To remove weeds around a grape vine without damaging the stem, we propose a contact recognition technology in this paper. The developed technique can reliably determine the position of the vine's lower stem, which may be deformed under sufficient force and may be buried in the weeds. We have developed a new method to detect the position of grape stems as objects in which its deformation is considered. By calculating the deformation from the relations between the geometry and mechanics of manipulator's link postures before and after contact point detection, the position of the object can be determined with high accuracy. Because stem position detection commences after tactile recognition has been performed, any deformation of the object has already occurred. The effect of this initial deformation on the detection accuracy cannot be ignored. Therefore, position detection of object is divided into two operations : reducing the object deformation and detecting the object position. In this way, highly accurate recognition of its position is achieved. The paper continues by giving a detailed explanation of the proposed method.

View full abstract

Gaussian filter to extract the roughness profile from a primary profile suffers strongly from the outliers mingled in the data. To solve this problem, several schemes of robust Gaussian filter have been proposed. However there are several fatal problems that a mean line determined with respect to the measurement data containing no outliers is not coincident with the mean line of the Gaussian filter output. To solve this problem, we have been proposed the robust Gaussian filter by using the fast M-Estimation method. This proposed method could provide completely the same results of the original Gaussian filter method. However, it is not confirmed whether a transmission characteristic of this method is corresponding to the transmission characteristic of the Gaussian filter. Therefore, this method cannot be applied to the real applications for the measurement of roughness profile where the transmission characteristic of Gaussian filter is strictly treated. In this paper, we confirm that the transmission characteristic of this method is corresponding to the transmission characteristic of the Gaussian filter.

View full abstract

New technologies are required to obtain information related to humans, such as their physical position or posture. Many methods are used for human body detection, although image processing is the most popular because it can obtain more information than the others. However, image processing cannot detect a human who does not move when the background is complicated. We propose combining image processing with an estimation of the size of a human body with a scanning pyroelectric electric sensor. When the pyroelectric sensor scans, the volume of the heat source in the rotatory direction is measured. A heat source domain is provided from the image information of the camera photography range. In addition, the position of the person can be measured similarly to stereo-camera measurement if scanned by two sensor units at the same time. In this work, results of measurement by a prototype are discussed.

View full abstract

Path planning is necessary for an autonomous moving robot to achieve an effective indoor moving and this planning is conducted following the map of moving environment. To develop those maps, robot localization is very important. In this study, robot localization area is extended without depending on the natural features of the environment as conventional approaches. Here, the door positions in the environment are used as landmarks and robot position is measured according to them. The door positions and baseboards in a wall align on a straight line, therefore, long distance landmark positions can be detected following baseboards. These baseboards are detected by an on-robot camera and walls are also detected following them. On the other hand, an on-robot LRF measures the distance between walls and robot. In this paper, the measuring distance of LRF is extended by combining the detected baseboards by the camera. With this extension, long distance landmark detection can be conducted, as a result robot localization area could be extended. Experiments were conducted by operating a robot to confirm the door position detection, and then robot localization area extension was confirmed.

View full abstract

In this research, we have been developing the denoising method by use of wavelet transform. In previous report, this method was investigated by comparison with conventional low-pass filter and Fourier transform. It was concluded that wavelet transform was more effective for elimination of white noise and this method was capable of removing the noise not only in the frequency domain, but also in amplitude domain. In this report, the denoising method by use of wavelet transform is applied into evaluation of Si wafer profile data which is acquired by an actual measurement system. As the acquired data includes not only the white noise but also the inherent noise introduced by the measurement system itself, it is found that the proposed method of filtering white noise in both frequency and amplitude domains is insufficient to improve GBIR (Global Backside Ideal Range which is also equal to peak-to-valley). The cutoffs used in the proposed method are then fine-turned via least-square procedure. The results demonstrate the validation of newly designed filter in GBIR improvement.

View full abstract

This study deals with a method of determining suitable cutting conditions without using a database. It involves deriving the cutting conditions from the drilling conditions by considering the energy efficiency. The energy efficiency is the ratio of the supplied energy to the removed volume. This can be used as a parameter to compare drilling with cutting. In addition, the energy efficiency ratio given by the ratio of both parameters is used for analyses. Mild steel was used to examine the relation between the cutting energy efficiency and drilling energy efficiency under various processing conditions. As a result, the relation between the perforation time and the energy efficiency ratio was found. A system used to derive the cutting conditions from this relation was constructed from the drilling data.

View full abstract

This paper describes the elucidation about the removal of surface oxidation layer of copper iron alloy for lead frame after heating processing using Na₂SO₄-electrolyzed oxidizing water (It abbreviates as Na₂SO₄ EO water). First, the etching characteristics of Na₂SO₄ EO water against the surface of copper iron alloy for lead frame was clarified using H₂SO₄ solution and NaCl EO water for comparative immersion experiments. The results showed that as for the etching efficiency on copper iron alloy, Na₂SO₄ EO water was superior to H₂SO₄ solution, while inferior to NaCl EO water. Also, Na₂SO₄ EO water more uniformly to the composition element of copper iron alloy by the etching compared with H₂SO₄ solution. As for the relationship between immersing time and etching speed, Na₂SO₄ EO water is same as NaCl EO water. Next, as for test pieces after heating processing, it did the surface oxidation layer removal experiment by the immersion. The results show that Na₂SO₄ EO water doesn't pollute the test pieces surface compared with NaCl EO water and it is possible to apply to the surface oxidation layer removal process.

View full abstract

Although many welding processes in metal work have been robotized recently, manual welding is still indispensable in custom-made production and high-mix low-volume production. Hence, skilled welders are needed widely in metal work industry. On the other hand, many people have little inducement to be a new welder because of high physical risk and severeness in skill training. Virtual Reality technology may give an effective solution on it and various high-tech VR systems for the training have been proposed so far. However, high tech is not always necessary to build a VR training system, as long as skill to be trained is analyzed and situation is confined. In this paper, the authors have proposed a simple VR training system for steady state electrode manipulation skill in manual arc welding, which adopted neither force display nor mixed reality, and verified its effectiveness. As a result, a good correlation is found between skills in VR welding and in actual welding from experimental tasks, and consequently the proposed system can train novice welders safely.

View full abstract