In this research, the optimal manufacturing conditions of fine Si powders from Si scrap were investigated as a function of different initial powder size using the high-energy ball milling equipment, which produces the fine powder by means of an ultra high-energy within a short duration. The morphological change of the powders according to the milling time was observed by Scanning electron microscopy (SEM). With the increasing milling time, the size of Si powder was decreased. In addition, more energy and stress for milling were required with the decreasing initial powder size. The refinement of Si scrap was rapidly carried out at 10min ball milling time. However, the refined powder started to agglomerate at 30 min milling time, while the powder size became uniform at 60 min milling time.

This paper is a study on data augmentation for small dataset by using deep learning. In case of training a deep learning model for recognition and classification of non-mainstream objects, there is a limit to obtaining a large amount of training data. Therefore, this paper proposes a data augmentation method using perspective transform and image synthesis. In addition, it is necessary to save the object area for all training data to detect the object area. Thus, we devised a way to augment the data and save object regions at the same time. To verify the performance of the augmented data using the proposed method, an experiment was conducted to compare classification accuracy with the augmented data by the traditional method, and transfer learning was used in model learning. As experimental results, the model trained using the proposed method showed higher accuracy than the model trained using the traditional method.

This paper presents a sensitivity optimization of a MEMS (microelectromechanical systems) gyroscope for a magnet-gyro system. The magnet-gyro system, which is a guidance system for a AGV (automatic or automated guided vehicle), uses a magnet positioning system and a yaw gyroscope. The magnet positioning system measures magnetism of a cylindrical magnet embedded on the floor, and AGV is guided by the motion direction angle calculated with the measured magnetism. If the magnet positioning system does not measure the magnetism, the AGV is guided by using angular velocity measured with the gyroscope. The gyroscope used for the magnet-gyro system is usually MEMS type. Because the MEMS gyroscope is made from the process technology in semiconductor device fabrication, it has small size, low-power and low price. However, the MEMS gyroscope has drift phenomenon caused by noise and calculation error. Precision ADC (analog to digital converter) and accurate sensitivity are needed to minimize the drift phenomenon. Therefore, this paper proposes the method of the sensitivity optimization of the MEMS gyroscope using DEAS (dynamic encoding algorithm for searches). For experiment, we used the AGV mounted with a laser navigation system which is able to measure accurate position of the AGV and compared result by the sensitivity value calculated by the proposed method with result by the sensitivity in specification of the MEMS gyroscope. In experimental results, we verified that the sensitivity value through the proposed method can calculate more accurate motion direction angle of the AGV.

This paper presents a study of path-planning method for AGV(automated guided vehicle) based on path-tracking. It is important to find an optimized path among the AGV techniques. This is due to the fact that the AGV is conditioned to follow the predetermined path. Consequently, the path-planning method is implemented directly affects the whole AGV operation in terms of its performance efficiency. In many existing methods are used optimization algorithms to find optimized path. However, such methods are often prone with problems in handling the issue of inefficiency that exists in system's operation due to inherent undue time delay created by heavy load of complex computation. To solve such problems, we offer path-planning method using modified binary tree. For the purpose of our experiment, we initially designed a AGV that is equiped with laser navigation, two encoders, a gyro sensor that is meant to be operated within actual environment with given set of constrictions and layout for the AGV testing. The result of our study reflects the fact that within such environments, the proposed method showed improvement in its efficiency in finding optimized path.

This paper presents to study the path tracking method of AGV(autonomous guided vehicle) which has a laser guidance system. An existing automatic guided vehicles(AGVs) which were able to drive on wired line only had a automatic guidance system. However, the automatic guidance systems that those used had the high cost of installation and maintenance, and the difficulty of system change according to variation of working environment. To solve such problems, we make the laser guidance system which is consisted of a laser navigation and gyro, encoder. That is robust against noise, and flexible according to working environment through sensor fusion. The laser guidance system can do a perfect autonomous driving. However, the commercialization of perfect autonomous driving system is difficult, because the perfect autonomous driving system must recognize the whole environment of working space. Hence, this paper studied the path tracking of AGV using laser guidance system without wired line. The path tracking method is consisted of virtual path generation method and driving control method. To experiment, we use the fork-type AGV which is made by ourselves, and do a path tracking experiments repeatedly on same experimental environment. In result, we verified that proposed system is efficient and stable for actual fork-type AGV.

A probability-based data generation is a typical context-generation method that is a not only simple and strong data generation method but also easy to update generation conditions. However, the probability-based context-generation method has been found its natural-born ambiguousness and confliction problems in generated context data. In order to compensate for the disadvantages of the probabilistic random data generation method, a situation propagation network is proposed in this paper. The situation propagating network is designed to update parameters of probability functions are included in probability-based data generation model. The proposed probability-based context-generation model generates two kinds of contexts: one is related to independent contexts, and the other is related to conditional contexts. The results of the proposed model are compared with the results of the probability-based model with respect to performance, reduction of ambiguity, and confliction.