• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2007.04a
• /
• pp.253-256
• /
• 2007

본 논문은 로봇이 영상을 통해 획득한 특정물체의 기하학적 정보를 이용하여 이족로봇이 안정적으로 보행할 수 있게 하기 위한 균형제어기법을 제안한다. 영상은 핀 홀 카메라 모델을 통해 획득하였으며, 영상에 포함되는 특정물체의 특징성분에 대한 변위와 로봇의 자세와의 상관관계는 핀 홀 카메라 모델을 이용하여 공간좌표계의 특징정보를 평면좌표계의 영상정보에 매칭시킨 후, 특징들의 변위에 따른 로봇 관절 좌표계의 변위를 추정하는 방법으로 구할 수 있었다. 본 논문에서 제안하는 균형제어기법은 별도의 센서없이 카메라만을 이용하여 이족보행 로봇의 균형제어가 가능하다는 장점을 가지며, 소형이족로붓을 이용한 실험을 통해 그 효율성을 검증하였다.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2002.05a
• /
• pp.446-449
• /
• 2002

Quadruped walking robot requires dynamic control to keep its stability in high speed walking. To keep its walking stability by control of only legs' Joint angle lowers energy efficiency. It is known that an animal or a human use the moving of the mass center of one's upper body to keep the stability. We have developed a quadruped walking robot with weight balancing oscillator that have high energy efficiency. In this study, walking tests are performed for the robot to verify the validity of the weight balancing oscillator.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2004.10a
• /
• pp.948-953
• /
• 2004

For stable walking of various biped walking robots(BWR), we need to know the kinematics, dynamics and the Zero Moment of Point(ZMP) which are not easy to analyze analytically. In this reason, we developed a simulation program for BWRs composed of 4 degree-of-freedom upper-part body and 12 degree-of-freedom lower-part of the body. To operate the motion simulator for analyzing the kinematics and dynamics of BWES, inputs for the distance between legs, base angle, choice of walking type, gaits, and walking velocity are necessary. As a result, if stability condition is satisfied by the simulation, angle data for each actuator are generated automatically, and the data are transmitted to BWRS and then, they are actuated by the motion data. Finally, we validate the performance of the proposed motion simulator by applying it to a constructed small sized BWR.

• Proceedings of the KIEE Conference
• /
• 2006.10c
• /
• pp.205-207
• /
• 2006

Since a humanoid robot inherently suffers from instability and always risks tipping itself over, or topping to the ground, it is necessary to ensure high stability and reliability of walk. An unexpected ground condition is one of the principal factors of instability. This paper proposes a walk stabilization method consisting of a Fuzzy algorithm and geometry under uneven terrain. The ground reaction forces that are measured by the FSR sensors on the sole are used to check the ground condition and the robot posture. The effectiveness of proposed method is verified by computer simulations.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.30 no.6 s.249
• /
• pp.643-652
• /
• 2006

This paper is concerned with computer simulations of a biped robot walking in static and dynamic gaits. To this end, a three-dimensional robot is considered possessing a torso and two identical legs of a typical design. For such limbs, a set of inverse kinematic solutions is analytically derived between the torso and the feet. Specific walking patterns are off-line generated meeting stability based on the VPCG or ZMP condition. Subsequently, to verify whether the robot can walk as planned in the presence of mass and ground effects, a multi-body dynamics CAE code has been applied to the resulting joint motions determined by inverse kinematics. As a result, the key parameters to successful gaits could be identified including inherent characteristics as well. Upon comparisons between the two types of gaits, dynamic gaits are concluded more desirable for larger humaniods.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2016.04a
• /
• pp.851-853
• /
• 2016

본 논문은 물류 작업을 위한 지상 보행이 가능한 드론의 제어기를 개선시키기 위한 방법 중 하나로 비선형 자세제어기를 적용한 결과를 보여준다. 로봇은 물류 작업을 위하여 설계되었는데, 페이로드의 변화에 따라서 모델 파라미터가 변하기 때문에 특별한 제어기가 필요하다. 따라서 기존 연구에서 제안되었던 비선형 적응제어기를 채택하여 실험 모델에 적용시켰다. 실험은 실내에서 진행되었고 초기 조건에서 30초 간 자세 안정화 명령의 응답 특성을 확인하는 방식으로 진행되었고, 제어기의 실효성과 안정성을 입증했다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2022.11a
• /
• pp.740-742
• /
• 2022

본 프로젝트에서는 다양한 지형에 구애받지 않고 전천후로 활동할 수 있는 로봇을 구현하기 위해 바퀴형 로봇 보다는 4족 보행 로봇을 채택하여 지형 극복에 더 유리하고 안정적인 자세 제어와 보행을 할 수 있는 동시에 LiDAR 센서와 카메라 모듈을 이용한 SLAM(동시적 위치 추정 및 지도작성)과 원격으로 사물과 사람들을 파악할 수 있는 원격조종 탐사로봇을 개발하고자 한다.

• Journal of Platform Technology
• /
• v.6 no.3
• /
• pp.23-30
• /
• 2018

A motion pattern generation is a process of calculating a certain stable motion trajectory for stably operating a certain motion. A motion control is to make a posture of a robot stable by eliminating occurring disturbances while a robot is in operation using a pre-generated motion pattern. In this paper, a general method of motion pattern generation for a biped walking robot using universal approximator, learning neural networks, is proposed. Existing techniques are numerical methods using recursive computation and approximating methods which generate an approximation of a motion pattern by simplifying a robot's upper body structure. In near future other approaches for the motion pattern generations will be applied and compared as to be done.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.46 no.4
• /
• pp.97-107
• /
• 2009

In image-based gait recognition systems, physical factors, such as the camera angle and the lens distortion, and environmental factors such as illumination determines the performance of recognition. In this paper we present a robust gait recognition method by compensating various types of image distortions. The proposed method is compared with existing gait recognition algorithm with consideration of both physical and environmental distortion factors in the input image. More specifically, we first present an efficient compensation algorithm of image distortion by using the projective transform, and test the feasibility of the proposed algorithm by comparing the recognition performances with and without the compensation process. Proposed method gives universal gait data which is invariant to both distance and environment. Gained data improved gait recognition rate about 41.5% in indoor image and about 55.5% in outdoor image. Proposed method can be used effectively in database(DB) construction, searching and tracking of specific objects.

• Journal of Internet Computing and Services
• /
• v.20 no.4
• /
• pp.91-102
• /
• 2019

In autonomous driving systems, the ability to classify pedestrians in images captured by cameras is very important for pedestrian safety. In the past, after extracting features of pedestrians with HOG(Histogram of Oriented Gradients) or SIFT(Scale-Invariant Feature Transform), people classified them using SVM(Support Vector Machine). However, extracting pedestrian characteristics in such a handcrafted manner has many limitations. Therefore, this paper proposes a method to classify pedestrians reliably and effectively using CNN's(Convolutional Neural Network) deep features and transfer learning. We have experimented with both the fixed feature extractor and the fine-tuning methods, which are two representative transfer learning techniques. Particularly, in the fine-tuning method, we have added a new scheme, called M-Fine(Modified Fine-tuning), which divideslayers into transferred parts and non-transferred parts in three different sizes, and adjusts weights only for layers belonging to non-transferred parts. Experiments on INRIA Person data set with five CNN models(VGGNet, DenseNet, Inception V3, Xception, and MobileNet) showed that CNN's deep features perform better than handcrafted features such as HOG and SIFT, and that the accuracy of Xception (threshold = 0.5) isthe highest at 99.61%. MobileNet, which achieved similar performance to Xception and learned 80% fewer parameters, was the best in terms of efficiency. Among the three transfer learning schemes tested above, the performance of the fine-tuning method was the best. The performance of the M-Fine method was comparable to or slightly lower than that of the fine-tuningmethod, but higher than that of the fixed feature extractor method.