• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.10 no.1
• /
• pp.43-48
• /
• 2010

This article presents the experimental studies of controlling angle and position of the inverted pendulum system using neural network to compensate for errors caused due to fuzzy controller. Although fuzzy control method can deal with nonlinearities of the system, fixed fuzzy rules may not work and result in tracking errors in some cases. First, a nominal Takagi-Sugeno (TS) type fuzzy controller with fixed weights is used for controlling the inverted pendulum system. Then the neural network is added at the reference input to form the reference compensation technique (RCT)control structure. Neural network modifies the input trajectories to improve system performances by updating internal weights in on-line fashion. The back-propagation learning algorithm for neural network is derived and used to update weights. Control hardware of a DSP 6713 board to have real time control is implemented. Experimental results of controlling inverted pendulum system are conducted and performances are compared.

• Proceedings of the KIEE Conference
• /
• 2000.07d
• /
• pp.2380-2382
• /
• 2000

Fuzzy system which are based on membership functions and rules, can control nonlinear, uncertian, complex system well. However, Fuzzy controller has problems: It is difficult to design a stable for amateur. To update the then-part membership functions of the fuzzy controller can be designed using the error back-propagation algorithm to be minimized error. Then we could be optimized the system choosing a good performance index. The proposed fuzzy controller based on neural network is applied to control an inverted pendulum for demonstration of the robustness of proposed methodology.

• Proceedings of the Korean Information Science Society Conference
• /
• 2005.07b
• /
• pp.13-15
• /
• 2005

P2P(Peer-to-Peer) 시스템은 대용량의 데이타를 공유하는데 유용하며, 네트워크 구조에 따라 중앙 집중형, 구조적 분산형, 그리고 비구조적 분산형으로 분류된다. 이 중 Gnutella와 같은 비구조적 분산형 P2P 시스템은 확장성과 신뢰성 측면에서 장점을 갖지만, 참여하는 노드의 수가 증가함에 따라 원하는 자원을 액세스하는 비용도 증가하는 문제를 가진다. 데이터 중복을 이용해 이러한 문제를 해결할 경우 중복된 데이타들의 일관성 유지를 위한 기법이 필요하다. 본 논문에서는 특정 노드가 갱신한 데이타를 중복된 사본을 저장하고 있는 다른 노드에 전파하기 위한 하이브리드 push/pull 기반의 갱신 전파 기법을 제안한다.

• Journal of KIISE:Databases
• /
• v.27 no.2
• /
• pp.175-184
• /
• 2000

A data warehouse is an integrated and summarized collection of data that can efficiently support decision making process. The summarized data at the data warehouse is often stored in materialized views. These materialized views need to be updated when source data change. Since the propagation of updates to the views may impose a significant overhead, it is very important to update the warehouse views efficiently. Though various strategies have been proposed to maintain views in the past, they typically require too much accesses to the data sources when the changes of multiple data sources have to be reflected in the view. In this paper we propose an efficient view update strategy that uses relatively small number of accesses to the data sources. We also show the performance advantage of our method over other existing methods through experiments using TPC-D data and queries.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.16 no.10 s.101
• /
• pp.965-972
• /
• 2005

In this paper, we first review LDPC codes in general and a belief propagation algorithm that works in logarithm domain. LDPC codes, which is chosen for second generation digital video broadcasting standard, are required a large number of computation due to large size of coded block and iteration. Therefore, we presented two kinds of low computational algorithm for LDPC codes. First, sequential decoding with partial group is proposed. It has same H/W complexity, and fewer number of iteration's are required at same performance in comparison with conventional decoder algerian. Secondly, early detection method for reducing the computational complexity is proposed. Using a confidence criterion, some bit nodes and check node edges are detected early on during decoding. Through the simulation, we knew that the iteration number are reduced by half using subset algorithm and computational complexity of early detected method is about $50\%$ offs in case of check node update, $99\%$ offs in case of check node update compared to conventional scheme.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.35 no.2C
• /
• pp.148-154
• /
• 2010

In this paper, we first review LDPC codes in general and a belief propagation algorithm that works in logarithm domain. LDPC codes, which is chosen 802.11n for wireless local access network(WLAN) standard are required a large number of computation due to large size of coded block and iteration. Therefore, we presented three kinds of low computational algorithm for LDPC codes. First, sequential decoding with partial group is proposed. It has same H/W complexity, and fewer number of iteration's are required at same performance in comparison with conventional decoder algorithm. Secondly, we have apply early stop algorithm. This method is reduced number of unnecessary iteration. Third, early detection method for reducing the computational complexity is proposed. Using a confidence criterion, some bit nodes and check node edges are detected early on during decoding. Through the simulation, we knew that the iteration number are reduced by half using subset algorithm and early stop algorithm is reduced more than one iteration and computational complexity of early detected method is about 30% offs in case of check node update, 94% offs in case of check node update compared to conventional scheme.

• Journal of Intelligence and Information Systems
• /
• v.14 no.1
• /
• pp.117-129
• /
• 2008

The cost-based scheduling work has been done in both the Operations Research (OR) and Artificial Intelligence (AI) literature. To deal with more realistic problems, AI-based heuristic scheduling approach with non-regular performance measures has been studied. However, there has been little research effort to develop a full inter-job cost propagation algorithm (CPA) for different jobs having multiple downstream and upstream activities. Without such a CPA, decision-making in scheduling heuristics relies upon local, incomplete cost information, resulting in poor schedule performance from the overall cost minimizing objective. For such a purpose, we need two types of CPAs : intra-job CPA and inter-job CPA. Whenever there is a change in cost information of an activity in a job in the process of scheduling, the intra-job CPA updates cost curves of other activities connected through temporal constraints within the same job. The inter-job CPA extends cost propagation into other jobs connected through precedence relationships. By utilizing the cost information provided by CPAs, we propose cost-based scheduling heuristics that attempt to minimize the total schedule cost. This paper develops inter-job CPAs that create and update cost curves of each activity in each search state, and propagate cost information throughout a whole network of temporal constraints. Also we propose various cost-based scheduling heuristics that attempt to minimize the total schedule cost by utilizing the cost propagation algorithm.

• Journal of KIISE:Databases
• /
• v.32 no.1
• /
• pp.1-11
• /
• 2005

An aggregation query computes aggregate information over a data cube in the query range specified by a user Existing methods based on the prefix-sum approach use an additional cube called the prefix-sum cube(PC), to store the cumulative sums of data, causing a high space overhead. This space overhead not only leads to extra costs for storage devices, but also causes additional propagations of updates and longer access time on physical devices. In this paper, we propose a new prefix-sum cube called 'SPEC' which drastically reduces the space of the PC in a large data warehouse. The SPEC decreases the update propagation caused by the dependency between values in cells of the PC. We develop an effective algorithm which finds dense sub-cubes from a large data cube. We perform an extensive experiment with respect to various dimensions of the data cube and query sizes, and examine the effectiveness and performance ot our proposed method. Experimental results show that the SPEC significantly reduces the space of the PC while maintaining a reasonable query performance.

• Journal of Institute of Control, Robotics and Systems
• /
• v.19 no.8
• /
• pp.702-708
• /
• 2013

Underwater TRN (Underwater Terrain Referenced Navigation) estimates an underwater vehicle state by measuring a distance between the vehicle and undersea terrain, and comparing it with the known terrain database. TRN belongs to absolute navigation methods, which are used to compensate a drift error of dead reckoning measurements such as IMU (Inertial Measurement Unit) or DVL (Doppler Velocity Log). However, underwater TRN is different to other absolute methods such as USBL (Ultra-Short Baseline) and LBL (Long Baseline), because TRN is independent of the external environment. As a magnetic-field-based navigation, TRN is a kind of geophysical navigation. This paper develops an EKF (Extended Kalman Filter) formulation for underwater TRN. A filter propagation part is composed by an inertial navigation system, and a filter update is executed with echo-sounder measurement. For large-initial-error cases, an adaptive EKF approach is also presented, to keep the filter be stable. At the end, simulation studies are given to verify the performance of the proposed TRN filter. With simplified sensor and terrain database models, the simulation results show that the underwater TRN could support conventional underwater navigation methods.

• Proceedings of the KIEE Conference
• /
• 2003.11b
• /
• pp.171-174
• /
• 2003

In this paper, a Learning module for a reconfigurable neural network processor(ERNIE) was proposed for an On-chip learning. The existing reconfigurable neural network processor(ERNIE) has a much better performance than the software program but it doesn't support On-chip learning function. A learning module which is based on Back Propagation algorithm was designed for a help of this weak point. A pipeline structure let the learning module be able to update the weights rapidly and continuously. It was tested with five types of alphabet font to evaluate learning module. It compared with C programed neural network model on PC in calculation speed and correctness of recognition. As a result of this experiment, it can be found that the neural network processor(ERNIE) with learning module decrease the neural network training time efficiently at the same recognition rate compared with software computing based neural network model. This On-chip learning module showed that the reconfigurable neural network processor(ERNIE) could be a evolvable neural network processor which can fine the optimal configuration of network by itself.