AdaManip: Adaptive Articulated Object Manipulation Environments and Policy Learning

Authors: Yuanfei Wang, Xiaojie Zhang, Ruihai Wu, Yu Li, Yan Shen, Mingdong Wu, Zhaofeng He, Yizhou Wang, Hao Dong

ICLR 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The effectiveness of our designs and proposed method is validated through both simulation and real-world experiments. Based on our proposed novel environment, we have conducted extensive experiments on 9 categories of 277 different objects, covering 5 types of mechanisms, showcasing the necessity of the proposed environment and dataset, and the effectiveness of the proposed policy learning framework in efficiently and intelligently adapting the manipulation.
Researcher Affiliation Academia 1 Center on Frontiers of Computing Studies, School of Computer Science, Peking University 2 Beijing University of Posts and Telecommunications 3 Inst. for Artificial Intelligence, Peking University 4 Nat l Eng. Research Center of Visual Technology, Peking University 5 State Key Laboratory of General Artificial Intelligence, Peking University
Pseudocode No The paper describes the methodology in narrative text and mathematical equations (Eq. 1, Eq. 2) but does not include any clearly labeled pseudocode or algorithm blocks.
Open Source Code No Our project page is available at: https://adamanip.github.io. The paper provides a link to a project page but does not explicitly state that the source code for the methodology is provided at this link, nor is it a direct link to a code repository.
Open Datasets No We introduce a new dataset that encompasses more realistic adaptive manipulation mechanisms. Our dataset includes 9 categories of 277 objects: Bottle, Pen, Coffee Maker, Window, Pressure Cooker, Lamp, Door, Safe, and Microwave. The object assets in our dataset are handcrafted from materials primarily obtained from 3D Warehouse (Trimble). More details can be found in Appendix B. While the paper describes the creation of a new dataset and mentions some sources for object assets, it does not provide concrete access information (link, DOI, repository) for their collected dataset.
Dataset Splits No We conduct experiments in the category level covering all the 9 object categories, and collect 20 adaptive manipulation demonstrations for each object as the training data. The paper mentions collecting training data but does not specify how this data is split into training, validation, or test sets for reproducibility.
Hardware Specification Yes Table 6: Parameters for training and diffusion model, Hardware Configuration NVIDIA Ge Force GTX 4090
Software Dependencies No The paper refers to various methods and models like Diffusion Policy, DDPM, and Point Net++, but does not explicitly list specific software dependencies with their version numbers (e.g., Python, PyTorch, CUDA versions) needed to replicate the experimental environment.
Experiment Setup Yes Table 6: Parameters for training and diffusion model. Training Values include Weight Decay 1e-6, Batch Size 64, Optimizer Adam, Learning Rate 1e-4, Epochs 500.