• The Journal of Korea Robotics Society
• /
• v.15 no.2
• /
• pp.197-204
• /
• 2020

In this paper, we present a learning platform for robotic grasping in real world, in which actor-critic deep reinforcement learning is employed to directly learn the grasping skill from raw image pixels and rarely observed rewards. This is a challenging task because existing algorithms based on deep reinforcement learning require an extensive number of training data or massive computational cost so that they cannot be affordable in real world settings. To address this problems, the proposed learning platform basically consists of two training phases; a learning phase in simulator and subsequent learning in real world. Here, main processing blocks in the platform are extraction of latent vector based on state representation learning and disentanglement of a raw image, generation of adapted synthetic image using generative adversarial networks, and object detection and arm segmentation for the disentanglement. We demonstrate the effectiveness of this approach in a real environment.

• Journal of Korea Multimedia Society
• /
• v.6 no.3
• /
• pp.549-555
• /
• 2003

CA is cost-effective to generate pseudorandom patterns than LFSR. Based on the effectiveness of a CA based pseudorandom pattern generator, CA have been employed successfully in several applications. Especially Nongroup CA is applied to efficient hash function generation, cryptography and image compression. In this paper we analyze the properties of TPNCA and by using basic paths in the 0-tree of a linear TPNCA we analyze the structure of the state-transition graph. Also by showing the structure of the complemented CA which have the acyclic state of the 0-tree as the complement vector is isomorphic to the structure of the original TPNCA, we reduce the time in analyzing the CA-states.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.19 no.5
• /
• pp.1104-1109
• /
• 2015

Due to the rapid growth in digital contents, it is important for us to design a high speed and secure encryption algorithm which is able to comply with the existing and future needs. This paper proposes an alternative approach for self-decimated LM-128 summation sequence generator, which will generate a higher throughput if compared to the conventional generator. We design and implement a threshold clock-controlled LM-128 and prove that it has a lower clock cycle and hence giving a higher key stream generation speed. The proposed threshold clock-control LM-128 generator consists of 256 bits inner state with 128 bits secret key and initialization vector. The cipher achieves a security level of 128 bits to be adapted to the digital contents security with high definition and high quality.

• Journal of Multimedia Information System
• /
• v.2 no.2
• /
• pp.233-240
• /
• 2015

Mobile devices are becoming increasingly sophisticated and the latest generation of smartphones now incorporates many diverse and powerful sensors. These sensors include acceleration sensor, magnetic field sensor, light sensor, proximity sensor, gyroscope sensor, pressure sensor, rotation vector sensor, gravity sensor and orientation sensor. The availability of these sensors in mass-marketed communication devices creates exciting new opportunities for data mining and data mining applications. In this paper, we describe and evaluate a system that uses phone-based accelerometers to perform activity recognition, a task which involves identifying the physical activity that a user is performing. To implement our system, we collected labeled accelerometer data from 10 users as they performed daily activities such as "phone detached", "idle", "walking", "running", and "jumping", and then aggregated this time series data into examples that summarize the user activity 5-minute intervals. We then used the resulting training data to induce a predictive model for activity recognition. This work is significant because the activity recognition model permits us to gain useful knowledge about the habits of millions of users-just by having them carry cell phones in their pockets.

• The Transactions of the Korea Information Processing Society
• /
• v.6 no.4
• /
• pp.988-998
• /
• 1999

This paper is to investigate the method of reducing the number of fuzzy rules with the help of LVQ. a large number of training patterns usually leads to a large set of fuzzy rules that require a large computer memory and take a long time to perform classification. so, in order to solve these problems, it is necessary to study to minimize the number of fuzzy rules. However, so as to minimize the performance degradation resulting from the reduction of fuzzy rules, fuzzy rules are generated after training the high-quality initial reference pattern. Through the simulation, we confirm that the proposed method is very effective.

• Journal of Power Electronics
• /
• v.14 no.6
• /
• pp.1081-1092
• /
• 2014

Renewable energy is being spotlighted as the electric power generating source for the next generation. Due to an increase in renewable energy systems in the grid system, their impact on the grid has become non-negligible. Thus, many countries in the world, including Europe, present their own grid codes for grid power conversion devices. In order to experiment with these grid codes, grid fault test equipment is required. This paper proposes both equipment and a control method, which are constructed with a 7-level cascaded H-bridge converter, that are capable of generating various grid faults. In addition, the Pulse Width Modulation (PWM) method for multilevel converters is compared and analyzed. The proposed structure, the control method, and the PWM method are verified through simulation and experimental results.

• Proceedings of the KIEE Conference
• /
• 2000.07a
• /
• pp.363-365
• /
• 2000

This paper presents an improved method to estimate the maximum load level for heavily loaded power systems with the load-generation variation vector by using the elliptic pattern of the P-e curve. The previous study suggested a simple technique of removing e-f coupling, where only high voltage load flow solutions to calculate transforming angle of system reference is needed. The proposed algorithm is improved to require only one load flow solution at a specific load level in addition to the operating point at the beginning stage, which reduces the computation time and the iteration number of estimation. The proposed method can be efficiently applied to heaviIy loaded systems with the combination of CPFlow when the reactive power limit and ULTC are considered. In this paper, the effect of ULTC on the estimation of maximum loadability index is also investigated. The proposed algorithm is tested on New England 39 bus system and IEEE 118 bus system.

• Proceedings of the KIEE Conference
• /
• 2005.11b
• /
• pp.347-349
• /
• 2005

This paper presents an efficient algorithm to estimate the maximum load level for heavily loaded power systems with the load-generation variation vector obtained by ELD (Economic Load Dispatch) and/or short term load forecasting while utilizing the elliptic pattern of the P-e curve. It is well known the power flow equation in the rectangular coordinate is fully quadratic. However, the coupling between e and f makes it difficult to take advantage of this quadratic characteristic. In this paper, a simple technique is proposed to reflect the e-f coupling effects on the estimation of maximum loadability with theoretical analysis. An efficient estimation algorithm has been developed with the use of the elliptic properties of the P-e curve. The proposed algorithm is tested on IEEE 14 bus system, New England 39 bus system and IEEE 118 bus system, which shows that the maximum load level can be efficiently estimated with remarkable improvement in accuracy.

• Spatial Information Research
• /
• v.8 no.1
• /
• pp.69-83
• /
• 2000

Cadastral map is the basic data that prescribe a lot number, the classification of land category, a boundary and ownerships of the parcels. Because the analogue cadastral map is not appropriate for the Parcel Based Land Information System, computerization of cadastral map is needed. When considering other automatic vectorizing softwares, we conclude that they can not satisfy the accuracy needed in cadastral map. Also screen digitizing methods demand lots of time. So we developed semi-automatic vectorizing program that realized almost capacities, such as overlay display which is needed for screen digitizing , window link, vector file generation , and so forth. As comparing screen digitizing method using AutoCAD with our developed program, we could obtain not only almost same accuracy , but also 35 minute reduction in vectorizing.

• The Journal of the Acoustical Society of Korea
• /
• v.14 no.1
• /
• pp.13-21
• /
• 1995

In this paper, we improve an image classifier algorithm based on neural network learning. It consists of two steps. The first is input pattern generation and the second, the global neural network implementation using an improved back-propagation algorithm. The feature vector for pattern recognition consists of the codebook data obtained from self-organization feature map learning. It decreases the input neuron number as well as the computational cost. The global neural network algorithm which is used in classifier inserts a control part and an address memory part to the back-propagation algorithm to control weights and unit-offsets. The simulation results show that it does not fall into the local minima and can implement easily the large-scale neural network. And it decreases largely the learning time.