GPS/GNSS (Global Positioning System/Global Navigation Satellite System) is a 3D positioning system using space satellites for measuring a receiver's current position. Recently, many people use GPS as the navigation system in car and cellular phone, so the positioning accuracy of several meters is required to satisfy the user's need. However the measurement error reaches hundreds of meters in urban areas. One of the reasons is that the receiver fails to measure pseudo range accurately due to multipath from the buildings and so on. The other reason is that the satellite constellation is biased because of the decreasing number of observable satellites. Therefore, we proposed methods for reducing the multipath error and the lack of visible satellites. At the present day, although multipath error is reduced by the choke ring antenna and the correlators, this method has a problem that the antenna is expensive, big or complex. We devise methods to reduce the multipath error by only using measurement data. By these methods, we can reduce the size of the receiver and to use the satellite that contains the multipath error for the measurement. We achieved the improvement from 35.3m to 30.5m in 2drms by this method. We achieved about 69% improvement in 2drms and about 5% increase in measurement rate. We can describe that we succeeded not only in improving the measurement accuracy but also in increasing the measurement rate in urban area. The results show that our proposed method is effective in urban areas measurement.

抄録全体を表示

The problem of constructing model reference adaptive H_∞ control for distributed parameter systems of hyperbolic type preceded by unknown input nonlinearity such as dead zone or backlash, is considered in this paper. Distributed parameter systems are infinite dimensional processes, but the proposed control scheme is constructed from finite dimensional controllers. An adaptive inverse model is introduced to estimate and compensate the input nonlinearity. The stabilizing control signal is added to regulate the effect of spill-over terms, and it is derived as a solution of certain H_∞ control problem where the residual part of the inverse model and the spill-over term are considered as external disturbances to the process.

抄録全体を表示

With the rapid development of robotic hardware and software platform, robots are more easily to provide services at home or office, and it is necessary to instruct robots to perform tasks on the spot. This paper presents a user interface for manipulating structured furniture and electric equipments. The contribution of this papers is summarized as follows. 1) We propose a minimally invasive system in which users can instruct robots to operate variable structure intuitively. 2) In order to obtain a 3D point that corresponds to the user-touched point on the interface, we show a method to estimate 3D points from screen points of the robot's view images. 3) We propose a system for robots to learn tasks by memorizing operation target.

抄録全体を表示

In this paper, a new learning scheme named Virtual Feedback Error Learning is proposed to improve conventional FEL schemes. Conventional schemes for FEL have a drawback that a time variant block is inserted into the closed loop system as a side effect from eliminating an assumption of positive realness for stability. This paper separates the inserted block from the closed loop by means of cascade connection of an inverse system. As a result, designers do not need to pay attention to the effect of the block to stability. The paper shows the parameters converge to true values under a PE condition, and verifies its effectiveness based on simulation.

抄録全体を表示

This paper proposes a parameter estimation method for state-space models based on the variational Bayes method. The proposed algorithm employs the prior and posterior distributions in the same form. This allows us to use it iteratively to obtain accurate estimation whereas the existing algorithms cannot be used iteratively. Furthermore a numerical example demonstrates the effectiveness of the proposed method.

抄録全体を表示

This paper considers finite-time position control of an one-link mechanical system. The system is modeled by discontinuous differential equations. In this paper, we prove that the Nakamura's local homogeneous controller based on a control Lyapunov function is valid to the position control of the robot manipulators, and show the effectiveness of the controller by experiments.

抄録全体を表示

Up to the present time, estimation of Beef Marbling Standard (BMS) number based on ultrasound echo imaging of live beef cattle has been studied. However, previous attempts to establish the objective and high accurate estimation method have not been satisfactory. Our previous work showed that estimation of BMS number was achieved by neural network modeling with non-linear mapping ablity. This paper reports a significant improvement of the estimation method based on dynamic ultrasound image. The proposed method consists of four processes: the extraction of dynamic and static texture features, frequency analysis, principal component analysis, and the estimation of BMS number by neural network. In order to evaluate the effectiveness of the proposed method, the experiments were conducted with or without dynamic image information. The number of target regions was set to 1 or 2, and two groups of samples, Case 1 and Case 2, were used for the experiments. Case 1 and Case 2 included 18 and 27 samples, which were measured at Saga Livestock Experiment Station and Nagasaki Agricultural and Forestry Technical Development Center, respectively. The image analysis was performed using only Case 1 or using the mixed group of Case 1 and 2. The experimental results with Case 1 showed the correlation coefficient of the estimated and the actual BMS number was improved from r=0.55 to r=0.79 by adding dynamic image information. Moreover, the correlation coefficient was further raised to r=0.84 with the number of target region increased from 1 to 2. Similarly, as for the mixed group of Case 1 and 2, the correlation coefficients were r=0.77, r=0.76, and r=0.88, respectively. These results suggested that a high estimation accuracy was achieved by adding dynamic image information and increasing target region.

抄録全体を表示