• The Journal of Korea Robotics Society
• /
• v.15 no.4
• /
• pp.404-412
• /
• 2020

Typically everyday human life tasks involve at least two people moving objects such as tables and beds, and the balancing of such object changes based on one person's action. However, many studies in previous work performed their tasks solely on robots without factoring human cooperation. Therefore, in this paper, we propose cooperative robot for table balancing using Q-learning that enables cooperative work between human and robot. The human's action is recognized in order to balance the table by the proposed robot whose camera takes the image of the table's state, and it performs the table-balancing action according to the recognized human action without high performance equipment. The classification of human action uses a deep learning technology, specifically AlexNet, and has an accuracy of 96.9% over 10-fold cross-validation. The experiment of Q-learning was carried out over 2,000 episodes with 200 trials. The overall results of the proposed Q-learning show that the Q function stably converged at this number of episodes. This stable convergence determined Q-learning policies for the robot actions. Video of the robotic cooperation with human over the table balancing task using the proposed Q-Learning can be found at http://ibot.knu.ac.kr/videocooperation.html.

• International Journal of Internet, Broadcasting and Communication
• /
• v.12 no.2
• /
• pp.37-44
• /
• 2020

The fifth generation (5G) mobile communication has been commercialized and the 5G applications, such as the artificial intelligence (AI) and the internet of things (IoT), are deployed all over the world. The 5G new radio (NR) wireless networks are characterized by 100 times more traffic, 1000 times higher system capacity, and 1 ms latency. One of the promising 5G technologies is non-orthogonal multiple access (NOMA). In order for the NOMA performance to be improved, sometimes the additive signal-dependent Gaussian noise (ASDGN) channel model is required. However, the channel capacity calculation of such channels is so difficult, that only lower and upper bounds on the capacity of ASDGN channels have been presented. Such difficulties are due to the specific constraints on the dependency. Herein, we provide the capacity of ASDGN channels, by removing the constraints except the dependency. Then we obtain the ASDGN channel capacity, not lower and upper bounds, so that the clear impact of ASDGN can be clarified, compared to additive white Gaussian noise (AWGN). It is shown that the ASDGN channel capacity is greater than the AWGN channel capacity, for the high signal-to-noise ratio (SNR). We also apply the analytical results to the NOMA scheme to verify the superiority of ASDGN channels.

• KIEE International Transaction on Electrical Machinery and Energy Conversion Systems
• /
• v.4B no.1
• /
• pp.29-35
• /
• 2004

Locomotive scheduling in railway systems experiences many difficulties because of the complex interrelations among resources, knowledge and various constraints. Artificial intelligence technology has been applied to solve these scheduling problems. These technologies have proved to be efficient in representing knowledge and rules for complex scheduling problems. In this paper, we have applied the CSP (Constraints Satisfaction Problems) programming technique, one of the AI techniques, to solve the problems associated with locomotive scheduling. This method is more effective at solving complex scheduling problems than available mathematical programming techniques. The advanced locomotive scheduling system using the CSP programming technique is realized based on the actual timetable of the Saemaul type train on the Kyong-bu line. In this paper, an overview of the CSP programming technique is described, the modeling of domain and constraints is represented and the experimental results are compared with the real-world existing schedule. It is verified that the scheduling results by CSP programming are superior to existing scheduling performed by human experts. The executing time for locomotive scheduling is remarkably reduced to within several decade seconds, something requiring several days in the case of locomotive scheduling by human experts.

• Communications for Statistical Applications and Methods
• /
• v.27 no.3
• /
• pp.313-326
• /
• 2020

Deep Learning is the most important key to the development of Artificial Intelligence (AI). There are several distinguishable architectures of neural networks such as MLP, CNN, and RNN. Among them, we try to understand one of the main architectures called Recurrent Neural Network (RNN) that differs from other networks in handling sequential data, including time series and texts. As one of the main tasks recently in Natural Language Processing (NLP), we consider Neural Machine Translation (NMT) using RNNs. We also summarize fundamental structures of the recurrent networks, and some topics of representing natural words to reasonable numeric vectors. We organize topics to understand estimation procedures from representing input source sequences to predict target translated sequences. In addition, we apply multiple translation models with Gated Recurrent Unites (GRUs) in Keras on English-Korean sentences that contain about 26,000 pairwise sequences in total from two different corpora, colloquialism and news. We verified some crucial factors that influence the quality of training. We found that loss decreases with more recurrent dimensions and using bidirectional RNN in the encoder when dealing with short sequences. We also computed BLEU scores which are the main measures of the translation performance, and compared them with the score from Google Translate using the same test sentences. We sum up some difficulties when training a proper translation model as well as dealing with Korean language. The use of Keras in Python for overall tasks from processing raw texts to evaluating the translation model also allows us to include some useful functions and vocabulary libraries as well.

• International journal of advanced smart convergence
• /
• v.9 no.2
• /
• pp.68-75
• /
• 2020

Non-orthogonal multiple access (NOMA) has gained the significant attention in the fifth generation (5G) mobile communication, which enables the advanced smart convergence of the artificial intelligence (AI), the internet of things (IoT), and many of the state-of-the-art technologies. Recently, correlated superposition coding (SC) has been proposed in NOMA, to achieve the near-perfect successive interference cancellation (SIC) bit-error rate (BER) performance for the stronger channel users, and to mitigate the severe BER performance degradation for the weaker channel users. In the correlated SC NOMA scheme, the stronger channel user BER performance is even better than the perfect SIC BER performance, for some range of the power allocation factor. However, such excessively good BER performance is not good for the user-fairness, i.e., the more power to the weaker channel user and the less power to the stronger channel user, because the excessively good BER performance of the stronger channel user results in the worse BER performance of the weaker channel user. Therefore, in this paper, we propose the power splitting to establish the user-fairness between both users. First, we derive a closed-form expression for the power splitting factor. Then it is shown that in terms of BER performance, the user-fairness is established between the two users. In result, the power splitting scheme could be considered in correlated SC NOMA for the user-fairness.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2004.10a
• /
• pp.225-232
• /
• 2004

Recently, there has been a rapid development in computer technology, which has in turn led to develop the automated welding system using Artificial Intelligence (AI). However, the automated welding system has not been achieved duo to difficulties of the control and sensor technologies. In this paper, the classification of the optimized bead geometry such as bead width, height penetration and bead area in the Gas Metal Arc (GMA) welding with fuzzy logic is presented. The fuzzy C-Means algorithm (FCM), which is best known an unsupervised fuzzy clustering algorithm is employed here to analysis the specimen of the bead geometry. Then the quality of the GMA welding can be classified by this fuzzy clustering technique and the choice for obtaining the optimal bead geometry can also be determined.

• Informatization Policy
• /
• v.24 no.4
• /
• pp.79-92
• /
• 2017

As city functions develop more complex and advanced, interests in smart cities are also increasing. Smart cities refer to the cities effectively solving urban problems such as traffic, safety, welfare, and living issues by utilizing ICT. Recently, many countries are attempting to introduce big data, Internet of Things, and artificial intelligence into smart cities, but they have not yet developed into comprehensive urban services. In this paper, we review the current status of domestic and overseas smart cities and suggest ways to solve issues of data sharing and service compatibility. To this end, we propose a "Deep Learning City Framework" that incorporates the deep learning technology into smart city services, and propose a new smart city strategy that safely shares spatial and temporal data in cities and converges learning data of various cities.

• Journal of Venture Innovation
• /
• v.2 no.1
• /
• pp.1-11
• /
• 2019

Many research or consulting institute refered to Artificial Intelligence, Internet of Things, Blockchain technology and 3D Printing as key driving forces and technologies of 4th industrial revolution. Compared with traditional manufacturing as a subtractive manufacturing(SM), 3D printing technology as an additive manufacturing(AM) will revolutionary impacts on many industries. This study compared 3D printing with traditional manufacturing in the economic, manufacturing, and marketing perspectives. This study also analyzed issues of 3D printing for the purpose of building business ecosystem. Finally agenda for the further research were suggested.

• Journal of Korea Game Society
• /
• v.13 no.6
• /
• pp.43-52
• /
• 2013

Path finder is one of the very important algorithm of artificial intelligence and is a process generally used in many game fields. Path finder requires many calculation, so it exerts enormous influences on performances. To solve this, many researches on the ways to reduce the amount of calculate operations have been made, and the typical example is A* algorithm but it has unnecessary computing process, reducing efficiency. In this paper, to reduce the amount of calculate operations such as node search with costly arithmetic operations, we proposes the weight based pre-processing A* algorithm. The simulation was materialized to measure the efficiency of the weight based pre-process A* algorithm, and the results of the experiments showed that the weight based method was approximately 1~2 times more efficient than the general methods.

• Proceedings of the Korea Inteligent Information System Society Conference
• /
• 2002.05a
• /
• pp.284-293
• /
• 2002

XML(eXtensible Markup Language, XML)은 인터넷에서의 자료 교환을 위해 고안된 언어이다. 본 논문에서는 XML의 개념을 발전시킨 확장형 규칙 표식 언어(extensible Rule Markup Language, XRML)를 제안하고 있다. XRML은 웹 페이지에 내재된 암묵적 규칙의 식별, 구조적인 규칙으로의 변환, 사람과 소프트웨어 에이전트간의 지식 공유를 가능하게 하며, 이를 통해 지식기반시스템(Knowledge Based System)과 지식관리시스템(Knowledge Management System)의 통합을 실현할 수 있는 새로운 언어가 될 것이다. 본 고에서는 XRML이 이상과 같은 능력을 갖기 위해 반드시 갖춰야 할 6가지 설계 기준과, 이들 기준을 반영한 XRML 구성 요소로서 RIML(Rule Identification Markup Language), RSML(Rule Structure Markup Language)과 RTML(Rule Triggering Markup Language)을 설계하였으며, 개별 요소들의 기능 및 특성과 함께 태그와 DTD(Document Type Definition)도 식별하였다. 나아가 전술한 구조를 기반으로 하여 XRML을 워크플로우 시스템상의 폼처리에 적용한 Form/XRML이라는 프로토타입 시스템을 설계하고 구현하였다. 본 프로토타입의 개발을 통해, 지식기반시스템의 지식을 활용하는 RTML이 폼을 비롯한 다양한 응용시스템에 내재될 수 있으며, 웹 페이지의 암묵적 규칙과 지식기반시스템의 규칙이 일관성 있게 유지될 수 있음을 보여 주었다. 요컨대 본 연구는 XRML이 지능형 웹으로 발전하기 위한 새로운 도구이며, KBS와 US의 통합을 위한 중요한 도구임을 입증하였다는 점에서 큰 의의를 갖는다고 하겠다.