Symmetry-Aware Robot Design with Structured Subgroups

Authors: Heng Dong, Junyu Zhang, Tonghan Wang, Chongjie Zhang

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We further empirically evaluate SARD on various tasks, and the results show its superior efficiency and generalizability.
Researcher Affiliation Academia 1Institute for Interdisciplinary Information Sciences, Tsinghua University 2Huazhong University of Science and Technology 3Harvard University.
Pseudocode Yes Algorithm 1 SARD: Symmetry-Aware Robot Design
Open Source Code Yes And our code is available at Git Hub2.
Open Datasets Yes Environments. All six tasks are adapted from Gupta et al. (2021b), which are created based on Mu Jo Co (Todorov et al., 2012) physics simulator.
Dataset Splits No The paper describes the reinforcement learning setup and experimental evaluation but does not specify explicit training, validation, and test dataset splits for the environment data.
Hardware Specification Yes Experiments are carried out on NVIDIA GTX 2080 Ti GPUs. ... SARD requires approximately 10G of RAM and 4G of video memory...
Software Dependencies No The paper mentions software like Py Torch (Paszke et al., 2019), Py Torch Geometric package (Fey & Lenssen, 2019), and Mu Jo Co (Todorov et al., 2012) with citations, but does not provide explicit version numbers (e.g., PyTorch 1.9, MuJoCo 2.0) for reproducibility.
Experiment Setup Yes Here we provide the hyperparameters needed to replicate our experiments in Table 1, and we also include our codes in the supplementary. Table 1: Hyperparameters of SARD and Transform2Act. [Lists hyperparameters such as Learning Rate, Batch Size, Discount Factor, etc.]