Details of a Researcher - Adachi, Shuichi

Papers - Adachi, Shuichi

Division display 21 - 40 of about 132 ／ All the affair displays >>

Identification of Decentralized System with Common Parameters

Takeuchi M., Kawaguchi T., Inoue M., Naruoka M., Adachi S.

2018 IEEE Conference on Control Technology and Applications, CCTA 2018 (2018 IEEE Conference on Control Technology and Applications, CCTA 2018) 1508 - 1513 2018.10

ISSN 9781538676981

© 2018 IEEE. This paper addresses a system identification problem for multiple isolated systems whose parameters are partially identical. An approach to the problem is a centralized identification; a single estimator collects all the input-output data from all of the systems to estimate their parameters at once. However, the approach can be practically infeasible when it is applied to a large number of systems due to the limitation of computational and communication resources. To solve the issue, we propose a method of a networked identification composed of two stages: 1) Multiple estimators collect the data from their own target systems to independently derive temporary estimates of their parameters and a covariance matrix of the estimates. 2) Then, with communicating the temporary estimates and the covariance matrix, they update their estimates. We also show the optimality of the proposed networked identification method with respect to the modeling accuracy, which is equivalently achieved by the centralized identification. Finally, we show the effectiveness of the proposed method in a numerical simulation.

Access to Document (DOI)

Aperiodic Sampled-Data Control via Explicit Transmission Mapping: A Set Invariance Approach

Kazumune Hashimoto, Shuichi Adachi, and Dimos V. Dimarogonas

IEEE Transactions on Automatic Control (IEEE Transactions on Automatic Control) 63 ( 10 ) 3523 - 3530 2018.10

Research paper (scientific journal), Joint Work, Accepted, ISSN 00189286

© 1963-2012 IEEE. Event-triggered and self-triggered control have been proposed in recent years as promising control strategies to reduce communication resources in networked control systems (NCSs). Based on the notion of set-invariance theory, this note presents new self-triggered control strategies for linear discrete-time systems subject to input and state constraints. The proposed schemes not only achieve communication reduction for NCSs, but also ensure both asymptotic stability of the origin and constraint satisfactions. A numerical simulation example validates the effectiveness of the proposed approaches.

Access to Document (DOI)

Model Matching Control for Binaural Reproduction Over Loudspeakers

Shohei Mori, Masaki Inoue, Kentaro Matsui, and Shuichi Adachi

SICE Journal of Control, Measurement, and System Integration 11 ( 3 ) 249 - 255 2018.05

Research paper (scientific journal), Joint Work, Accepted

Robust safety controller synthesis using tubes

Hashimoto K., Adachi S., Dimarogonas D.

2017 IEEE 56th Annual Conference on Decision and Control, CDC 2017 (2017 IEEE 56th Annual Conference on Decision and Control, CDC 2017) 2018-January 535 - 541 2018.01

ISSN 9781509028733

© 2017 IEEE. In this paper, we investigate the reachability analysis and safety controller synthesis problem for linear discrete-Time systems under additive bounded disturbances. We consider the following problem; design a state feedback controller such that any state trajectories starting from an initial set can be robustly controlled towards a target one in finite time, while at the same time avoiding any prohibited regions. One of the potential disadvantages of existing reachability algorithms when external disturbances are taken into account, may be that the solution to guarantee reachability becomes conservative. Motivated by this, this paper provides a new controller synthesis framework based on the notion of tube-based control strategy, in which a suitable sequence of polytopes is generated according to a convex feasibility problem. An illustrative simulation validates the effectiveness of our proposed method.

Access to Document (DOI)

Difference-based Identification Method for Systems with Rate-independent Hysteresis and Its Application to Rechargeable Battery

Kawaguchi T, Maruta I, Osamura K, Adachi S

IFAC-PapersOnLine (IFAC-PapersOnLine) 51 ( 15 ) 150 - 155 2018.01

© 2018 In this paper, we consider identification of a linear system connected in parallel with the rate-independent hysteresis. Since the parallel system provides only the combined output of the two systems, it is required to extract the signal related to the target system from the observed data to identify the systems separately. In order to achieve this, we propose a system identification method, in which the identification experiment is performed twice with appropriately designed input signals. The inputs make the responses of the hysteresis identical in the two experiments, therefore the difference of the two observed outputs is related to only the target linear system. Then, the linear system is identified using the difference. The proposed method is applied to a rechargeable battery, whose output originates from the nonlinear and the linear elements: open circuit voltage with hysteresis and overpotential. The effectiveness of the proposed method is illustrated through a numerical simulation of the rechargeable battery.

Access to Document (DOI)

Recursive Identification of Fractional Time-Delay Systems Using Discrete-Time Model

Higo T, Kawaguchi T, Itamoto M, Adachi S

IFAC-PapersOnLine (IFAC-PapersOnLine) 51 ( 15 ) 939 - 944 2018.01

© 2018 In this paper, a recursive identification method for fractional time-delay systems using a discrete-time model is proposed. The fractional time delay is a non-integer multiple of the sampling period for the discrete-time system. We describe the discretization of the fractional time-delay systems and propose a new formulation for system identification. Single-input, single-output systems are transformed to two-input, single-output systems: one transfer function represents system dynamics and the other represents the influence of the time delay. Then, we propose a new recursive identification method for fractional time-delay systems. Finally we verify the effectiveness of the proposed method through numerical examples.

Access to Document (DOI)

SigsofTmax: Reanalysis of the softmax bottleneck

Kanai S, Yamanaka Y, Fujiwara Y, Adachi S

Advances in Neural Information Processing Systems (Advances in Neural Information Processing Systems) 2018-December 286 - 296 2018

ISSN 10495258

© 2018 Curran Associates Inc..All rights reserved. Softmax is an output activation function for modeling categorical probability distributions in many applications of deep learning. However, a recent study revealed that softmax can be a bottleneck of representational capacity of neural networks in language modeling (the softmax bottleneck). In this paper, we propose an output activation function for breaking the softmax bottleneck without additional parameters. We re-analyze the softmax bottleneck from the perspective of the output set of log-softmax and identify the cause of the softmax bottleneck. On the basis of this analysis, we propose sigsoftmax, which is composed of a multiplication of an exponential function and sigmoid function. Sigsoftmax can break the softmax bottleneck. The experiments on language modeling demonstrate that sigsoftmax and mixture of sigsoftmax outperform softmax and mixture of softmax, respectively.

Design and implementation of a binaural reproduction controller applying output tracking control

Ito A, Matsui K, Ono K, Hattori H, Suenaga T, Iwauchi K, Adachi S

145th Audio Engineering Society International Convention, AES 2018 (145th Audio Engineering Society International Convention, AES 2018) 2018

© 2018 KASHYAP. We have been studying a design method for a controller for binaural reproduction with loudspeakers. The gain of the controller amplifies errors due to external disturbances and system perturbations, and this leads to deterioration of the sound quality. Therefore, the gain should be suppressed to as low as possible. For this purpose, we formulate the design of the controller as a minimization problem of the gain, in which the H∞ norm of the controller is adopted as a measure of the gain. In this article, we also introduce a binaural reproduction system as an implementation example. This system virtually reproduces multichannel audio such as 22.2 multichannel audio using line array loudspeakers.

ロバストゲインスケジュールオブザーバによるヒステリシス特性をもつ二次電池の充電率推定

ADACHI Shuichi

計測自動制御学会論文集 Vol.53 ( No.11 ) 574 - 582 2017.11