2026-09-21T17:38:24Z https://www.ideals.illinois.edu/oai-pmh

oai:www.ideals.illinois.edu:2142/125573 2025-02-06 col_2142_5131 col_2142_8888 com_2142_5130 com_2142_8887 com_2142_234

application/pdf en text Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-02-04 without embargo terms The student, Yixuan Wang, accepted the attached license on 2024-07-18 at 16:03. The student, Yixuan Wang, submitted this Thesis for approval on 2024-07-18 at 16:11. This Thesis was approved for publication on 2024-07-19 at 11:50. DSpace SAF Submission Ingestion Package generated from Vireo submission #20967 on 2025-02-04 at 21:04:18 Dynamic and structured scene representation for robotic manipulation Wang, Yixuan 2024-07-19 Li, Yunzhu Driggs-Campbell, Katie Hajek, Bruce Robotic Manipulation Robot Learning Representation Learning eng Representation is an essential component of typical robotic manipulation frameworks. An ideal representation should be both efficient for computation and sufficient for downstream tasks. However, existing representations typically have fixed dimensions, which may not be optimal for different tasks. Therefore, in the first part of the thesis, we propose a dynamic representation that can dynamically adapt its dimensions to current the observation and the goal. Through various pile manipulation experiments, we demonstrate that dynamic representation can significantly improve the performance of robotic manipulation tasks compared to fixed- size representations. In the second part of the thesis, we propose another novel implicit representation, D$^3$Fields, that is 3D, semantic, and dynamic. Such a representation can be used for zero-shot generalizable rearrangement tasks, where the goal is specified by 2D images. We demonstrate the effectiveness of our D$^3$Fields through a wide range of robotic rearrangement tasks, including organizing shoes, collecting debris, and organizing office desks. Compared to state-of-the-art implicit 3D representations, such as FeatureNeRF and DistilledNeRF [1], [2], our D$^3$Fields is more computationally efficient and effective for novel scenes. 2024-08 Thesis https://hdl.handle.net/2142/125573 Copyright 2024 Yixuan Wang Electrical & Computer Eng Electrical & Computer Engr University of Illinois at Urbana-Champaign M.S. Thesis