• Proceedings of the Korean Society of Computer Information Conference
• /
• 2024.01a
• /
• pp.375-377
• /
• 2024

본 논문에서는 OpenAI Gym 환경에서 프로그램으로 간단한 제어가 가능한 Mountain-Car-v0 게임에 대해 DQN(Deep Q-Networks) 강화학습을 진행하였다. 본 논문에서 적용한 DQN 네트워크는 입력층 1개, 은닉층 3개, 출력층 1개로 구성하였고, 입력층과 은닉층에서의 활성화함수는 ReLU를, 출력층에서는 Linear함수를 활성화함수로 적용하였다. 실험은 Mountain-Car-v0에 대해 DQN 강화학습을 진행했을 때 각 에피소드별로 획득한 보상 결과를 살펴보고, 보상구간에 포함된 횟수를 분석하였다. 실험결과 전체 100회의 에피소드 중 보상을 50 이상 획득한 에피소드가 85개로 나타났다.

• Fashion & Textile Research Journal
• /
• v.24 no.4
• /
• pp.372-385
• /
• 2022

In this study, the patterns expressed on nineteenth-century blue and white porcelain among Joseon white porcelain were selected as the material for the development of the car seat fabric design. It was intended to be applied to car seat design by incorporating Korea's own traditional patterns to fit modern sensibility. First, seven pieces of nineteenth-century blue and white porcelain were selected through the literature, and motifs were produced using adobe illustrator, a computer graphic program. Seven car seat fabric designs were developed according to the construction method and development method of the produced motif. Work 1 was designed to elicit a soft and feminine atmosphere using the peony pattern shown in Table 1-1. Work 2 aimed to express a luxurious atmosphere using the image of the mountain expressed in Table 1-2 as a design material. Works 3 was designed by freely arranging the letters of luck expressed in Table 1-3 to form a free and dynamic image. Work 4 was intended to express a stable and rhythmic atmosphere by horizontally arranging the images of the gently curved wings, tail, and rhythmical tail feathers of the phoenix expressed in Table 1-4. Work 5 was designed in a vertical arrangement using the patterns and silhouettes of the tiger's back expressed in Table 1-5. Work 6 was designed using the wave pattern expressed in Table 1-6 to replicate the rhythmic atmosphere. Work 7 was designed using the images of rocks, waves, and the sun in Table 1-7 to express a calm and antique atmosphere.

• Journal of KIISE:Software and Applications
• /
• v.31 no.11
• /
• pp.1431-1438
• /
• 2004

Reinforcement learning learns policies for accomplishing a task's goal by experience through interaction between agent and environment. Q-learning, basis algorithm of reinforcement learning, has the problem of curse of dimensionality and slow learning speed in the incipient stage of learning. In order to solve the problems of Q-learning, new function approximation methods suitable for reinforcement learning should be studied. In this paper, to improve these problems, we suggest Fuzzy Q-Map algorithm that is based on online fuzzy clustering. Fuzzy Q-Map is a function approximation method suitable to reinforcement learning that can do on-line teaming and express uncertainty of environment. We made an experiment on the mountain car problem with fuzzy Q-Map, and its results show that learning speed is accelerated in the incipient stage of learning.

• Journal of Korean Society of Rural Planning
• /
• v.20 no.4
• /
• pp.137-143
• /
• 2014

Car exhaust emissions are recognized as one of the key sources for climate change and electric vehicles have no emissions from tailpipe. However, the limited charging infrastructures could restrict the propagation of electric vehicles. The purpose of this study is to find the vulnerable districts limited to the charging station services after meeting the goal of Ministry of Knowledge Economy(12%). We assumed that the charging service can be provided by current gas stations. The range of the vulnerable grades was determined by the accessibility to current gas stations and the vulnerable regions were classified considering the optimal number of charging stations estimated by the efficiency function. We used 4,827 sub-municipal divisions and 11,677 gas station locations for this analysis. The results show that most of mountain areas are vulnerable and the fringe areas of large cities generally get a good grade for the charging infrastructure. The gangwon-do, jeollanam-do, gyeongsangbuk-do, and chungcheongnam-do include more than 40% vulnerable districts.

• Korean Journal of Environment and Ecology
• /
• v.1 no.1
• /
• pp.66-82
• /
• 1987

The user's composition of socia-economic characteristics in Bukhan Mountain National Park showed that male. twenties and students were 65.4 percent. 62.7 percent and 37.4 percent respectively by sex, age and occupation. In visiting purpose, the nature-oriented motive was 67.1 percent of the total. hut the picnic patterns as of neighbourhood park and the recreation patterns as of recreation ground were appeared simultaneously. In preferable place of visitors. the well-known mountain hut, camp sites and summits were prefered. The level of scenic satisfaction was 7.8 and comparatively high on considering the maximum level of 10.0. The level of total satisfaction. however, was no more than 6.3 owing to lack of accomodation facilities along trails, 63.4 percent of visitors were opposed to construction cable-car and visitors were anxious seriously about the nature deterioration. In Jeongnung valley, the number of users was the lagest, so that the maximum number of passangers a day attained to 20.000. The peak seasons of visiting were Spring and Fall, and the peak hours during a day was 10-11 hours A.M. and 3-5 hours P.M. Therefore partitioned spatial management in consideration of hiking pattern of nature park. picnic pattern as of neighbourhood park and recreation pattern as of recreation ground shall be necessary to solve the conflicts among functions.

• Water for future
• /
• v.8 no.2
• /
• pp.30-40
• /
• 1975

Hydrogeological survey and geophysical prospecting have been carried out in Saudi Arabia for the purpose of finding groundwater in the soil and rock at the request of General trading company in Jeddah, Saudi Arabia. The surveyed area is located on $38^{\circ}-39^{\circ}$ 30' in longitude and $26^{\circ}-26^{\circ}$ 30' in latitude. The topography of this area is dominated by northwest southeast mountain range composed mostly of precambrian rocks and basalt of tertiary period. Geology is mainly composed of greenstone, granite, andesite, diorite rhyolite of pre-cambrian era and sandstone of cambrian period which are underlained by basalt and andesite of tertiary period and alluvium of quaternary unconformably. The instruments used in this investigation are TR-18B2 radioactivity unit which isjapanese patented and A.C. Terrameter, a resistivity meter manufactured by ABEM of Stockholm, Sweden. Radioactivity method has been conducted along the Alula-Khaybar road, totally 164Km by the car-borne. As a result of the above survey 16 places have been selected and these anomalies show 1.2N-1.6N compared to background of each area in intensity with width of 10-50m. Resistivity vertical profiling which made use of Schlumberger configuration method has been made over selected areas by radioactivity method to provide hydrogeological information for a water resources survey. The result of resistivity shows that good aquifers are located in the western part of surveyed area where sedimentary rock is distributed. The strata showing 10-50, ${\Omega}-m$ in resistivity are thought to be waterbearing layer. The variations in aquifer resistivity found, are thought to be due to verying clay content, which could be related to aquifer yield. It has proved impossible to detect small salinity variation in the buried aquifer by geophysics. As a result of resistivity prospecting 10 places are recommended to be drilled at the anomalies as shown attached map. yields from the proposed holes have been estimated approximately from $20m^3$ to $200m^3$ per day. Prior to drilling for groundwater, test boring using ${\c}4"$ should be drilled in order to obtain more reliable hydrogeological information for the construction of perfect wells.ells.

• Journal of the Korean Institute of Landscape Architecture
• /
• v.35 no.4
• /
• pp.57-70
• /
• 2007

As a basic element of Jeju landscape, Oreum offers a beautiful and aesthetic view. Considering topographical and geological research achievements, however, an effort to discover implicit value in terms of landscape characteristics and value has been ignored. This paper has investigated the characteristics and value of landscape by Oreum focusing on Jeju landscape characteristics and eco-touristic value and discussed a scheme to maximize the values. Under a theme of 'Sustainable Development' of the RIO Declaration, tour industry has recently changed its focus from eco-tourism to gee-tourism. Fortunately, Jeju Oreum has very distinctive and unique landscape with depressed crater at a crest. Nevertheless, it's very difficult to see a true aspect of Oreum from the street or over the car window. Therefore, it's urgent to begin a research on how to make advantage of and preserve Oreum landscape in order to maximize its landscape values and improve its potential as a tourist attraction. Through diverse programs such as sky leisure sports(ex: light airplane and helicopter riding, paragliding), sky watching, and mountain hiking, in particular, a possibility that Oreum can succeed as LBD(Learning by Doing)-based tour program with volcanic features needs to be examined. Besides, it's also a good idea to develop Oreum tour program or Oreum Museum as an alternative plan. Above all, however, it's most urgent to protect the existing Oreum and restore ecological and landscape beauty of Oreum through proper land use.

• The KIPS Transactions:PartB
• /
• v.10B no.6
• /
• pp.587-592
• /
• 2003

Many real world control problems have continuous states and actions. When the state space is continuous, the reinforcement learning problems involve very large state space and suffer from memory and time for learning all individual state-action values. These problems need function approximators that reason action about new state from previously experienced states. We introduce Fuzzy Q-Map that is a function approximators for 1 - step Q-learning and is based on fuzzy clustering. Fuzzy Q-Map groups similar states and chooses an action and refers Q value according to membership degree. The centroid and Q value of winner cluster is updated using membership degree and TD(Temporal Difference) error. We applied Fuzzy Q-Map to the mountain car problem and acquired accelerated learning speed.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.33 no.7
• /
• pp.51-58
• /
• 2005

In 1990, GPS which had been developed for the military purposes became available to the civilian community. Since then these satellite navigation systems have been used extensively in the industrial areas such as car navigation, airplanes, communications, science and surveying. If we are dependent on GPS, however, there are some foreseeable problems in the areas of national security and sovereignty. Current GPS satellite constellation provides limited performance for the country like Korea and Japan where mountain area and urban canyon do not allow the wide skyline. To solve these problems, many countries plan to make other alternative navigation systems.In this paper, RNSS(Regional Navigation Satellite System) is designed to provide communication service with high elevation angle. It is shown, that system does not only have good navigation performance, but also improve GPS performance in Korea and its neighboring areas.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.13 no.6
• /
• pp.635-642
• /
• 2003

Reinforcement learning got successful results in a lot of applications such as control and scheduling. Various function approximation methods have been studied in order to improve the learning speed and to solve the shortage of storage in the standard reinforcement learning algorithm of Q-Learning. Most function approximation methods remove some special quality of reinforcement learning and need prior knowledge and preprocessing. Fuzzy Q-Learning needs preprocessing to define fuzzy variables and Local Weighted Regression uses training examples. In this paper, we propose a function approximation method, Fuzzy Q-Map that is based on on-line fuzzy clustering. Fuzzy Q-Map classifies a query state and predicts a suitable action according to the membership degree. We applied the Fuzzy Q-Map, CMAC and LWR to the mountain car problem. Fuzzy Q-Map reached the optimal prediction rate faster than CMAC and the lower prediction rate was seen than LWR that uses training example.