of harmonics are continuous in time, and (3) the power spectrum for a mixed tone can be approximated by summing the power spectra of each tone. The “fundamental-ness” of the partials is introduced by these features. Duet and trio tones performed by natural instruments are used as test tones. The averages of the correct identification rates evaluated by performed musical notes are 84%-96% for duet tones, 80%-87% for trio tones.