BROWSE
Detail
Conference Paper
디지털 트윈 기반 실·가상 연계 로봇 시연 데이터 생성 및 궤적 생성 기반 가상 데이터 증강
Cited - time in 
Share 
Share
- Issue Date
- 2026-06
- Citation
- 대한전자공학회 학술 대회 (하계) 2026, pp.1-3
- Publisher
- 대한전자공학회
- Language
- Korean
- Type
- Conference Paper
- Abstract
- This paper presents a digital-twin-based framework for generating, replaying, and augmenting robot demonstration data. The proposed system combines a real-to-sim linked demonstration pipeline with trajectory-driven virtual data augmentation. First, a task scene is reconstructed as digital assets and instantiated in Isaac Sim and Isaac Lab, where teleoperated or scripted demonstrations are recorded, replayed, and validated in a physics-based digital twin environment. Second, the replayable seed demonstrations are converted into Mimic-compatible subtask-annotated trajectories and used to generate additional synthetic demonstrations through trajectory adaptation, retargeting, and stitching. Unlike conventional scripted generation that plans and executes the entire motion sequence from scratch, the proposed augmentation pipeline reuses existing demonstration structures and produces diverse trajectories more efficiently.
- KSP Keywords
- Data Augmentation, Digital Twin, Digital assets, Physics-based, Trajectory-driven, Virtual data, motion sequence