• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2022년도 추계학술발표대회
• /
• pp.535-537
• /
• 2022

AI 가속기는 머신 러닝 및 딥 러닝을 포함한 인공 지능 및 기계 학습 응용 프로그램의 연산을 더 빠르게 수행하도록 설계된 일종의 하드웨어 가속기 또는 컴퓨터 시스템이다. 가속기를 설계하기 위해선 설계 영역 탐색(Design Space Exploration)을 하여야 하고 여러 인공지능 중에서도 합성 곱 신경망(CNN)에 대한 설계 영역 탐색을 소개한다.

• Journal of Astronomy and Space Sciences
• /
• 제32권3호
• /
• pp.257-268
• /
• 2015

Characteristics of delta-V requirements for deploying an impactor from a mother-ship at different orbital altitudes are analyzed in order to prepare for a future lunar CubeSat impactor mission. A mother-ship is assumed to be orbiting the moon with a circular orbit at a 90 deg inclination and having 50, 100, 150, 200 km altitudes. Critical design parameters that are directly related to the success of the impactor mission are also analyzed including deploy directions, CubeSat flight time, impact velocity, and associated impact angles. Based on derived delta-V requirements, required thruster burn time and fuel mass are analyzed by adapting four different miniaturized commercial onboard thrusters currently developed for CubeSat applications. As a result, CubeSat impact trajectories as well as thruster burn characteristics deployed at different orbital altitudes are found to satisfy the mission objectives. It is concluded that thrust burn time should considered as the more critical design parameter than the required fuel mass when deducing the onboard propulsion system requirements. Results provided through this work will be helpful in further detailed system definition and design activities for future lunar missions with a CubeSat-based payload.

• 교육시설
• /
• 제26권5호
• /
• pp.26-31
• /
• 2019

• 로봇학회논문지
• /
• 제17권1호
• /
• pp.40-47
• /
• 2022

A reward function suitable for a task is required to manipulate objects through reinforcement learning. However, it is difficult to design the reward function if the ample information of the objects cannot be obtained. In this study, a demonstration-based object manipulation algorithm called stochastic exploration guided by demonstration (SEGD) is proposed to solve the design problem of the reward function. SEGD is a reinforcement learning algorithm in which a sparse reward explorer (SRE) and an interpolated policy using demonstration (IPD) are added to soft actor-critic (SAC). SRE ensures the training of the critic of SAC by collecting prior data and IPD limits the exploration space by making SEGD's action similar to the expert's action. Through these two algorithms, the SEGD can learn only with the sparse reward of the task without designing the reward function. In order to verify the SEGD, experiments were conducted for three tasks. SEGD showed its effectiveness by showing success rates of more than 96.5% in these experiments.

• 한국정보과학회논문지:컴퓨팅의 실제 및 레터
• /
• 제13권4호
• /
• pp.231-239
• /
• 2007

본 논문에서는 국내 표준 128비트 블록 암호화 알고리즘인 SEED를 하드웨어로 설계할 경우 면적-성능간의 trade-off 관계를 보여준다. 본 논문에서 다음 4가지 유형의 설계 구조를 비교한다. (1) Design 1 : 16 라운드 완전 파이프라인 방식, (2) Design 2 : 단일 라운드의 반복 사용 방식 (3) Design 3 : G 함수 공유 및 반복 사용 방식 (4) Design 4 : 단일 라운드 내부 파이프라인 방식. (1),(2),(3)의 방식은 기존의 논문들에서 제안한 각기 다른 설계 방식이며 (4)번 설계 방식이 본 논문에서 새롭게 제안한 설계 방식이다. 본 논문에서 새롭게 제안한 방식은, F 함수 내의 G 함수들을 파이프라인 방식으로 연결하여 면적 요구량을 (2)번에 비해서 늘이지 않으면서도 파이프라인과 공유블록 사용의 효과로 성능을 Design 2와 Design 3보다 높인 설계 방식이다. 본 논문에서 4가지 각기 다른 방식을 각각 실제 하드웨어로 설계하고 FPGA로 구현하여 성능 및 면적 요구량을 비교 분석한다. 실험 분석 결과, 본 논문에서 새로 제안한 F 함수 내부 3단 파이프라인 방식이 Design 1 방식을 제외하고 가장 throughput 이 높다. 제안된 Design 4 가 단위 면적당 출력성능(throughput)면에서 다른 모든 설계 방식에 비해서 최대 2.8배 우수하다. 따라서, 새로이 제안된 SEED 설계가 기존의 설계 방식들에 비해서 면적대비 성능이 가장 효율적이라고 할 수 있다.

• Journal of Astronomy and Space Sciences
• /
• 제31권4호
• /
• pp.347-351
• /
• 2014

Planetary exploration rovers are likely to make a trip on a winding and sloping road of irregular surfaces to the destination in order to accomplish scientific missions. One of the key technologies for rovers is a suspension for traveling and performing exploration missions; the suspension is an essential area of technology for a stable movement of a rover. In this study, an 8-wheel suspension is designed to enable efficient climbing of slopes on a passage to the destination. For the two front wheels among the eight wheels, the moment at the pivot connecting two wheels is derived when the distance between the wheels and the torque of wheels are same. A test experiment was performed to compare the magnitude of moment according to the change in tilt angle and the position of the pivot. Finally, a suspension design considering the position of the pivot was proposed to enhance the hill-climbing performance.

• 한국추진공학회:학술대회논문집
• /
• 한국추진공학회 2008년 영문 학술대회
• /
• pp.109-114
• /
• 2008

This paper presents the development status of a subscale precooled turbojet engine "S-engine" for the hypersonic cruiser and space place. S-engine employs the precooled-cycle using liquid hydrogen as fuel and coolant. It has $23cm{\times}23cm$ of rectangular cross section, 2.6 m of the overall length and about 100 kg of the target weight employing composite materials for a variable-geometry rectangular air-intake and nozzle. The design thrust and specific impulse at sea-level-static(SLS) are 1.2 kN and 2,000 sec respectively. After the system design and component tests, a prototype engine made of metal was manufactured and provided for the system firing test using gaseous hydrogen in March 2007. The core engine performance could be verified in this test. The second firing test using liquid hydrogen was conducted in October 2007. The engine, fuel supplying system and control system for the next flight test were used in this test. We verified the engine start-up sequence, compressor-turbine matching and performance of system and components. A flight test of S-engine is to be conducted by the Balloon-based Operation Vehicle(BOV) at Taiki town in Hokkaido in October 2008. The vehicle is about 5 m in length, 0.55 m in diameter and 500 kg in weight. The vehicle is dropped from an altitude of 40 km by a high-altitude observation balloon. After 40 second free-fall, the vehicle pulls up and S-engine operates for 60 seconds up to Mach 2. High altitude tests of the engine components corresponding to the BOV flight condition are also conducted.

• 디지털산업정보학회논문지
• /
• 제7권1호
• /
• pp.63-74
• /
• 2011

LED design works make user experience the conditions of time and space, and enable maximized spatial exploration and pursuit of ultimate sources. Light itself includes an ample dimension of time and, through the process of on and off, time-duration, similarity, and discontinuation is reorganized. LED design works make people experience the conditions of time and space, and enable maximized spatial exploration and pursuit of ultimate sources. LED design have been approached more from the viewpoint of material than from the viewpoint of aesthetics. LED based environmental design will be embedded deeply into our life with the technology of ubiquitous city. As the problem of the light intensity, tempo of change LED motion graphic and colors, substitutability is almost completely resolved now, LED is have more potentials than any other material as a tool of motion graphic. These result well make that using of LED is could be better useful from now on. In edition, users could be offered better beneficial lighting environment, and they will enjoy their better lives in the future.

• International Journal of Fluid Machinery and Systems
• /
• 제10권3호
• /
• pp.240-253
• /
• 2017

Latin hypercube sampling is widely used design-of-experiment technique to select design points for simulation which are then used to construct a surrogate model. The exploration/exploitation properties of surrogate models depend on the size and distribution of design points in the chosen design space. The present study aimed at evaluating the performance characteristics of various surrogate models depending on the Latin hypercube sampling (LHS) procedure (sample size and spatial distribution) for a diverse set of optimization problems. The analysis was carried out for two types of problems: (1) thermal-fluid design problems (optimizations of convergent-divergent micromixer coupled with pulsatile flow and boot-shaped ribs), and (2) analytical test functions (six-hump camel back, Branin-Hoo, Hartman 3, and Hartman 6 functions). The three surrogate models, namely, response surface approximation, Kriging, and radial basis neural networks were tested. The important findings are illustrated using Box-plots. The surrogate models were analyzed in terms of global exploration (accuracy over the domain space) and local exploitation (ease of finding the global optimum point). Radial basis neural networks showed the best overall performance in global exploration characteristics as well as tendency to find the approximate optimal solution for the majority of tested problems. To build a surrogate model, it is recommended to use an initial sample size equal to 15 times the number of design variables. The study will provide useful guidelines on the effect of initial sample size and distribution on surrogate construction and subsequent optimization using LHS sampling plan.

• 천문학회보
• /
• 제34권1호
• /
• pp.119.1-119.1
• /
• 2009