In order to reduce the computation amount of FIR filters, we consider design sparse FIR filters by using combinatorial optimization. We also consider solve this optimization problem by using branch and bound method. The computation time for this optimization depends on the initial reference value of branch and bound method. This paper proposes the method to estimate the initial cost value to improve the efficiency for the sparse FIR filter design.
IoT (Internet of Things) platforms have been drastically changing in the recent years. And a competitive axis of controller business has been also changing while edge computing attracts attention in the field of FA (Factory Automation) in particular. Control technology on the IoT platform seems to need industrial applicability evaluation from the view point of structured approach and unstructured approach. In this paper, consideration on industrial applicability evaluation of the temperature control system is discussed.
A database-driven PID (DD-PID) control method is one of the effective control methods for nonlinear systems. This scheme requires a database to tune control parameters adaptively. In the conventional DD-PID control method, there is a problem that the calculation cost and required memory for creating an optimal database are large, and it is difficult to implement the controller into small-seized embedded computers. This paper proposes a method to implement DD-PID controller using small-sized sub-databases based on the self-organazation map (SOM). In the proposed method, sub-databases are constructed by clustering a large scale database using the SOM. Moreover, it is possible to realize the reduction of the required memory and the calculation cost compared to the conventional method by selecting sub-databases adaptively. A numerical simulation example evaluates the effectiveness of the proposed scheme.
Systems Engineering (SE) and Model-based Development (MBD) attract attention from manufacturers to produce more complex products with less rework in order to satisfy higher demands. However, concrete methodology is not defined well. This paper proposes a methodology of target assignment with respect to a special functional cascading for integrated atchitectural products. This paper also proposes “abstracted energy model” for assignment in top layer in the cascade. An example for air conditioner development demonstrates the efficiency of this methodology.
In this paper, a distributed event-triggered formation control algorithm for a multi-agent system which consists of linear discrete-time agents is proposed. For agents to achieve a formation, it is necessary to communicate with each other to feedback their information. For communication, wireless communications are desirable for various tasks since wired networks can constrain their movement. Also, each agent have to observe their states from sensors. Those energy consumption due to unneecessary calculation and communication of agents may shorten battery life of agents. By using the event-triggered protocol, which updates the control input aperiodically only when certain triggering condition is satisfied, we aim to reduce those calculation and communication by reducing the input update frequency. Also the multi-agent system is guaranteed to achieve a formation by determining the triggering condition of the control input based on Lyapunov's stability theorem. At the end of the paper, the effectiveness of the proposed algorithm is verified by numerical simulation.
In the database-driven control scheme, the fixed number of neighbors' data is selected to calculate controller parameters. Therefore, the inappropriate neighbors' data are occasionally chosen when query is not included in the database. In this study, the database-driven control scheme using a kernel density estimation is newly proposed. The kernel density estimation can calculate the similarity between the query and database. According to the proposed scheme, it is possible to calculate the degree of abnormality based on the aforementioned similarity. The effectiveness of the control scheme is numerically evaluated by a numerical example.
This paper proposes the opposite reference filter to realize the desired transient property in load control of hydraulic cylinder for Tunnel Boring Machine. In the case of load increase at the start of excavation, it is desirable to reduce load transiently in order to suppress overshoot, and to reduce load quickly in case of emergency. The reference filter is designed to have opposite characteristics when the reference signal is increased and decreased. Finally, the force control experiment of TBM hydraulic cylinder is performed to verify the usefulness of the proposed method.
Several researches have formulated economic performance optimization problems taking account of a trade-off relation between process input and output variance. The present work introduces theoretical formulations that relate the weighting parameter with each input and output variance on the condition that the Linear Quadratic Gaussian (LQG) controllers are implemented as lower layer controllers. The proposed approach is applied to a two-input, two-output separation process model, and solves optimal weighting parameter based on LQG control. The numerical example also shows that the obtained optimal weighting parameter is effective for a MPC based learning algorithm.
Drowsiness can reduce working efficiency and result in dozing while driving. It is known that short naps are effective for eliminating drowsiness, so further analysis of short naps is required to clarify the relationship between nap's length and quality. EEG in short naps is characterized by the appearance of a sleep spindle and detection of this sleep spindle is thus required for analysis of this sleep stage. In this paper, we propose a method to detect sleep spindles in the time domain using a Long Short-Term Memory (LSTM) network.
This study proposes a new design method for the mobile robot trajectory generation considering obstacle avoidance using mixed integer linear programming (MILP). In the proposed method, the velocity of a mobile robot as well as the position of the mobile robot are used in the initial condition of MILP. Numerical examples show that the mobile robot collides with obstacles using a conventional method, in which the velocity is not taken into account, and on the other hand, the problem is resolved using the proposed method.
The grid connection inverter is built in the PCS (Power Conditioning Subsystem) for the distributed power supply, such as photo voltaic system. Recently, there has been considerable advances in the research on power flow using PCS. In order to adjust the power in the PCS, it is necessary that the grid connection inverter of the PCS predicts the load and output fluctuation in the customer side. Therefore, this paper investigates a load estimation in the grid connection inverter of the PCS. The proposed estimation method adds a superimposing pseudo random binary sequence to the inverter reference. The validity of the proposed estimation method is verified experimentally. As a result, it is experimentally clarified that the estimation is possible even for inductive loads.
Generalized Predictive Control (GPC) is one of the model-based control methods. The control law is derived through the performance index calculated by sum of squares about the error between reference signal and output prediction and the control input. Although coprime factorization approach has been used in order to extend the control law in the previous researches, there has been a possibility that the order of the derived control law becomes high. Therefore, this paper extends GPC through newly-defined output prediction and proposes the method to re-design the control law or the characteristic from noise to output with keeping the closed-loop transfer function.
This paper shows that the plant impulse response can be estimated by the inverse discrete Fourier transform of the estimated plant frequency response that is the ratio of the discrete Fourier transforms of the prefiltered output and input of the plant, and it corresponds well with the estimated plant impulse response that is the plant output filtered by the inverse filter of the plant input.
Microwave power dividers play important roles in microwave circuit systems. As an in-phase divider, the Wilkinson power divider which consists of two quarter-wavelength transmission lines and a resistor connected between two output ports is well known. However, the circuit size becomes large because this divider is designed based on a distributed circuit theory. This problem is extremely serious at low frequency regions such as VHF and UHF bands. Although some power dividers utilizing Π- or T-networks have been already reported in order to overcome this problem, these dividers have relatively narrow bandwidths. On the other hand, authors have been also proposed broadband lumped-element power dividers utilizing LC-ladder circuits.
This paper describes a design method of multi-section LC-ladder dividers at VHF band. The multi-section LC-ladder divider is a broadband lumped-element power divider composed of multi-section LC-ladder circuits and an isolation resistor. By designing the divider based on multi-section impedance transformer and L-type matching techniques, broadband characteristics of a relative bandwidth of over 100% can be obtained at VHF band. Electromagnetic simulation and experiment for the designed divider have been performed in order to verify the design procedure. The measured results of the fabricated LC-ladder divider at a center frequency of 178MHz utilizing commercial chip elements are good agreement with simulation results.
A novel frequency-tunable high power resonator is proposed for use in the feedback-path of the feedback-type high power oscillator to realize a frequency-tunable performance. With the combination of a fixed-frequency high power resonator and a multi-section high breakdown-voltage varactor diode, a 2.385 to 2.456GHz GaN-HFET VCO having an output power of 20W and a drain efficiency of 55% has been successfully achieved for use in the microwave oven.
Internet of Things (IoT) data processing systems must handle massive and many kinds of data. Hence, it is important for designing IoT data processing systems to evaluate a performance of GI/G/s/s typed systems. The exact solutions of GI/G/s/s typed systems have not been yet developed. Alternatively, we apply a discrete simulation method to evaluate the systems. However, the method spends much time to evaluate the performance with any conditions. In our previous study, we have evaluated a performance of GI/G/s typed systems with infinite capacity by using machine learning. However, we have not evaluated a performance of loss traffic systems (GI/G/s/s typed systems with finite capacity). In this paper, we evaluate a performance of the systems by using machine learning and validate what kind of training data should we use.
This paper proposes a joint vector model and a motion check rule format for an exercise guidance system to help beginners to learn exercise by themselves. Our joint vector model can describe the difference between correct/incorrect motions by the angles and the distances between vectors from joints to joints. Our format can describe planning knowledge that consists of precondition, time and actions for distinguishing correct/incorrect motions based on the joint vector model, so that it can generate guidance sentences depending on the types of error motions. We evaluate the description ability of the joint vector model by describing the check points in a text book concerning the Japanese popular radio exercise No. 1. Furthermore, we evaluate the usability of guidance, the times required for creating motion check rules, and the execution times by implementing an exercise training support system.
In the field of rehabilitation, Stroke Impairment Assessment Set (SIAS) is used to evaluate the body dysfunctions for the post-stroke that is the disease of the brain. We have been developing automatic evaluation systems for the motor dysfunction in SIAS by using the detected joints of the body by using depth sensor. However, the joint detection is a specific function to Kinect and is not always possible for other depth sensors. Moreover, even when using Kinect, the positions of the joints are not correctly estimated when observed from the backward or with irregular poses. Thus, it is often inevitable to analyze the body poses from depth data directly. In this study, we directly apply depth data to the trunk function tests and joint movements of SIAS. The measuring method of angles from the raw depth data will be proposed, and the results are compared to the traditional test method of SIAS.
It has been shown that it is possible to construct an ideal ANC system in which the noise source signal is a white signal by applying a dither signal having a power spectrum that is the inverse of the noise source signal. The noise source signal is a random signal that has reached to the microphone via the plural number of long distance acoustic paths. Therefore, the deviation of the frequency characteristic of the power spectrum is considered to be large. The nature of the system that the noise source signal is equivalently replaced by a white signal is extremely significance. The main aim of the paper is to verify the basic principle of this high speed ANC system to the actual noise. A method of generating an external dither signal with a power spectrum that is in reciprocal relation to the power spectrum of the noise source signal is discussed. Also, a practical normalized adaptive algorithm and a range of stable step size are derived. Finally, numerical verifications with actual noise source signal is performed by computer simulation, and the features and effectiveness of the proposed system are summarized.
A method for the estimation of wave-making resistance from the hull form and Froude number through deep learning is proposed. At the same time, this research also gives a solution when the data are skewed, which solves the problem of low generalization performance. The reduction of wave-making resistance is an essential issue in hull form design. However, the estimation of wave-making resistance is a time-consuming task that depends on experimental measurements. To enable direct estimation of the wave resistance from hull form, deep learning, which enables end-to-end learning, is an effective approach. The proposed method has two phases. First, auto-encoders, which reduce the dimension of the offset and the profile data, are generated, while performing to the skewed offset data, use an improved sampling method. Subsequently, after the regularization of these data, a deep neural net for regression estimation of wave-making resistance is generated. The results of evaluation experiments show that the proposed method can estimate wave-making resistance with high precision.
The demand for navigation is rising as railway transportation services is becoming complex. Real-time data integration from multiple business systems is required for passenger-oriented navigation. Conventional mediators, real-time data integration technique which converts queries and data between business systems and applications based on their each schemes and mapping between them, can not follow expansion both of applications and business systems. In this paper, we propose a new mediator to solve the scalability problem with core-global scheme, which allows independednt mappings with each schemes of them and expresses principal components of business systems and applications. The 2 features of core-global scheme reduce effects of scheme changes on other mappings so that the mediator has scalability. Furthermore, we propose an architecture of the mediator that reduces effectivity loss of queries. We implemented a prototype and evaluated its scalability and query effectivity for an application of railway navigation.
A large number of traffic fatalities are caused by falling asleep at the wheel. Several drowsiness detection technologies have been developed in recent years. A previous study describes how hemodynamics can vary significantly due to drowsiness. However, it was difficult to estimate drowsiness from the time series of hemodynamics. In this study, general models for estimating three drowsiness levels (i.e., high, medium, and low) based on hemodynamics were constructed using a convolutional neural network for detecting the condition before a state of complete sleepiness is reached, the goal being traffic accident prevention. The results showed that the accuracy of the model was 68.9%.