• ICROS
• /
• v.15 no.2
• /
• pp.17-25
• /
• 2009

본 고에서는 이동로봇의 동시간 위치인식 및 지도작성(Simultaneous Localizaton and Mapping;SLAM) 기술에 대하여 다룬다. 이동로봇의 SLAM을 위하여, 로봇과 랜드마크의 상태를 상태공간 영역에서 같이 기술하는 방법과 센서로부터 입력된 정보를 이용하여 로봇이 상태를 추정하는 기법을 소개한다. 실제 로봇을 통한 예제를 통하여 로봇의 상태와 특징점을 동시에 추정하는 것을 보여준다.

• Transactions of the Korean Society of Mechanical Engineers B
• /
• v.35 no.3
• /
• pp.237-244
• /
• 2011

An experimental study was carried out to investigate the effect of the inclination angle on the nucleate pool boiling of saturated water at atmospheric pressure. We considered an annulus with a gap of 5 mm and a bottom opening. The inner tube of the annulus was heated, and the outer diameter and the length of the tube were 25.4 mm and 500 mm, respectively. The inclination angle was varied from horizontal to vertical. The results were compared to those for an annulus with a larger gap and a single tube. In the small-gap annulus, the effect of the inclination angle on the heat transfer was not significant. However, an early onset of the critical heat flux was observed at 80 kW/$m^2$ when the annulus was horizontal. Liquid agitation and bubble coalescence were considered to be the major heat-transfer mechanisms.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.33 no.9
• /
• pp.56-65
• /
• 2005

A dynamic analysis method that freezes a time domain by discretization and solves the spatial propagation equation has a unique feature that provides a degree of freedom on spatial domain compared with the space discretization or space-time discretization finite element method. Using this feature, the time finite element analysis can be effectively applied to optimize the spatial characteristics of distributed type actuators. In this research, the time domain finite element method was used to discretize the model. A state variable vector was used in the discretization to include arbitrary initial conditions. A performance index was proposed on spatial domain to consider both potential and vibrational energy, so that the resulting shape of the distributed actuator was optimized for dynamic control of the structure. It is assumed that the structure satisfies the final rest condition using the realizable control scheme although the initial disturbance can affect the system response. Both equations on states and costates were derived based on the selected performance index and structural model. Ricatti matrix differential equations on state and costate variables were derived by the reconfiguration of the sub-matrices and application of time/space boundary conditions, and finally optimal actuator distribution was obtained. Numerical simulation results validated the proposed actuator shape optimization scheme.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.36 no.12
• /
• pp.1543-1549
• /
• 2012

This study focuses on the state space model and the design of a state observer for the slip dynamics between rolls in STS cold cluster mills. First, a mathematical model of the roll slip is given as a nonlinear differential equation. Then, by using a Taylor series expansion, it is linearized as a state space model. Next, by using Gopinath's algorithm, a minimal-order state observer based on the state space model is designed to estimate the angular speed of all idle rolls except for an actuated roll that is measureable. Finally, a computer simulation is used to validate that the proposed state space model very well describes slip dynamics between, and moreover, the state observer very well estimates the angular speed of the idle roll.

• Journal of KIISE:Software and Applications
• /
• v.30 no.11
• /
• pp.1054-1061
• /
• 2003

Q-Learning, a representative algorithm of reinforcement learning, experiences repeatedly until estimation values about all state-action pairs of state space converge and achieve optimal policies. When the state space is high dimensional or continuous, complex reinforcement learning tasks involve very large state space and suffer from storing all individual state values in a single table. We introduce Q-Map that is new function approximation method to get classified policies. As an agent learns on-line, Q-Map groups states of similar situations and adapts to new experiences repeatedly. State-action pairs necessary for fine control are treated in the form of rule. As a result of experiment in maze environment and mountain car problem, we can achieve classified knowledge and extract easily rules from Q-Map

• Proceedings of the Korea Air Pollution Research Association Conference
• /
• 2000.04a
• /
• pp.301-303
• /
• 2000

오늘날 경제적 생활환경의 개선으로 인하여 현대인의 일상생활 대부분이 다양한 실내공간에서 이루어지고 있고 인구 및 산업의 도시집중화에 따른 지하생활공간의 이용이 확대되는 둥 생활양식과 거주환경에 많은 변화를 가져오고 있다. 실내공간은 현대 도시인들에게 있어서 대부분의 시간을 소비하는 공간으로서(Wily et at., 1991) 실내공간의 공기오염이 보건학적인 중요한 관심사로 대두된 것은 1970년대 이후이며(Spongier, 1983) 현재 국내에서도 실내공간을 대상으로 오염물질에 대한 오염도가 조사되고 있으나 아직 미진한 상태이다(김윤신, 1994). (중략)

• Proceedings of the KAIS Fall Conference
• /
• 2011.05a
• /
• pp.400-403
• /
• 2011

다양한 특성의 도시공간의 확장은 경제 활동의 활성화는 물론 도시공간에서의 범죄발생 확대로 이어지고 있다. 이에 따라 국민의 생명과 재산보호, 사회질서유지 등을 위한 각종 대책과 경찰의 역할을 확대되고 있으나 효율적인 범죄예방 활동이 부족한 상태이다. 따라서 이 연구에서는 체계적인 범죄 모니터링을 위한 기초연구의 일환으로 도시공간에서의 범죄가 발생하는 공간에 대한 패턴을 이해하고 분석하여 도시 내 안전구역을 분석하여 GIS 공간분석기법의 효용성을 나타내었다.

• The KIPS Transactions:PartB
• /
• v.15B no.2
• /
• pp.159-170
• /
• 2008

Nowadays, one of the important functionalities required from robot task planners is to generate plans to compose existing component services into a new service. In this paper, we introduce the design and implementation of a heuristic search planner, JPLAN, as a kernel module for component service composition. JPLAN uses a local search algorithm and planning graph heuristics. The local search algorithm, EHC+, is an extended version of the Enforced Hill-Climbing(EHC) which have shown high efficiency applied in state-space planners including FF. It requires some amount of additional local search, but it is expected to reduce overall amount of search to arrive at a goal state and get shorter plans. We also present some effective heuristic extraction methods which are necessarily needed for search on a large state-space. The heuristic extraction methods utilize planning graphs that have been first used for plan generation in Graphplan. We introduce some planning graph heuristics and then analyze their effects on plan generation through experiments.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.10 no.7
• /
• pp.1188-1195
• /
• 2006

This paper presents an efficient design method of 2-D state-space digital filter based on an improved branch-and -bound algorithm. The resultant 2-D state-space digital filters whose coefficients are represented as the sum of two power-of-two terms, are attractive for high-speed operation and simple implementation. The feasibility of the proposed method is demonstrated by several experiments. The results show that the approximation error and group delay characteristic of the resultant filters are similar to those of the digital filters which designed in the continuous coefficient space.

• The KIPS Transactions:PartD
• /
• v.10D no.5
• /
• pp.813-820
• /
• 2003

This paper solves Push-Push game with the model checker NuSMY which exhaustively explores all search space to determine whether a model satisfies a property. In case a model doesn't satisfy properties to be checked, NuSMV generates a counterexample which tells where this unsatisfaction occurs. However, the algorithm for generating counterexample in NuSMV traverses a search space twice so that it is inefficient for solving the game we consider here. To save the time to be required to complete the game, we revise the part of counterexample generation so that it traverses a search space once. As a result, we obtain 62% time improvement and 11% space improvement in solving the game with modified NuSMV.