This paper concerns the development of a new type of force sensor called Gyroscopic Force Measuring System (simply called GFMS) for measuring a force vectorialy. The GFMS realized a highly accurate measurement using an estimator for the unknown angles of incidence of a force vector. The estimator, however, assumed an exact knowledge on the dynamical equations in the steady-state. This caused much degradation in the measurement accuracy unless the steady-state of the GFMS could be attained. The estimation errors of the angle of incidence directly affect the accurate measuring of the force. We can get the better performance by choosing the adequate damping coefficient of gyroscope (for high frequency mode). From the error analysis, some essential information for designing the GFMS is obtained.

抄録全体を表示

This paper describes a fast method for detecting the phase difference of two M-arrays by use of property of M-array that there exist only one all-zero column for even-order M-arrays. Two methods for this purpose are proposed here; one is the row coincidence method, and the other is the row correlation method. In the row coincidence method, the difference of all-zero columns of the two M-arrays is firstly detected, and the columns of one of the M-arrays are cyclically shifted to the right so as the all-zero rows of both M-arrays coincide. Then the rows of one of the M-arrays are shifted cyclically to the down side, until both M-arrays become the same. In order to reduce the effect of noise, the authors also propose the row correlation method in which correlation method is used as the means of detecting the coincidence of row in an noisy environment. Simulation results show that both the row coincidence method and the row correlation method are much faster than the conventional two dimensional correlation method. Therefore, this method would find wide applications in the signal or image processing in which a phase difference of M-array is to be calculated with less computing time and robustness to noise.

抄録全体を表示

This paper describes a new type of motion capture system for virtual reality (VR) that uses rotating magnetic field and differential magnetic field. This system can provide enough resolution, accuracy and response speed for VR with low cost compared with typical motion capture systems that uses CCD cameras, magnetic field or mechanical links. Our method uses a cubic exciting coil that consists of six square shaped coils and pickup coils that are arranged on three axes of Cartesian coordinate fixed on target. Angular posture of target is detected by measuring phase of induced signal on pickup coils that are attached on the target in rotational magnetic field. The linear position of target is detected by measuring amplitude of magnetic field at target as differential transformer. Cross-correlations between induced signals and exciting current signals are calculated to obtain phase difference and amplitude with good tolerance to noises. These three dimensional angular and linear detection are performed simultaneously by multiple frequencies method. A compensation method for angular position that uses redundancy of detection and an idea to compose each output of pickup coils are also proposed to reduce errors. Experiments were carried out to confirm effectiveness of our method. We have obtained resolution of 0.1[deg] in angular detection and 0.5[mm] in linear detection.

抄録全体を表示

In this paper we investigate parameter dependent Lyapunov functions for the linear systems with a uncertain constant real parameter in terms of geometry. First, we formulate the surface on which all parameter dependent Lyapunov matrices for a given uncertain system exist. Next, after defining a Riemannian metric on the surface, we derive a method to obtain a parameter dependent Lyapunov matrix as a geodesic on the surface.

抄録全体を表示

GPC (Generalized Predictive Control) has been formulated in state space approach as well as in conventional polynomial equation approach. The equivalence of control laws derived in two approaches has also been proven. In this paper, a new formulation of GPC control law is derived by dynamic programming, where the control law in state space approach is decomposed into low order components and expressed by recursive Riccati equation. The equivalence between the newly-derived and the original control law is proven in this paper. Futhermore, in order to guarantee the stability of GPC, a terminal penalty term is introduced to the performance index as we intoroduced to the preview servo system. The resulting stabilized GPC law becomes a generalized version of the preview servo control law.

抄録全体を表示

We study output stabilization or stability enhancement of output for a class of linear parabolic systems by means of a finite-dimensional compensator. In regular state stabilization problem, the number of sensors and actuators are chosen large enough so that they satisfy the observability and controllability conditions regarding the associated basis. We then ask the following: What control theoretic results do we have, when an enough number of sensors and actuators are not expected? Output stabilization (stability enhancement), a concept weaker than regular state stabilization, answers the question. New algebraic conditions are proposed, the feature of which is fairly different from those of state stabilization. The result is applied to a parabolic system with boundary observation/control.

抄録全体を表示

In this paper, we analyze the geometric structure of nonlinear systems with a viewpoint of the fact that the corresponding Hamilton-Jacobi (HJ) inequality has the set of solution. At first, we investigate “semi-solvable region” of the HJ inequality, which includes a true solvable region and gives us information on a part of the uncontrollable domain. Next, we introduce a non-Euclidean metric on the state space, and show that a system corresponds to a submanifold in a certain expanded manifold. We show that in the case of linear systems the submanifold is flat independent of the metric under adequate assumptions, and therefore the nonlinearity of systems can be represented by the curvature of this submanifold.

抄録全体を表示

In this paper, we consider the robust stability problem for uncertain linear time-delay systems which are given by retarded differential difference equations. By introducing an auxiliary linear system as a kind of multipliers we derive a much less conservative robust stability condition in the form of linear matrix inequalities (LMIs) compared with previously reported LMI conditions. To illustrate its less conservativeness numerical examples are shown.

抄録全体を表示

A design method of a memoryless feedback law for neutral systems is proposed. The feedback gain is determined with a solution of a matrix inequality, which can be solved via the LMI technique. The resulting closed loop system is assured to be a linear quadratic regulator for some quadratic cost functional, and in consequence, it has a robustness property against a class of static non-linear perturbation or a class of dynamic linear perturbation in the input channel. A numerical example is given to illustrate the effectiveness of the regulator.

抄録全体を表示

A method is given for calculating the extinction angle β and induced voltage e_F during open circuited period of the capacitor motor controlled by a triac. Three kinds of control circuits, (1) connected with the main phase winding, (2) with auxiliary winding, (3) with both windings, are investigated. The terminal voltage e_T during open circuited period is obtained from e_T=e_F in the circuit (1). The difference voltage V_D between two phases, in the circuit (3), is obtained, and superposed on the circuit. Values of β and e_F are calculated by a iterative solution using the symmetrical components method. Effects of the firing angle α and slip s on β are investigated on a test motor. Variations of the fundamental component voltages as a function of α, controlled by the three kinds of circuits, are investigated.

抄録全体を表示

This paper proposes a design method of frequency shaped Sliding Mode Control System using output feedback and application to follow-up control of a fluid temperature and liquid level interacting plant to a time dependent reference input. This method is characterized as that the Moore-Penrose's inverse matrix is used to deduce the equivalent control input because the numbers of inputs and outputs are not equal in general, then the order of the reduced system is equal to the number of outputs, and the chattering can be suppressed by introducing the switching hyperplane with dynamics. First, switching hyperplane is introduced as that it includes an integral element for the controlled deviation to remove the steady-state error, and a switching condition of the reachability to the switching surface is deduced using the eventual sliding mode. Next, a linear operator with dynamics is introduced to the switching hyperplane, and a frequency dependent weight matrix is introduced as that controlled deviation can be suppressed for certain frequencies. The switching hyperplane has frequency characteristics, so that the chatter motion is decreased. Finally, the frequency-shaped sliding mode system is applied to follow-up control of 2inputs-3outputs carriages, and a fluid temperature and liquid level interacting plant to confirm the effectiveness.

抄録全体を表示

This paper introduces a co-evolutionary approach to Genetic Algorithms (GAs) for exploring not only solution candidates but also their genetic expressions. Two GAs cooperate each other; one explores the arrangement of genes while the other explores the attributes of genes. The genetic expression of a solution candidate is determined by a couple of individuals from the two GAs, therefore it is not modified directly but modified through genetic operations on the individuals. This improves the intrinsic difficulty of GAs that the genetic expressions of solution candidates should be determined beforehand so that unknown better schemata will have the shorter expressions. According to the schemata theorem, the shorter expressions are rather robust to the destruction by the genetic operations and will cause an exponentially increasing number of the solution candidates containing the schemata with the short expressions in successive generations. Two typical ways of determining genetic expressions of the solution candidates from the arrangement and the attributes of genes are formalized and discussed in terms of the length of the expressions for the schemata. They are applied to the 3-bit deceptive problems introduced by D.E. Goldberg which are well-known by that the efficiency of the genetic expression can be controlled by the fitness evaluation procedure. As a result of computer simulations, the proposed method showed a better performance than the simple GA. The efficiency of the proposed method, especially from the view point of adaptive search of the genetic expression, is also discussed.

抄録全体を表示

In this paper, we propose a method of constructing a central pattern generator (CPG) model. A CPG is a network of neurons capable of generating a rhythmic output. In this method, the CPG model can generate undulating patterns. These generated patterns depend on the network topology of the CPG model. Furthermore, transitions between the different undulating patterns arise from changes of the network energy. We can consider the network energy as some driving intention of locomotion. We model CPGs as graphs, and introduce two time-evolution systems; one is a wave equatin which is Hamilton system and the other is a gradient system. The former generates intrinsic oscillating modes, and they correspond to gait patterns naturally. And the latter selects and transits in modes by bifurcating its potential functional. The superposed system of them is the objective CPG model. Some computer simulations show that the proposed CPG model with twelve oscillators (teice of six legs) can generate hexapodal gait patterns. And transitions between gait patterns are accomplished in the simulations.

抄録全体を表示

In this paper, an ontological analysis of operations and physical causal relations in artifacts is done, which is particularly beneficial for modeling interactive artifacts. Recently, human-centered design methods have been receiving attentions. We will introduce what we call “Interactive Artifacts” which are designed for promoting appropriate interactions between operators and artifacts. We have already proposed a hierarchical model for representing interactive artifacts involving “intentions”, “operations, states” and “physical phenomena” which are tightly related with each other from teleological, causal and temporal aspects. In this model, artifacts are interpreted as interfaces between humans and environments. In this sense, the proposed model is suitable for representing interactive artifacts. One of the effective processes for designing interactive artifacts is the so called “Participatory Design”. In order to apply the proposed model to the participatory design process, we will examine the concepts that are suitable for representing “physical causalities” and “operations” based on “Ontology Engineering” and will define some “conceptual classes”. In ontology engineering, some basic concepts, which are dependent on device structures, are defined. Based on these primitive concepts, we will introduce new classes for representing artifacts. In order to elucidate the effectiveness of the proposed model and classes, we will show an example of improvemental design through participatory design processes.

抄録全体を表示

The adaptability of resolution to the complexity of approximated function has a great influence on the performance of learning in the function approximation for reinforcement learning. We propose applying the reaction-diffusion equation on a graph to function approximation for reinforcement learning.The function approximator expressed by nodes can change its resolution adaptively by distributing them densely in the complex region of the state space with the proposed algorithm. A function is expressed in a plane. The successive least square method is adopted to approximate the function from the data. Each plane corresponds to a node, which is an element of the graph. Each node moves to diffuse the complexity of the approximated function in the neighborhood based on the reaction-diffusion equation. The complexity of the function is defined by the change of gradient. The simulation shows the two points: 1) The proposed algorithm provides the adaptability for function approximation. 2) The function approximation improves the efficiency of the reinforcement learning.

抄録全体を表示

This paper considers a stochastic discrete algebraic Riccati equation, which is a generalized version of the well-known standard discrete algebraic Riccati equation, and has additional linear terms. Under controllability, detectability and the assumption that the additional terms are not too large, the existence of a positive semi-definite solution is guaranteed.

抄録全体を表示

The present paper describes an efficient approach to preventively diagnose the burning conditions for a diesel engine equipped with 12 cylinders. The time series data were acquired with accelerometers which were installed on the outer surface of the engine cylinders, and analyzed by wavelet transform.The diagnosis of burning conditions of each cylinder was visually conducted on time-frequency diagrams in connection with a conventional method.

抄録全体を表示