http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
오성회(Songhwai Oh) 大韓電子工學會 2012 電子工學會論文誌-SC (System and control) Vol.49 No.3
데이터 연관은 지능시스템의 자율적인 작동에 매우 중요한 문제이다. 본 논문에서는 데이터 연관 문제를 Bayesian 방식으로 구성하고 이를 성공적으로 지능시스템에 응용한 예를 설명한다. 먼저 데이터 연관 문제가 어떻게 Bayesian 방식으로 구성하여 혼잡한 환경에서의 다 물체 추적 문제에 적용되는지 알아본다. 그리고 데이터 연관이 지능시스템에 어떻게 응용될 수 있는지 정체 관리를 이용한 항공 교통 관제, 카메라 네트워크 위치 및 관점 자동 보정, 멀티 센서 퓨젼의 세 가지 예를 이용해 살펴본다. Data association plays an important role in intelligent systems. This paper presents the Bayesian formulation of data association and its applications to intelligent systems. We first describe the Bayesian formulation of data association developed for solving multi-target tracking problems in a cluttered environment. Then we review applications of data association in intelligent systems, including surveillance using wireless sensor networks, identity management for air traffic control, camera network localization, and multi-sensor fusion.
Robust Localization Using RGB-D Images
Yoonseon Oh,Songhwai Oh 제어로봇시스템학회 2014 제어로봇시스템학회 국제학술대회 논문집 Vol.2014 No.10
Visual information extracted from RGB images has been successfully used for mobile robot localization. The main difficulty with localization using RGB images is that visual features from RGB images are not completely invariant against changes in viewpoints and lighting conditions. This problem can be overcome using features from RGB-D images. In this paper, we evaluate two depth features, depth patches and histograms of oriented normal vectors, extracted from RGB-D images for localization of a mobile robot and demonstrate that robust localization is possible under varying lighting conditions.
Vision-Based 3D Reconstruction Using a Compound Eye Camera
Wooseok Oh,Hwiyeon Yoo,Timothy Ha,Songhwai Oh 제어로봇시스템학회 2021 제어로봇시스템학회 국제학술대회 논문집 Vol.2021 No.10
The vision-based 3D reconstruction methods have various advantages and can be used in various applications such as navigation. Although various vision-based methods are being studied, it is difficult to reconstruct many parts at once with a general camera because of a small FOV. To solve this problem, we propose a coarse but lightweight reconstruction method using a camera with a unique structure called a compound eye with various advantages such as large FOV. In the process, we devise a network that performs depth estimation on a compound eye structure to obtain a depth image containing 3D information from an RGB image. We tested our methods by collecting data using a compound eye camera implemented in a Gazebo simulation and simulation scenes we created. As a result, our 3D reconstruction method using the data we collected and the confidence score from our depth estimation result, can capture the environment with a high recall of 97.51%.
Semi-Supervised Imitation Learning with Mixed Qualities of Demonstrations for Autonomous Driving
Gunmin Lee,Wooseok Oh,Jeongwoo Oh,Seungyoun Shin,Dohyeong Kim,Jaeyeon Jeong,Sungjoon Choi,Songhwai Oh 제어로봇시스템학회 2022 제어로봇시스템학회 국제학술대회 논문집 Vol.2022 No.11
In this paper, we consider the problem of autonomous driving using imitation learning in a semi-supervised manner. In particular, both labeled and unlabeled demonstrations are leveraged during training by estimating the quality of each unlabeled demonstration. If the provided demonstrations are corrupted and have a low signal-to-noise ratio, the performance of the imitation learning agent can be degraded significantly. To mitigate this problem, we propose a method called semi-supervised imitation learning (SSIL). SSIL first learns how to discriminate and evaluate each state-action pair’s reliability in unlabeled demonstrations by assigning higher reliability values to demonstrations similar to labeled expert demonstrations. This reliability value is called leverage. After this discrimination process, labeled and unlabeled demonstrations with estimated leverage values are utilized while training the policy in a semi-supervised manner. The experimental results demonstrate the validity of the proposed algorithm using unlabeled trajectories with mixed qualities. Moreover, the hardware experiments using an RC car are conducted to show that the proposed method can be applied to real-world applications.
Learning to use Topological Memory for Visual Navigation
Obin Kwon,Songhwai Oh 제어로봇시스템학회 2020 제어로봇시스템학회 국제학술대회 논문집 Vol.2020 No.10
We propose a new method of using topological memory in visual navigation problem. In our method, an agent builds a topological map during the navigation and simultaneously reasoning on the built map. The proposed method can efficiently use the elements of the graph memory using GCN and Transformer network. We evaluated our method on the visual target navigation problem which need some exploration strategies. The agent is trained using Deep Reinforcement Learning in photo-realistic Habitat simulator with Matterport 3d dataset. Using much smaller memory than the baseline, our method achieved competitive performance in visual target navigation problem.
Efficient Environmental Monitoring Using Cost-Aware Path Planning
Junghun Suh,Songhwai Oh 제어로봇시스템학회 2013 제어로봇시스템학회 국제학술대회 논문집 Vol.2013 No.10
This paper presents an efficient environmental monitoring strategy that considers the information gain along the trajectory of a robot. In order to monitor environmental parameters such as temperature and chemical concentration, an estimation method based on Gaussian process regression is used. The goal of this paper is to model accurate spatio-temporal phenomena by reducing the uncertainty over the surveillance region. A cost-aware path planning based environmental monitoring is desirable for mobile sensor networks since robots are coordinated to follow a trajectory with the maximum accumulated information gain, as well as the traveling distance. The proposed method with respect to different sampling methods is demonstrated in simulations.