• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2005.11a
• /
• pp.335-338
• /
• 2005

로봇 에이전트는 변화하는 환경에서 센서정보를 바탕으로 적절한 행동을 선택하며 동작하는 것이 중요하다. 행동 선택 네트워크는 이러한 환경에서 변화하는 센서정보에 따라 실시간으로 행동을 선택할 수 있다는 점에서, 장시간에 걸친 최적화보다 단시간 내 개선된 효율성에 초점을 맞추어 사용되어 왔다. 하지만 행동 선택 네트워크는 초기 문제에 의존적으로 설계되어 변화하는 환경에 유연하게 대처하지 못한다는 맹점을 가지고 있다. 본 논문에서는 행동 선택 네트워크의 연결을 LCS를 기반으로 진화 학습시켰다. LCS는 유전자 알고리즘을 통해 만들어진 규칙들을 강화학습을 통해 평가하며, 이를 통해 변화하는 환경에 적합한 규칙을 생성한다. 제안하는 모델에서는 LCS의 규칙이 센서정보를 포함한다. 진화가 진행되는 도중 이 규칙들이 모든 센서 정보를 포함하지 못하기 때문에 현재의 센서 정보를 반영하지 못하는 경우가 발생할 수 있다. 본 논문에서는 이를 해결하기 위해 센서정보 별로 개체를 따로 유지하는 방법을 제안한다. 제안하는 방법의 검증을 위해 Webots 시뮬레이터에서 케페라 로봇을 이용해 실험을 하여, 변화하는 환경에서 로봇 에이전트가 학습을 통해 올바른 행동을 선택함을 보였고, 일반LCS를 사용한 것보다 조건별 개체 유지를 통해 더 나은 결과를 보이는 것 또한 확인하였다.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.6 no.2
• /
• pp.13-21
• /
• 1996

The necessity and usefulness of higher-order neural networks have been well-known since early days of neurocomputing. However the explosive number of terms has hampered the design and training of such networks. In this paper we present an evolutionary learning method for efficiently constructing problem-specific higher-order neural models. The crux of the method is the neural tree representation employing both sigma and pi units, in combination with the use of an MDL-based fitness function for learning minimal models. We provide experimental results in classification and prediction problems which demonstrate the effectiveness of the method. I. Introduction topology employs one hidden layer with full connectivity between neighboring layers. This structure has One of the most popular neural network models been very successful for many applications. However, used for supervised learning applications has been the they have some weaknesses. For instance, the fully mutilayer feedforward network. A commonly adopted connected structure is not necessarily a good topology unless the task contains a good predictor for the full *d*dWs %BH%W* input space.

• Proceedings of the KIEE Conference
• /
• 2008.10b
• /
• pp.341-342
• /
• 2008

GA(Genetic Algorithm)는 NP-Complete 도메인이나 NP-Hard 도메인 내의 문제들에 대해서 최적의 해를 찾기 위해서 많이 사용되어 지는 진화 컴퓨팅 방법 중 하나이다. 모바일 로봇 기술 중 경로계획은 NP-Complete 도메인 영역의 문제 중 하나로 이를 해결하기 위해서 Dijkstra 등의 그래프 이론을 이용한 연구가 많이 연구되었고 최근에는 GA등 진화 컴퓨팅 기법을 이용하여 최적의 경로를 찾는 연구가 많이 수행되고 있다. 그러나 모바일 로봇이 처리해야 될 공간 정보 크기가 증가함에 따라 기존 GA의 개체의 크기가 증가되어 게산 복잡도가 높아져 시간 지연등의 문제가 발생할 수 있다. 이는 모바일 로봇의 잠재적 오류로 발생될 수 있다. 공간 정보에는 동적이 장애물들이 예측 불허하게 나타 날 수 있는데 이것은 전역 경로 계획을 수립할 때 또한 반영되어야 된다. 본 논문에서는 k-means 클러스터링 기법을 이용하여 장애물 밀집도 및 거리 정보를 기반으로 공간정보를 k개의 군집 공간으로 재분류하여 이를 기반으로 N*M개의 그리드 개체 집단을 생성하여 최적 경로계획을 수립하는 GA를 제시한다.

• 연구논문집
• /
• s.34
• /
• pp.113-119
• /
• 2004

Recently, the conventional concept of a machine-tool in manufacturing systems is changing from the target of integration to the autonomous manufacturing device based on a knowledge evolution. Subsequently, a machine-tool has been the subject of a cooperation through an advanced environment where an open architecture controller, high speed network and internet technology are contained In the future, a machine-tool will be more improved in the form of a knowledge evolution based device. In order to develop the knowledge evolution based machine-tool, this paper proposes the structure of knowledge evolution and the scheme of a dialogue agent among agent-based modules such as a sensory module, a dialogue module, and an expert system. The dialogue agent has a role of interfacing with another machine for cooperation. To design of the dialogue agent module in M2M(Machine To Machine)environment, FIPA-OS and ping agent based on FIPA-OS are analyzed in this study. Through this, it is expected that the dialogue agent module can be more efficiently designed and the knowledge evolution based machine-tool can be hereafter more easily implemented.

• Proceedings of the KIEE Conference
• /
• 1999.11c
• /
• pp.576-578
• /
• 1999

본 논문에서는 생물학적인 DNA와 유전자 알고리즘의 진화 메커니즘에 근거를 둔 DNA 코딩방법을 변형하여 새로운 DNA 코딩 방법을 제안한다. 이 방법은 기존의 DNA 코딩 방법이 DNA 유전자의 Redundancy와 Over-lapping 성질 때문에 갖고 있는 DNA 자체의 특성인 염색체의 길이를 자유자재로 변화시킬 수 있는 코딩 기술에 진화단계에서 변형을 가할 수 있는 새로운 유전자 알고리즘을 추가하여, 초기에 국소해로 접근하는 일반적인 유전자 알고리즘의 위험 부담률을 줄이고, 전역 해로의 접근 가능성을 높이는 방법을 제시한다. 또한. 이 변형된 DNA 코딩 방법의 가능성을 입증하기 위하여 시스템 제어에 필요한 지식을 표현하는 적당한 퍼지 규칙을 후건부의 매개변수의 동조만을 통하여 획득하고, 이 규칙에 변형된 DNA 코딩 방법을 적용하여 최적화 된 새로운 퍼지규칙 획득 알고리즘을 개발한다. 제안된 알고리즘을 이용한 퍼지 제어기를 설계하고. 이 제어기의 유용성을 입증하기 위하여 병렬형 이중 도립진자 시스템에 적용하여 시뮬레이션을 실행한 결과 효과적으로 퍼지규칙을 획득하고 제어함을 알 수 있다.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.25 no.3
• /
• pp.289-295
• /
• 2014

In this paper a wideband active balun using advanced phase correction architecture is proposed. The proposed active balun is constructed with each different architecture of active balun combined with the cascode architecture to improve phase correction performance compared with conventional phase correction techniques. Operating over 0.6~2.0 GHz band, the proposed balun shows $10^{\circ}$ of phase error and 2 dB of gain error with 7 mW power consumption from 1.8 V supply voltage.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.19 no.1
• /
• pp.96-101
• /
• 2009

This paper presents a method for identifying the parameter set of inelastic constitutive equations, which is based on an Evolutionary Algorithm. The advantage of the method is that appropriate parameters can be identified even when the measured data are subject to considerable errors and the model equations are inaccurate. The design of experiments suited for the parameter identification of a material model by Chaboche under the uniaxial loading and stationary temperature conditions was first considered. Then the parameter set of the model was identified by the proposed method from a set of experimental data. In comparison to those by other methods, the resultant stress-strain curves by the proposed method correlated better to the actual material behaviors.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.20 no.4
• /
• pp.522-527
• /
• 2010

In this paper, a new MBHF(Morphology-Based Homomorphic filter) is presented to enhance contrast in mammographic images. The MBH filtering is performed based on the morphological sub-bands, in which an image is morphologically decomposed. The filter is designed to have optimal gain and structuring element in each sub-band through differential evolution. Experimental results show that the proposed method improves the contrast in mammographic images such that an evaluation criterion, WPSNR(Weighted Peak Signal to Noise Ratio) which takes into account human visual system is increased compared with a wavelet-based Homomorphic filter.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.11 no.4
• /
• pp.359-365
• /
• 2001

In this paper, we propose the design methodology of target tracking system using fuzzy basis function expansion(FBFE) based on virus evolutionary genetic algorithm (VEGA). In general, the objective of target tracking is to estimate the future trajectory of the target based on the past position of the target obtained from the sensor. In the conventional and mathematical nonlinear filtering method such as extended Kalman filter(EKF), the performance of the system may be deteriorated in highly nonlinear situation. To resolve these problems of nonlinear filtering technique, by appling artificial intelligent technique to the tracking control of moving targets, we combine the advantages of both traditional and intelligent control technique. In the proposed method, after composing training datum from the parameters of extended Kalman filter, by combining FDFE, which has the strong ability for the approximation, with VEGA, which prevent GA from converging prematurely in the case of lack of genetic diversity of population, and by idenLifying the parameters and rule numbers of fuzzy basis function simultaneously, we can reduce the tracking error of EKF. Finally, the proposed method is applied to three dimensional tracking problem, and the simulation results shows that the tracking performance is improved by the proposed method.

• Journal of the Korean Institute of Telematics and Electronics C
• /
• v.35C no.12
• /
• pp.85-98
• /
• 1998

This paper proposes a new design method of neuro-FLC by the Lamarckian co-adaptation scheme that incorporates the backpropagation learning into the GA evolution in an attempt to find optimal design parameters (fuzzy rule base and membership functions) of application-specific FLC. The design parameters are determined by evolution and learning in a way that the evolution performs the global search and makes inter-FLC parameter adjustments in order to obtain both the optimal rule base having high covering value and small number of useful fuzzy rules and the optimal membership functions having small approximation error and good control performance while the learning performs the local search and makes intra-FLC parameter adjustments by interacting each FLC with its environment. The proposed co-adaptive design method produces better approximation ability because it includes the backpropagation learning in every generation of GA evolution, shows better control performance because the used COG defuzzifier computes the crisp value accurately, and requires small workspace because the optimization procedure of fuzzy rule base and membership functions is performed concurrently by an integrated fitness function on the same fuzzy partition. Simulation results show that the Lamarckian co-adapted FLC produces the most superior one among the differently generated FLCs in all aspects such as the number of fuzzy rules, the approximation ability, and the control performance.