A5057994317- Soil Mechanics and Vehicle Dynamics
- Robot Manipulation and Learning
- Mechanical Engineering and Vibrations Research
- Agricultural Engineering and Mechanization
- Robotics and Sensor-Based Localization
- Reinforcement Learning in Robotics
- Spectroscopy and Chemometric Analyses
- Multimodal Machine Learning Applications
- Soft Robotics and Applications
- Domain Adaptation and Few-Shot Learning
- Advanced Vision and Imaging
- Advanced Neural Network Applications
- Agriculture and Farm Safety
- Gear and Bearing Dynamics Analysis
- Advanced Image and Video Retrieval Techniques
- Engineering Applied Research
- Adversarial Robustness in Machine Learning
- Image and Object Detection Techniques
- Hydraulic and Pneumatic Systems
- Agriculture, Soil, Plant Science
- Modular Robots and Swarm Intelligence
- Human Pose and Action Recognition
- Greenhouse Technology and Climate Control
- Muscle activation and electromyography studies
- 3D Surveying and Cultural Heritage
University of Minnesota
2019-2025
Twin Cities Orthopedics
2020-2023
Chungnam National University
2011-2020
Sungkyunkwan University
2007-2020
Kyushu University
2018
Massachusetts Institute of Technology
2015-2018
ETH Zurich
2017
Georgia Institute of Technology
2009-2015
Samsung Medical Center
2006-2013
Atlanta Technical College
2012
Duckietown is an open, inexpensive and flexible platform for autonomy education research. The comprises small autonomous vehicles ("Duckiebots") built from off-the-shelf components, cities ("Duckietowns") complete with roads, signage, traffic lights, obstacles, citizens (duckies) in need of transportation. offers a wide range functionalities at low cost. Duckiebots sense the world only one monocular camera perform all processing onboard Raspberry Pi 2, yet are able to: follow lanes while...
We propose a voting-based pose estimation algorithm applicable to 3D sensors, which are fast replacing their 2D counterparts in many robotics, computer vision, and gaming applications. It was recently shown that pair of oriented points, points on the object surface with normals, voting framework enables robust estimation. Although discriminative for objects sufficient curvature changes, they not compact enough industrial real-world mostly planar. As edges play key role registration, depth...
We present a three-dimensional deep convolutional neural network (3D CNN) approach for grasping unknown objects with soft hands. Soft hands are compliant and capable of handling uncertainty in sensing actuation, but come at the cost unpredictable deformation fingers. Traditional model-driven approaches, which assume known models objects, robot hands, stable grasps expected contacts, inapplicable to such since predicting contact points between is not straightforward. Our solution adopts CNN...
This paper presents a particle filtering approach for 6-DOF object pose tracking using an RGB-D camera. Our filter is massively parallelized in modern GPU so that it exhibits real-time performance even with several thousand particles. Given priori 3D mesh model, the proposed renders model onto texture buffers GPU, and rendered results are directly used by our likelihood evaluation. Both photometric (colors) geometric (3D points surface normals) features employed to determine of each respect...
In this paper, we present an object pose estimation algorithm exploiting both depth and color information. While many approaches assume that a target region is cleanly segmented from background, our approach does not rely on assumption, thus it can estimate of in heavy clutter. Recently, oriented point pair feature was introduced as low dimensional description surfaces. The has been employed voting scheme to find set possible 3D rigid transformations between model test scene features....
We study an emerging problem named “grasping the invisible” in robotic manipulation, which a robot is tasked to grasp initially invisible target object via sequence of pushing and grasping actions. In this problem, pushes are needed search for rearrange cluttered objects around it enable effective grasps. propose solve by formulating deep learning approach critic-policy format. The target-oriented motion critic, maps both visual observations information expected future rewards primitives,...
Phosphorylated AKT (pAKT) is a major contributor to radioresistance in human cancers. The aim of this study was investigate the association pAKT expression and radiation resistance cervical cancer. A retrospective review made records 27 women who received primary therapy due locally advanced cancer (LACC) with FIGO stage IIB-IVA. Nine patients regarded as resistant developed local recurrences median progression free interval 9 months. Eighteen did not show recurrences, were...
We present a 3D model-based visual tracking approach using edge and keypoint features in particle filtering framework. Recently, particle-filtering-based approaches have been proposed to integrate multiple pose hypotheses shown good performance, but most of the work has made an assumption that initial is given. To ameliorate this limitation, we employ for initialization filter. Given 2D–3D correspondences, randomly choose set minimum correspondences calculate possible hypotheses. Based on...
We propose a combined approach for 3D real-time object recognition and tracking, which is directly applicable to robotic manipulation. use keypoints features the initial pose estimation. This estimate serves as an edge-based tracking. The combination of these two complementary methods provides efficient robust tracking solution. main contributions this paper includes: 1) While most RAPiD style have used simplified CAD models or at least manually well designed models, our system can handle...
This paper presents an approach to textureless object detection and tracking of the 3D pose. Our schemes are coherently integrated in a particle filtering framework on special Euclidean group, SE(3), which visual problem is tackled by maintaining multiple hypotheses For detection, efficient chamfer matching employed so that set coarse pose estimated from between 2D edge templates query image. Particles then initialized randomly drawing based costs matching. To ensure particles at or close...
Development of nondestructive measurements soluble solids and firmness apples benefits the producers,processors, packers. The (Magness-Taylor (MT) maximum force) were predictedby diffuse reflectance measurement in visible/near-infrared (400-1800 nm) regions spectrum. Two apple cultivars,Gala (n = 800) Red Delicious 960), measured. spectroscopic firmnessfor each validated by a refractometer reading on juice puncture test using universal testing instrument,respectively. could be predicted NIR...
In this paper we present algorithms and experiments for multi-scale assembly of complex structures by multi-robot teams. We also focus on tasks where successful completion requires multiple types operations with a range precision requirements. develop hierarchical planning approach to perception in support manipulation, which the resolution operation is matched required manipulation operation. demonstrate these techniques context multi-step task robots assemble large box-like objects,...
As artificial intelligence becomes an increasingly prevalent method of enhancing robotic capabilities, it is important to consider effective ways train these learning pipelines and leverage human expertise. Working towards goals, a master-apprentice model presented evaluated during grasping task for effectiveness perception. The apprenticeship augments self-supervised with by demonstration, efficiently using the human's time expertise while facilitating future scalability supervision...
Motivated by the stringent requirements of unstructured real-world where a plethora unknown objects reside in arbitrary locations surface, we propose voxel-based deep 3D Convolutional Neural Network (3D CNN) that generates feasible 6-DoF grasp poses unrestricted workspace with reachability awareness. Unlike majority works predict if proposed pose within restricted will be successful solely based on stability, our approach further learns predictor evaluates is reachable or not from robot's...
Grasping a novel target object in constrained environments (e.g., walls, bins, and shelves) requires intensive reasoning about grasp pose reachability to avoid collisions with the surrounding structures. Typical 6-DoF robotic grasping systems rely on prior knowledge environment planning computation, which is ungeneralizable inefficient. In contrast, we propose Collision-Aware Reachability Predictor (CARP) for systems. The CARP learns estimate collision-free probabilities poses significantly...
Interactive robotic grasping using natural language is one of the most fundamental tasks in human-robot interaction. However, can be a source ambiguity, particularly when there are ambiguous visual or linguistic contents. This paper investigates use object attributes disambiguation and develops an interactive system capable effectively resolving ambiguities via dialogues. Our approach first predicts target scores attribute through vision-and-language grounding. To handle objects commands, we...
This article presents an integrated engine-hydro-mechanical transmission control algorithm for a tractor considering the efficiency. First, hydro-mechanical efficiency was obtained by network analysis based on hydrostatic unit constructed from test. Using map and thermal of engine, optimal operating line obtained, which provides higher total system Based line, proposed, powertrain To evaluate performance proposed algorithm, AMESim-MATLAB/Simulink-based co-simulator developed. From simulation...
Journal Article Calcium Oxalate Crystals in the Thyroid: Their Identification, Prevalence, Origin, and Possible Significance Get access John D. Reid, M.D., M.D. Department of Pathology, Robinson Memorial Hospital, Ravenna, Ohio, Northeast Ohio Universities College Medicine, Rootstown, Cleveland Metropolitan General Cleveland, Research Department, Louis C. Herring Company,Orlando, Florida Search for other works by this author on: Oxford Academic Google Scholar Chang-Hyun Choi, Norman O....
In this paper, we present a 3D model-based object tracking approach using edge and keypoint features in particle filtering framework.Edge points provide 1D information for pose estimation it is natural to consider multiple hypotheses.Recently, based approaches have been proposed integrate hypotheses shown good performance, but most of the work has made an assumption that initial given.To remove assumption, employ initialization filter.Given 2D-3D correspondences, choose set minimum...