Our purpose of this paper is to derive simultaneous state and parameter estimators for nonlinear discrete-time systems. We linearize the nonlinear systems by unscented statistical linearization (USL) and decouple unknown parameter vectors from state vectors. That is to say, we treat the unknown parameter vector as unknown input to the approximated linear models. Then, we apply unbiased minimum variance estimation (UMVE) method for the approximated linear systems with unknown input. We named this method as Approximated Unbiased Minimum Variance Filter (AUMVF). The advantage of AUMVF is that the dynamics of state estimation error is not affected by the parameter estimation error. Furthermore, we consider the linearization error due to the USL at the derivation of the AUMVF. Then, we derive the upper bound of error covariance matrix including effects of the parameter uncertainties and linearization error. Robust AUMVF (RAUMVF) is derived by using the upper bound of error covariance matrix. RAUMVF can provide more robust and accurate estimations. The validity of the proposed methods is illustrated by Monte Carlo simulations.

In this paper, disturbance estimation using parallel feed-forward compensator for non-minimum phase plant is considered. In general, it is difficult to improve the estimation performance for non-minimum phase system. In order to overcome this problem, parallel feed-forward compensator is designed to ensure minimum phase property of an augmented system, which consists of the plant and parallel feed-forward compensator, within gain of the parallel feed-forward compensator being small. A design method of the parallel feed-forward compensator using H_∞ control theory for descriptor systems is proposed in this paper. H_∞ control theory designs the parallel feed-forward compensator both to ensure the minimum phase property of the augmented system and to minimize the gain of the parallel feed-forward compensator. The verification is described that the parallel feed-forward compensator obtained by H_∞ control theory ensures the minimum phase property of the augmented system. The effectiveness of the proposed method is clarified through numerical examples.

Current robotic therapy for lower extremities is in its early stage of development. Aiming at exploring an efficacious intervention for gait rehabilitation, we investigate the gait-training device that combines saddle-seat-type body-weight support (BWS), a split-belt treadmill and functional electrical stimulation (FES). The task-oriented approach to restoring voluntary control of locomotion in patients with neuromuscular diseases focuses on three gait functions: (a) balancing, (b) hip extension, and (c) ankle push-off. The two gait functions of (a) and (b) are supported/trained by end-effector intervention combining the saddle-seat-type BWS and a split-belt treadmill, and another function of (c) is supported/trained by muscle intervention electrically stimulating ankle plantar flexion muscles. We evaluate the differences between walking with saddle-seat-type BWS and with harness-type BWS, in terms of the preferred walking speed, profiles of kinematics and ground reaction force, and the effectiveness of FES. The results indicate that the proposed gait-training device maintains subjects in a natural posture and supports important gait functions such as hip extension and ankle push-off effectively, in particular, at slow walking speed. The combined use of saddle-seat-type BWS treadmill training and FES has great potential for the next generation of neurorehabilitation technologies.

In this paper, we employ parallel computing techniques to reconstruct large-scale household compositions for micro-simulations (MS) or agent-based simulations (ABS). MS and ABS are recently employed for social simulations. For enabling MS or ABS, each household composition such as ages, occupations, or other properties of each member of a household should be prepared before simulations. However real household compositions are not available to researchers due to privacy or security reasons. Therefore reconstruction methods from statistics are developed to generate artificial populations for MS or ABS. In this paper, we propose a method to reconstruct a prefecture-level large-scale household compositions based on statistics using parallel computing techniques. In order to generate artificial populations as soon as possible, parallel computing techniques are essential in reconstruction methods. In this paper, we show a challenge in an application of parallel computing to a previously proposed reconstruction method, and show the effectiveness of the proposed method.

In this paper, we propose design policy for legged robots system software architecture. Quadruped robot systems is exemplified as the target legged robot, and its functional units are decomposed into some reusable subsystems with generic interfaces for legged robot. These subsystems are implemented as RT-Components, which are modular software components based on RT-Middleware. Finally, in order to show the validity of the proposed design policy, some experiments are conducted with two different quadruped robots and three different gait patterns.

The deteriorations of the social infrastructures such as bridges become a serious problem. Therefore, the automated inspections of the infrastructures are necessary. In this work, we developed the inspection multirotor to adhere to vertical walls based on Universal Vacuum Gripper, in short UVG. It's one of the vacuum sucker to adhere to uneven surfaces with deformable skirt. In order to optimize the adhesion force to the vertical wall, we evaluated the skirt shape of UVG such as cylinder, ellipse and inclination. As the inspection method for the multirotor, the hammering system of the light weight rod with a rolling ball was developed. We evaluated the performance of the skirt shapes of UVG for wall adhesion by measuring the sustainable weights. The developed hammering system was also evaluated by comparing the hammering results of human and multirotor.

This paper describes control design methods based on model predictive control for a system consisting of subsystems connected by time-delay elements. When the output of the upstream subsystem is measured in the middle of the time-delay element, it can be used for both feedback (FB) control of the upstream subsystem and feedforward (FF) control of the downstream subsystem. Since FB control requires offset-free characteristics and FF control requires fast transient response, a control system that has these different characteristics needs to be integrally designed. To this end, we propose two design methods. The first is to evaluate the control output with the smoothed error between the measured and predicted values. The second is to decompose the system into a FB control system and a FF control system and to evaluate the sum of performance indexes set for each system. Although the system is decomposed in terms of design, the input of each system does not affect the performance index of the other system. Consequently, the manipulated variable can be directly calculated as the sum of both controls. We also discuss a method of suppressing the violent oscillation of the manipulation reference so as to cancel the response delay of the actuators. By adding penalty terms of the reference change between the control steps and change from the previous control step to the performance index, the oscillation of the manipulation reference is reduced and the desired control response is achieved.

Based on classical density functional theory (cDFT) for colloidal systems, we attempted to derive the multiple light scattering properties of whole blood. The derived formula, OD_total=OD_id+OD_ex, for the theoretical transmittance optical density comprised the attenuation without angular scattering part, OD_id, and with angular scattering part, OD_ex. The former ideal part, OD_id, could be represented as the simple exponential decay of the scattering and absorption coefficients of whole blood with its sample thickness, and the latter excess part, OD_ex, as the conventional result of radiative transfer, such as photon diffusion theory and Monte Carlo photon propagation. The applicability of the formula were demonstrated at 800, 1300nm wavelengths using human whole blood samples with varying the blood volume fraction η from 0 to 65% (n=2). Consequently, the rms errors of two-wavelength ratio, OD₁₃₀₀/OD₈₀₀, were in the rage of 2.5-5.0 in η(%). These results suggested that the proposed cDFT method might be applicable for broader colloidal systems analysis.