• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.5
• /
• pp.179-186
• /
• 2021

Manipulation of complex objects with an anthropomorphic robot hand like a human hand is a challenge in the human-centric environment. In order to train the anthropomorphic robot hand which has a high degree of freedom (DoF), human demonstration augmented deep reinforcement learning policy optimization methods have been proposed. In this work, we first demonstrate augmentation of human demonstration in deep reinforcement learning (DRL) is effective for object manipulation by comparing the performance of the augmentation-free Natural Policy Gradient (NPG) and Demonstration Augmented NPG (DA-NPG). Then three DRL policy optimization methods, namely NPG, Trust Region Policy Optimization (TRPO), and Proximal Policy Optimization (PPO), have been evaluated with DA (i.e., DA-NPG, DA-TRPO, and DA-PPO) and without DA by manipulating six objects such as apple, banana, bottle, light bulb, camera, and hammer. The results show that DA-NPG achieved the average success rate of 99.33% whereas NPG only achieved 60%. In addition, DA-NPG succeeded grasping all six objects while DA-TRPO and DA-PPO failed to grasp some objects and showed unstable performances.

• Annual Conference of KIPS
• /
• 2020.11a
• /
• pp.858-861
• /
• 2020

로봇이 사람과 같이 다양하고 복잡한 사물 조작을 하기 위해서 휴먼형 로봇손의 사물 파지 작업이 필수적이다. 자유도 (Degree of Freedom, DoF)가 높은 휴먼형(anthropomorphic) 로봇손을 학습시키기 위하여 사람 데모(human demonstration)가 결합된 강화학습 최적화 방법이 제안되었다. 본 연구에서는 강화학습 최적화 방법에 사람 데모가 결합된 Demonstration Augmented Natural Policy Gradient(DA-NPG)와 NPG 의 성능 비교를 통하여 행동 복제의 효율성을 확인하고, DA-NPG, DA-Trust Region Policy Optimization (DA-TRPO), DA-Proximal Policy Optimization (DA-PPO)의 최적화 방법의 성능 평가를 위하여 6 종의 물체에 대한 휴먼형 로봇손의 사물 조작 작업을 수행한다. 그 결과, DA-NPG 와 NPG를 비교한 결과를 통해 휴먼형 로봇손의 사물 조작 강화학습에 행동 복제가 효율적임을 증명하였다. 또한, DA-NPG 는 DA-TRPO 와 유사한 성능을 보이면서 모든 물체에 대한 사물 파지에 성공하여 가장 안정적이었다. 반면, DA-TRPO 와 DA-PPO 는 사물 조작에 실패한 물체가 존재하여 불안정한 성능을 보였다. 본 연구에서 제안하는 방법은 향후 실제 휴먼형 로봇에 적용하여 휴먼형 로봇 손의 사물조작 지능 개발에 유용할 것으로 전망된다.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.17 no.3
• /
• pp.537-542
• /
• 2022

There is a limit to the current indoor air quality (IAQ) monitoring method using fixed sensors and devices. A mobile robot for IAQ monitoring was developed by mounting IAQ monitoring sensors on a small multi-legged robot to minimize vibration and protect the sensors from vibration while robot moves. The developed mobile robot used a simple gait mechanism to enable the robot to move forward, backward, and turns only with the combination of forward and reverse rotation of the two DC motors. Due to the simple gait mechanism, not only IAQ data measurements but also gait motion control were processed using a single Arduino board. Because the mobile robot has small number of electronic components and low power consumption, a relatively low-capacity battery was mounted on the robot to reduce the weight of the battery. The weight of mobile robot is 1.4kg including links, various IAQ sensors, motors, and battery. The gait and turning speed of the mobile robot was measured at 3.75 cm/sec and 14.13 rad/sec. The maximum height where the robot leg could reach was 33 mm, but the mobile robot was able to overcome the bumps up to 24 mm.

• Journal of the KSME
• /
• v.34 no.10
• /
• pp.787-795
• /
• 1994

이상과 같이 소개한 KIST 휴먼 로봇의 4족 보행시스템을 개발하는 데에 있어서 최대 목표는 보행제어의 지능화를 통해서 외관상 유사한 히타치사의 극한작업 로봇과 차별화시키는 것이다. 이를 위해서는 성능이 급격히 향상되어가고 있는 최신의 컴퓨터와 센서들을 적극 활용하면서 국내외 관련 전문가들이 축적하고 있는 노우 하우들을 최대한 활용하는 것이 필요하고, 기술적 으로는 걸음새의 지능제어, 반사제어, 소비동력의 최소화 등을 집중적으로 해결하는 것이 요구 된다. 휴먼 로봇 보행기구의 연구 성과는 궁극적으로 계단 등과 같은 단순 장애물은 물론 얼마나 악조건의 험로 위 자율 보행이 가능하고, 평제에서는 얼마나 빠른 최대 보행속도를 발휘할 수 있으며, 로봇에 탑재된 축전지만으로 얼마나 먼거리를 보행할 수 있는 지에 의해 그 질적 수 준이 판가름 나게 될 것이다. 이러한 점에서 이 과제의 추진은 국내의 로봇 연구 분야에 있어서 전혀 새로운 매우 거대한 도전이라고 할 수 있다.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.17 no.5
• /
• pp.941-948
• /
• 2022

Disaster response robots are deployed to disaster sites where human access is difficult and dangerous. The disaster response robots explore the disaster sites prevent a structural collapse and perform lifesaving to minimize damage. It is difficult to operate robots in the disaster sites due to rough terrains where various obstacles are scattered, communication failures and invisible environments. In this paper, we developed a series connectable wheeled robot module. The series connectable wheeled robot module was developed into two types: an active driven robot module and a passive driven robot module. A wheeled robot was built by connecting the two active type robot modules and one passive type robot module. Two robot modules were connected by one DoF rotating joint, allowing the wheeled robot to avoid obstructions in a vertical direction. The wheeled robot performed driving and obstacle avoidance using only pressure sensors, which allows the wheeled robot operate in the invisible environment. An obstacle avoidance experiment was conducted to evaluate the performance of the wheeled robot consisting of two active driven wheeled robot modules and one passive driven wheeled robot module. The wheeled robot successfully avoided step-shaped obstacles with a maximum height of 80 mm in a time of 24.5 seconds using only a pressure sensors, which confirms that the wheeled robot possible to perform the driving and the obstacle avoidance in invisible environment.

• The Journal of the Korea Contents Association
• /
• v.18 no.3
• /
• pp.592-600
• /
• 2018

As the field of robots is evolving to intelligent robots that can replace even humans' mental or emotional labor, 'robot ethics' needed in relationship between humans and robots is becoming a crucial issue these days. The purpose of this study is to consider the ethics of robots and humans that is essential in this post-human age. It will deal with the followings as the main contents. First, with the cases of developing ethics software intended to make robots practice ethics, the authors begin this research being conscious about the matter of whether robots can really judge what is right or wrong only with the ethics codes entered forcibly. Second, regarding robot ethics, we should consider unethicality that might arise from learning data internalizing human biasness and also reflect ethical differences between countries or between cultures, that is, ethical relativism. Third, robot ethics should not be just about ethics codes intended for robots but reflect the new concept of 'human ethics' that allows humans and robots to coevolve.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.17 no.6
• /
• pp.1137-1144
• /
• 2022

In this study, an artificial intelligence(AI) was developed to help with facial expression practice in order to express emotions. The developed AI used multimodal inputs consisting of sentences and facial images for deep neural networks (DNNs). The DNNs calculated similarities between the emotions predicted by the sentences and the emotions predicted by facial images. The user practiced facial expressions based on the situation given by sentences, and the AI provided the user with numerical feedback based on the similarity between the emotion predicted by sentence and the emotion predicted by facial expression. ResNet34 structure was trained on FER2013 public data to predict emotions from facial images. To predict emotions in sentences, KoBERT model was trained in transfer learning manner using the conversational speech dataset for emotion classification opened to the public by AIHub. The DNN that predicts emotions from the facial images demonstrated 65% accuracy, which is comparable to human emotional classification ability. The DNN that predicts emotions from the sentences achieved 90% accuracy. The performance of the developed AI was evaluated through experiments with changing facial expressions in which an ordinary person was participated.

• Proceedings of the Korean Information Science Society Conference
• /
• 2006.10b
• /
• pp.421-425
• /
• 2006

최근 들어서 휴머노이드 로봇을 비롯한 로봇에 대하여 관심이 증대되고 있다. 이에 따라, 외모를 닮은 로봇 뿐 만 아니라, 사람과 상호 작용을 할 수 있는 로봇 기술의 중요성이 부각되고 있다. 이러한 상호 작용을 위한 효율적이고, 가장 자연스러운 방법 중의 하나가 비전을 기반으로 한 제스처 인식이다. 제스처를 인식하는데 있어서 가장 중요한 것은 손의 모양과 움직임을 인식하는3차원 제스처 인식이다. 본 논문에서는 3차원 손 제스처를 인식하기 위하여3차원 손 모델 추정 방법과 명령형 제스처 인식 시스템을 소개하고, 수화, 지화 등으로의 확장성을 위한 프레임워크를 제안한다.

• Journal of the Korean Society for Precision Engineering
• /
• v.23 no.7 s.184
• /
• pp.67-75
• /
• 2006

One of challenging topics for humanoid hands is to modulate a human-like motion of humanoid fingers handling an object. To this end, recognizing the motion behavior of human fingers is very important aspect. Based on this concept, this paper identifies the .joint trajectories of human fingers for an operation of hand opening and closing, and specifies an empirical model that coordinates an inter-articular relationship of human fingers doing the given motion. It is expected that the inter-articular model presented in this paper is applicable for humanoid fingers to mimic the natural motion of human fingers.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2021.07a
• /
• pp.433-434
• /
• 2021

본 논문에서는 RPA 도구 개발 과정 중 자동화 및 데이터 전처리 기능을 이용한 양질의 데이터 추출 과정을 기술한다. 개발된 RPA 도구에서는 녹화 기능으로 자동화 기능을 향상시켰으며, 비대칭 데이터 변환 기능과 이상치 처리 기능을 통해 업무 생산효율 증가 및 휴먼에러 방지를 제공한다.