|
Nonlinear Stochastic Time-varying System Identification Based on Multi-dimensional Taylor Network with Optimal Structure |
|
Chao Zhang, Hong-Sen Yan(Southeast University, China) |
| A novel method for nonlinear stochastic time-varying systems identification based on MTN with optimal structure is proposed. In this paper, the connection weight coefficients of MTN are regarded as time-varying parameters, which are trained by the VFF-RLS algorithm, to reflect the input-output change. Moreover, to avoid the dimension explosion, the weight-elimination algorithm is introduced to choose effective regression items of MTN, thereby the simplest structure of network which has the best generalization ability is obtained. Results show that the method proposed in this paper is valid. |
|
|
A preliminary result on system identification of a floating offshore wind turbine |
|
Yuki Noma, Naoyuki Hara, Keiji Konishi(Osaka Prefecture University, Japan) |
| In this paper, we build a mathematical model of a floating offshore wind turbine using system identification. Input and output measurement signals of the wind turbine are used to obtain a linear state-space model. The model is validated through a nonlinear simulation using a high-fidelity aeroelastic simulator ``FAST." |
|
|
Evaluation of Steering Model depending on Gazing Distance by using Driving Simulator |
|
Daisuke Matsuno, Nobutomo Matsunaga, Yuki Shida, Hiroshi Okajima(Kumamoto University, Japan) |
| To analyze the driving behavior is important for providing comfortable driving assistance. The novelty of the paper is to give a modeling concept of a steering model depending on gazing distance. The steering model consists of measurement of gazing distance in real-time, and optimal estimation of PID gain by Particle Swarm Optimization. The effectiveness of proposed method was evaluated by using Honda driving simulator. Driving experiments experiment by isolated and following driving. From experiment, proposed model depending on gazing distance was more precise than conventional method.
|
|
|
Estimation Problem for Network Model at State and Measurements Attacks and Information Cost Criterion |
|
Vasyl Martsenyuk, Mikołaj Karpiński(University of Bielsko-Biala, Poland), Bakhytzhan Akhmetov, Nazym Zhumangalieva(Kazakh National Research Technical University after K.I. Satpayev, Kazakhstan), Iryna Gvozdetska(Ternopil State Medical University, Ukraine) |
| The purpose of this research is to offer constructive algorithm for estimator search in one network model under state and measurements attacks.
The model is nonstationary descriptor system including difference equations for node state variables and algebraic equations for measurements. State variables and measurements are considered as random vectors.
We use information cost criterion in order to find optimal estimator of inner product.
The main algorithm is based on reducing estimation problem to a problem of optimal control. Mean-squared error independent on attacks is obtained |
|
|
Road Roughness Modelling by Using Spectral Factorization Methods |
|
Semiha Turkay(Anadolu University, Turkey) |
| Road roughness model for the right and left tracks are constructed by two methods.In the first method,from time-domain measurements obtained from LTPP archives, the empirical auto and cross-power spectral densities are estimated by using the Welch method.The road roughness is decomposed into three stochastically independent random processes and the linear-shape filter parameters of the right and the left tracks are estimated via a subspace-based identification algorithm.In the second method,the subspace-based identification algorithm is directly used, without employing spectral decomposition |
| |