Domestic Conference Articles
    Development of VR Teleoperation System and Collecting Demonstration for Deformable Object Manipulation
    Minseok Song, Bonggyeong Park, Daehyung Park
    Institute of Control, Robotics and Systems (ICROS), 2024
    A VR teleoperation system to collect human demonstrations for DOM in high quality. We collected time-series demonstration data about four different O-ring object which are containing joint state information of robot, RGB images of camera on wrist and base of robot.
    [PDF]
    Reactive Task Planning using Scene Graph for Robust Robotic Manipulation
    Ulzhalgas Rakhman, Jaehoon Yoo, Yeseung Kim, Deokmin Hwang, Seunghoon Hong, Daehyung Park
    Korea Robotics Society Annual Conference (KRoC), 2022
    A reactive task planning-and-execution framework adopting scene-graph for automatic abstraction and behavior tree for robust task execution. We demonstrate the effectiveness and efficiency of our framework against human interventions in robotic assembly scenarios.
    [PDF]
    Outstanding Paper Award
    Transferable-Reward Decomposition for Inverse Reinforcement Learning
    Jaehwi Jang and Daehyung Park
    Korea Robotics Society Annual Conference (KRoC), 2022
    A transferable reward-decomposition method for inverse reinforcement learning (IRL) that returns a unique pair of primary goal and its residual rewards, separately. Our method lowers the computational complexity of reward decomposition by reusing the exploration results of IRL.
    [PDF]
    A Telemanipulation Suite for Deformable Object Manipulation
    Bonggyeong Park, Chanyoung Ahn, Daehyung Park
    Korean Artificial intelligence Association (KAIA), 2022
    Deformable object manipulation with learning from demonstration requires a dataset that absorbs human intelligence. We propose a virtual reality-based telemanipulation suite that allows an expert hand to manipulate simulated deformable objects while recording complete observations.
    A Framework for Natural Language-guided Semantic Mapping and Mobile Navigation
    Jinwoo Kim, Dohyun Kim, Daehyung Park
    Korea Robotics Society Annual Conference (KRoC), 2022
    A novel method for natural language-driven location labeling and mobile navigation. Our approach automatically labels the location description on a metric map via natural language grounding. By associating the semantic map with a world model, our method enables the robot to understand natural language instruction for indoor delivery navigation.
    [PDF]
  1. Minsung Yoon, Daehyung Park, Sung-Eui Yoon, "Bias tree expansion using reinforcement learning for efficient motion planning," Korea Robotics Society Annual Conference (KRoC), 2021 [PDF]
  2. Hyeongyeol Ryu, Minsung Yoon, Daehyung Park, Sung-Eui Yoon, "Robust Robot Navigation against External Disturbance using Deep Reinforcement Learning," Korea Robotics Society Annual Conference (KRoC), 2021 [PDF]