• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2013.05a
• /
• pp.867-868
• /
• 2013

• Proceedings of the Korea Information Processing Society Conference
• /
• 2015.04a
• /
• pp.822-823
• /
• 2015

본 논문에서는 휴머노이드 로봇이 더욱 지능적으로 보행할 수 있도록 강화학습을 적용하는 방법을 제안한다. 강화학습을 활용하면 로봇의 이동 경로에 동적으로 이동하는 장애물이 있을 경우 이를 인지하고 회피하여 목적지까지 문제 없이 이동할 수 있다. 실제 환경에서의 실험에 앞서 Unity3D를 활용한 가상 환경에서 이를 구현하여 실험해보았다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2022.05a
• /
• pp.428-429
• /
• 2022

최근에 자율 주행 자동차에 관련한 관심이 증가하면서 다양한 연구들이 도출되고 있다. 특히, 자율 주행 자동차를 시뮬레이터에서 검증하는 방법은 실 환경과 비교할 때 상대적으로 안전한 성능 검증 방법으로 많이 활용되고 있다. 시뮬레이터의 핵심 기술은 실 환경과 가상 시뮬레이션 환경의 차이를 줄이는 데 있다. 본 논문에서는 Generative Adversarial Imitation Learning(GAIL)[1] 기반으로 자율 주행 자동차 시뮬레이터 내에서 다수의 가상 동적 객체들의 움직임을 제어하는 방법을 제안한다. GAIL은 생성기와 판별기로 구성된다. 생성기는 강화학습 정책 생성기와 전문가 정책 생성기를 포함한다. 판별기는 보상 학습기를 포함한다. GAIL 기반으로 가상 자동차 및 가상 보행자를 제어함으로써 동영상에서의 이동경로를 학습해서 표현할 수 있다.

• The Journal of the Korea Contents Association
• /
• v.12 no.8
• /
• pp.274-282
• /
• 2012

This study conducted the following experiment to examine change of physicopsychological function on lumbar stabilization exercise(LSE) and virtual reality game training(Nintendo Wii Sport-NWS) to stroke patients subject for fall prevention. Psychological function was measured by falls efficacy with stroke patients and physical function was measured by static and dynamic balance on comparative analysis of pre, post exercise and each groups in 30 stroke patient subject. Static balance was measured by BBS, FRT, dynamic were measured by TUG, 10m walking test and falls efficacy with stroke patients was measured index of falls efficacy. These result lead us to the conclusion that each group were statistically improved at all physicopsychological test, but BBS, FRT, 10m walking test were more statistically improved at LSE group and falls efficacy with stroke patients were more improved at virtual reality game training group. Consequently, virtual reality game training would be lead to positive increment of physicopsychological function on stroke patient.

• Journal of rehabilitation welfare engineering & assistive technology
• /
• v.10 no.1
• /
• pp.47-57
• /
• 2016

In this paper, we measured human body signals in order to verify a active harness system that we developed for gait and balance training. The experimental procedure was validated by tests with 20 healthy male subjects. They conducted motions of Activities of Daily Living(ADL)(Normal Walking, Stand-to-Sit, Sit-to-Stand, Stair Walking Up, and Stair Walking Down) according to body weight support rates (0%, 30%, 50% of subjects' body weight). The effectiveness of the active harness system is verified by using the results of foot pressure distribution. In normal walking, the decrease of fore-foot pressure, lateral soleus muscle and biceps femoris muscle were remarkable. The result of stand-to-sit results motion indicated that the rear-foot pressure and tibialis anterior muscle activities exceptionally decreased according to body weight support. The stair walking down show the marked drop of fore-foot pressure and rectus femoris muscle activities. The sit-to-stand and stair walking up activities were inadequate about the effect of body weight support because the velocity of body weight support system was slower than male's activity.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2005.06a
• /
• pp.1780-1783
• /
• 2005

In this paper ZMP was considered in order to get a walking stability, so the gait in the stable domain was realized through putting the stability margin in the sole domain of a foot. It is assumed that the robot's legs have 12 joints to operate a open-loop drive and there was no external disturbance under walking phases, additionally, the robot is walking on the flat plane. It was observed that the robot's walking trajectory, locus of COM and ZMP after imposing the motion to each joint. For realizing the simulation considering ZMP and movement of mass center, it was checked if it is stable for the constraint robot model to walk in stability and the feasibility was estimated about its dynamic gait. Eventually it was shown that a constraint gait algorithm is able to realize. To verify the proper walking process, ZMP(Zero Moment Point) theory is applied and the simulation has been done by ADAMS.

• Journal of the Korean Institute of Illuminating and Electrical Installation Engineers
• /
• v.19 no.7
• /
• pp.73-80
• /
• 2005

Biped robot requires that an energy source and a control part should be installed on the body to realize active system. So, we choose the DC motor having high torque in compact size in this study. In the DC motor serve system, we choose power amplifier with analog configuration, developed the module combined the controller and the driver. We applied this module to robot actuator and studied the response characteristics in an action and a return. Main controller with serve system, loading PIC micro controller, can be load on the robot with light weight.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.23 no.4
• /
• pp.360-365
• /
• 2013

The paper introduces dynamic generation technique of foot trajectories using CPG(Central Pattern Generator). In this approach, the generated foot trajectories can be changeable according to variable outputs of CPG in various environments, because they are given as mapping functions of the output signals of the CPG oscillators. It enables to provide an adaptable foot trajectory for environmental change. To demonstrate the effectiveness of the proposed approach, experiments on humanoid robot Nao is executed in the Webot simulation. The performance and motion features of CPG based approach is analyzed.

• Journal of the Korean Society of Industry Convergence
• /
• v.17 no.4
• /
• pp.245-254
• /
• 2014

In this paper, stable and robust dynamic walking for a biped motion is proposed. To success this objective, the following structures are processed. In this paper, the proposed control method is one that adjusts actual zero moment position to move to the closest possible point in the stable area instead of following desired zero moment position. This minimizes energy consumption with the smallest joint movements. The proposed control method makes mechanical energy that drives lower limb of the bipedal robot efficient. In this paper, walking experiment is carried out with the three control structures mentioned above. The trajectory generated by off-line is illustrated by performing to walking on flat ground. experiment with an obstacle whose height is lower than that of trajectory is executed to validate dynamic motion.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2018.05a
• /
• pp.543-544
• /
• 2018

가상 현실 네비게이션을 위한 전방향 트레드밀은 사용자가 걷거나 달리면서 물리적으로 고정된 공간 내에 사용자를 유지할 수 있도록 지면 모션을 시뮬레이션하는 장비이다. 이러한 트레드밀 시스템의 성능이나 안정성을 정량적으로 측정하거나 분석하기가 어렵기 때문에 이전의 연구에서는 주관적 설문 조사와 같은 정성적 분석 방법을 사용하였다. 본 연구에서는 인간의 보행 경로와 유사한 궤도를 따라 움직이는 무인 차량 시스템을 이용한 새로운 정량적 데이터 측정 방법을 제안한다. 무인 차량 시스템은 미리 정의 된 인간의 보행 동작을 시뮬레이션하고 트레드밀 시스템에 대한 제어 입력을 제공하며, 다축 가속 및 방향과 같은 차량의 동적 데이터를 측정 할 수 있다. 또한 이 데이터는 평상시의 정기 또는 다른 제어 알고리즘과의 비교를 수행할 수 있다. 본 연구에서는 궤적 시뮬레이션 모듈, 데이터 수집 모듈, 성능 평가 모듈 등 전방향 트레드밀에 대한 정량 분석 방법의 설계 구조 및 초기구현 결과를 제시하고자 한다.