In this paper, we discuss the first version dynamical modeling method for drop by using simple mathematical equations in order to create real time computer graphics in case that one operator want to communicate with the others about visual image or we construct a reference visual image for automatic control system and so on, for such kind of plant on the assumption that plant operators evaluate the process state for the plant based on the visual data inside plant through ccd camera and manipulate based on the rules constructed using operators language with experience. This model has following three blocks, first block is for the occurrence period which means that drop in formed at the end of a small pipe. Second block is for the growing period which means that drop is extended by gravity. The third block is for falling period which means that drop is falling after split. The feature for this model is that it is constructed using simple mathematical equations without repeat calculation and is able to represent several phenomenon according to different viscosity and surface tension for liquid by tuning some parameter and carry out real time visualization using computer graphics. We construct this model based on observation real phenomenon and confirmed that this model can simulate the phenomenon according to milk with 2 dimensional image by tuning some parameter for the equations within about 5% error ratio.

View full abstract

We have analyzed the interaction of the air pollution gases concentration between neighboring places in Kitakyushu area by KM₂O-Langevin equations theorem, which provides the causality between different time series. From the analyses on SO₂, NO₂ and NO gases concentrations on 100 hours basis from 1980 to 1986, we have recognized a strong causality on SO₂ in 1980 and 1986, but not noticiable in other cases. We have also recognized a moderate causality on NO₂ in 1986.

View full abstract

Properties of Hankel operator are discussed for a system with delays in state. In the finite-horizon time setting, the Hankel singular values are characterized by the roots of transcendental equation and the singular vectors are constructively given based on the solution to differential equation. Then, employing the singular vectors for the base of state-space, we elaborate a calculation method of finite-dimensional approximation.

View full abstract

This paper discusses simultaneous design of a plant and controller for command responses, using two-degree-of-freedom control systems. First, we propose a new criterion for tracking performance, which are suitable for a simultaneous design problem. Next, we give a concrete design procedure of a feedforward controller, for a fixed plant. Numerical example demonstrates the effectiveness of the proposed criterion. It also shows that feedforward and feedback properties are not independent in case of simultaneous design.

View full abstract

This paper gives a new design method for dynamical modular systems based on LMIs. The whole system consisting of dynamical modules may become unstable by connecting and disconnecting dynamical modules even though each module is stable. Therefore, we should study the stability of the whole system at connecting and disconnecting dynamical modules and investigate whether dynamical modular systems can have fault-tolerance from the viewpoint of stability. In this paper, the stability condition is provided, and it is shown that the whole system has fault-tolerance if the stability condition is satisfied. The design problem of connecting dynamical modules is formulated as a combinatorial problem subject to the stability condition, which can be described by LMIs, and the design method for optimal connections among dynamical modules is proposed.

View full abstract

It has been proved that a least mean square (LMS) adaptive algorithm is H_∞ optimal. On the other hand, the recursive least-squares (RLS) algorithm does not have robustness property. First, robustness property of the LMS algorithm is shown through a numerical example which simulates an acoustic field. Next, an RLS algorithm with regularization method is proposed in order to add robustness property to the conventional RLS algorithm. Finally, the effectiveness of the proposed method is shown through a numerical example.

View full abstract

Gain-scheduling control of a system with time-varying delay is considered. The goal of control is to stabilize the closed-loop system and achieve L₂-gain performance by swiching LTI controllers with respect to time-varying delay that is measured or estimated real time. Our approach is based on approximation of continuous timevarying delay by a discrete-time linear parameter varying (LPV) system. Using this approximation, it is possible to design a Gain-scheduling controller by solving a finite number of LMIs. The effectiveness of our method is demonstrated by simulations.

View full abstract

In this paper, we propose an LMI-Loop shaping method where simultaneous controller design for multiple specifications is possible. First, we present a design of a reduced-order controller design for an frequency weighted problem. The problem is an extention of the robust stabilization problem which has been used for an H_∞-Loop Shaping method. It can be solved with coprime factorization representation and LMI optimization. Second, we show that the design can be applied to some H_∞ 2 block problems. Since the controller parameters are embedded in the same way and they are common for all LMI's (multiple LMI's can be regarded as a single large LMI), the control design can be done simultaneously for the multiple specifications. Last, we apply the design method to a vibration control system for a boom of truck-mounted concrete pump.

View full abstract

In order to design a type 2 control system, that can remove a steady-state error (offset) when set-points or disturbances have ramp dynamics, the Model Predictive Control (MPC) algorithm was extended by introducing filters into an output prediction equation. The proposed algorithm can be applied to integrating processes as well as stable processes. The conditions that are necessary and sufficient for eliminating a steady-state error were derived. On the basis of these conditions, the prediction filter was subsequently designed. The order of the filter was found to be crucial for control performance. The results of the control simulation demonstrated successful functioning of the proposed control algorithm.

View full abstract

A hand with dexterity and local autonomy is one of the key components of a space robot to perform precise in-orbit servicing. This paper presents performance of the three-finger multisensory hand boarded on the unmanned spacevehicle "Hikoboshi". The hand comprises a newly devised three-finger module with grip force sensors and a wrist compliance device with displacement sensors. It also has a hand-eye camera and three range sensors for non-contact sensing. The contact sensors display fine performance owing to zerogravity of space compared with the ground. The laser/PSD type proximity range sensors retain performance under the high intensity of sunlight in orbit. Although the hand-eye camera is useful for position measurement of objects through image processing in most case, some points are noticed to use it in satellite daytime owing to intensity and direction variations of sunlight. The robot performed high precision tasks in space utilizing these hand mechanisms and multisensory through sensor-fused telerobotics.

View full abstract

This paper presents an application of process modeling method in order to design an operation support system for the batch plants by means of Petri nets. PVC polymerization reaction process is used as an example of modeling. Many information of process operation is required to develop the operation support system. Process modeling contributes to summarize the process information and to develop the operation support system. Petri nets are useful tools to describe and analyze concurrent processes. Three kinds of model which are operation procedure model, component behavior model and state variable model are considered to model process information. The operation procedure models are used to indicate the information of the operation progress situation of the process. The component behavior models are used to indicate the relationships between the behavior of the process components and the operation by the operator. The state variable models are used to indicate the state values in the plant. Each model is related in order to give operators the information of the process operation progress situation. And the interlock function to prevent miss-operation can be modeled by relating each model. Furthermore, the fault propagation path in the process is modeled as a fault propagation model. Normal operation support model is constructed by relating each process model. Emergent operation support model is constructed based on the fault propagation model. The modeling described here is useful to develop the operation support system.

View full abstract

Most algorithms for reinforcement learning face difficulty in achieving optimal performance when the state of the environment is not completely known. The authors have proposed a method for overcoming this problem by using recurrent neural networks in a learning agent. In this paper, we discuss the implementation of the proposed method using several types of network architecture and supervised learning algorithms. Further, the internal representation of the environment acquired in the learning agent is examined using a technique of cluster analysis. The results show that the learning agent achieves optimal performance in reinforcement learning tasks by constructing an accurate internal model, despite incomplete perception of the state of the environment.

View full abstract

We propose a metric for learning processes of agents as a useful tool in experimental analysis of multiagent systems. This method employs quantity of information for decision making which quantifies how decision making reduces uncertainty of relations between input and output. We evaluate this method in Artificial Stock Market which consists of many autonomous learning agents. The results show the proposed metric is valuable in the following two points. One is that the metric gives a criterion for determining the size of initial learning. The other is that the metric can be used to extract dominant sets of attributes in decision making. We conclude this metric will prove to be a significant tool to investigate the behavior of multiagent systems.

View full abstract

An important problem when one develops the neural networks is to determine a suitable size of a neural network. In this paper, we will consider to find the number of hidden units thoretically, since the number of units in input and output layers has not been determined so much from thoretical viewpoint. We will introduce Kolmogorov's ε-entropy and ε-capacity in neural networks, and calculate ε-entropy and ε-capacity of neural networks to approximate Lipschitz continuous functions. Then by numerical simulations we will show the effectiveness of the proposed algolithm to find the number of units.

View full abstract

In this paper, we propose an adaptive method of the Delaunay triangulation for remotely sensed images using given corresponding points. Applying Delaunay diagonal flip recursively, this method realizes adaptive triangulation process. In algorithms where the corresponding points increase dynamically, the proposed method achieves higher efficiency than the Voronoi-Delaunay division does.

View full abstract

By placing a gap around the bases of the bluff body, a linearity of a vortex shedding flowmeter has been significantly improved. Measurement error caused by an upstream pipe fitting has also been remarkably reduced.

View full abstract

In this paper, we propose a parameter adjustment law using a Riccati equation which guarantees the asymptotic stability of the dynamic error system instead of the strictly positive realness condition. Moreover, we give another Riccati equation which guarantees that both of the error vector and the parameter vector have finite powers for error equations with the system disturbance.

View full abstract