Fault detection and diagnosis (FDD) of air-conditioning systems is very important for energy saving, safety and maintenance and in high demand. The air-conditioning system is a complex and dynamic system which makes the diagnosis difficult because the observed parameters do not directly express the state of the system. On the other hand, faults analysis and fault cases collected from the actual air-conditioning system are not practicable. With due consideration of the above points, the authors propose an effective FDD method for air-conditioning systems using model simulation and Recurrent Neural Network (RNN). An air-conditioning system of a certain building is used as a model in this research. The validity of the model is confirmed by comparing the simulation data with real observations. In this study, the authors concentrated on several typical faults in both chilled and hot water volume control valves. The authors obtained the data of the normal and some fault cases in time sequence using the model simulation. As a result of the current experiments, the recognition rate was over 90% and it was demonstrated that the proposed method is suitable for practical use.

抄録全体を表示

Pulse Radar system and FMCW radar system have already been practically used in the society as the distance measurement systems utilizing radar beam while the industrial use of them is often limited by the conditions of the Radio Transmission Regulation that stipulates mainly the allowable range of the beam intensity and the sweep band in each country. In this paper, a method to measure a short distance even in the narrow-band width is proposed and it utilizes the circuit to mix the wave reflected by the object with a transmitted wave. And also an in-phase signal and a quadrature signal of those mixed waves are utilized. As a technical reference to support the availability of the method, a case study with RF circuit model using a diode model is presented here with showing the possible achievement of measurement error ±2mm or less in a short distance 2m or less in the simulated bandwidth of 76MHz in 24.15 GHz band.

抄録全体を表示

This paper introduces a multirate sampling visual feedback control technique on an active vision system using parallel mechanism. Active vision systems are mechanical systems that consist of multidegrees of freedom machines and vision sensors, and the systems perform track target points. The proposed system is based on a six degrees of freedom parallel mechanism and a vision sensor which is set on the motion base plate of the mechanism. Even if the target is a point on 3D space, the point is expressed as a 2D point on the screen of the vision system. Then, the active vision system has four degrees of redundancy. Making use of zero space ralated to the redundancy, the proposed system has the tracking functions realized by basic, translation, and orientation control modes. On the other hand, there is a problem that control performance of the system is not high because of long operating time of the vision sensor relative to sampling ratio of the motor control. In order to improve the control performance, this paper introduces an image based visual feedback controller including multirate sampling disturbance observer. Finally, experimental results are shown for evaluation.

抄録全体を表示

This paper proposes a new method of getting Lyapunov functions using randomization, state discretization, and quantization of the Markov process. We deal with continuous deterministic systems as discrete stochastic systems and quantize the systems. The randomization of deterministic systems is introduced because we are aiming at the extension of the method to the stable stochastic case in future works. This procedure introduces Schrödinger-like equations. The problem of obtaining Lyapunov functions is replaced by the problem of solving eigen equations of the Schrödinger-like equations. In previous works of obtaining Lyapunov functions, for example, Fleming and Soner's method, there is the case that numerical approximated values diverge because of iteration of calculation. However, our method does not cause the problem because it is based on eigen analysis.

抄録全体を表示

This paper presents a new branch-and-bound (B & B) method to reduce the computational burden of online optimization in multi-vehicle formation control with collision avoidance that is formulated by model predictive control. One of the possible reasons for extremely heavy computation time of the standard B & B algorithm is that so many subproblems are generated since integer constraints are rarely satisfied in “relaxed” problems. The purpose of the proposed method is to decrease subproblems based on a new branching rule taking into account the inherent properties of collision avoidance problems. Numerical examples and experiments show that the proposed method drastically reduces computation time of online optimization for multi-vehicle formation control.

抄録全体を表示

This paper proposes a weighted balanced realization and model order reduction for nonlinear systems. This is an extension of a nonlinear balanced realization, which is based on singular value analysis of the nonlinear Hankel operator, to include input and output weights. The proposed method is a natural nonlinear extension of weighted balanced realization for linear systems. Furthermore, it is proved that a reduced order model derived by the proposed procedure is also stable if the original model is so. Finally, the effectiveness of the method is shown by a numerical simulation.

抄録全体を表示

This paper addresses an optimal design problem of dynamic quantizers for a class of 2D systems with discrete-valued input. First, we derive a closed form expression of the performance of a class of dynamic quantizers. Next, based on this, an optimal dynamic quantizer and its performance are provided. Finally, we apply the optimal dynamic quantizer to generate binary halftone images.

抄録全体を表示

In particle swarm optimization (PSO), each particle searches for an optimal solution in such a way that its candidate solution becomes closer to the global best and its personal best. All the particles in the swarm tend to search in the neighborhood of the global best, which makes it difficult to apply PSO to optimization of multimodal objective functions. In order to search widely over the solution space, this paper presents PSO with enhanced autonomous search ability of each particle (PSO-ASA). PSO-ASA is based on the simple idea that the personal best of each particle is not updated if its current candidate solution is close to the global best. Furthermore, we propose a hybrid method combining PSO-ASA and the original PSO in order to be able to treat objective functions with various landscape features.

抄録全体を表示

The estimate accuracy of state transition probability effects an obtained policy when reinforcement learning method is applied. Therefore, it is important to know how accurately we must estimate the probability. The objective of this study is to derive a sampling condition to guarantee an optimal policy with a desired reliability. First, we describe a relation of Q-factors and estimated probabilities to induce the right optimal policy. The relation provides a required accuracy for the probability estimation and leads to sampling conditions. Additionally, we propose the method to predict the required number of sample in advance. Numerical simulations examine the effectiveness of the proposed methods.

抄録全体を表示

In this paper we propose an approach to use tactile information in a robot's whole arm dynamic manipulation. Unlike those researches that only use tactile sensors to recognize objects or contact conditions, this approach considers the relation between a robot's motion trajectory and the tactile information for the succeeded tasks cases, and adjusts the robot's motion spatially and temporally so as to realize the required dynamic manipulations. Experimental results show the effectiveness of our approach. Our human interactive robot named RI-MAN can hold up a dummy weighing up tp18kg by this algorithm using real-time whole body tactile feedback.

抄録全体を表示

We introduce the concept of generative pointer, which is an extended subobject classifier with local identity (i.e. an incomplete indication or identification). Generative pointer is derived from a generalization of the emergence process of rational numbers. It metamorphoses hierarchical structure in dynamical systems into heterarchical one. Generally, intermittency arises as a critical state on a dynamical system. Thus it is sensitive with respect to parameter shifts. In contrast, on-off intermittency is ubiquitously observed for a wide range of parameter values when a generative pointer is applied to a Henon map system. This fact implies robustness of the criticality against parameter shifts. The concept of structural stability or trajectory stability of dynamical systems is based on the specific invariable map, but robustness, in contrast with stability, is based on the dynamic change of the map. Thus, in our view, the robustness does not conflict with the emergence but coincides with it.

抄録全体を表示

This paper presents a novel flocking strategy for a swarm of robots that enables the robots to autonomously navigate in an environment populated with obstacles. Robot swarms are often required to move toward an assigned goal while adapting to environmental changes in many applications. Specifically, we pay attention to the swimming behavior of a school of fishes and implement three basic behavior models such as maintenance, partition, and unification subject to the environment condition. The proposed approach is based on the local interaction among individual robots, which requires one robot to dynamically select two neighbors within its sensing range and maintain a uniform distance with each other. To provide full generality, all robots are not allowed to have individual identification numbers, a pre-determined leader, the memory of past actions, and the communication capability. We verify the validity of the proposed algorithm using our in-house simulator. It is demonstrated that a large swarm of robots can be split into multiple groups and/or merged into one passing through multiple narrow passageways.

抄録全体を表示