Curriculum-based Co-design of Morphology and Control of Voxel-based Soft Robots

Authors: Yuxing Wang, Shuang Wu, Haobo Fu, QIANG FU, Tiantian Zhang, Yongzhe Chang, Xueqian Wang

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

Reproducibility Variable Result LLM Response
Research Type Experimental 5 EMPIRICAL EVALUATION
Researcher Affiliation Collaboration 1Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, China 2Tencent AI Lab, Shenzhen, China
Pseudocode No The paper describes algorithms and processes but does not include structured pseudocode or algorithm blocks.
Open Source Code Yes The environment details can be found in Appendix A and our codes are available online1. 1https://github.com/Yuxing-Wang-THU/Modular Evo Gym
Open Datasets Yes We create a modular robot design space described in Section 4 and 8 task-related environments based on Evolution Gym (Bhatia et al., 2021).
Dataset Splits No The paper does not provide explicit training, validation, or test dataset splits in terms of percentages or sample counts. Reinforcement learning experiments typically differ from supervised learning in this regard, focusing on environment interaction and policy performance over dataset splitting.
Hardware Specification Yes For all the environments used in this paper, it takes around 1 day to train our model on a computer with 12 CPU cores and an NVIDIA RTX 3090 GPU.
Software Dependencies No We use Py Torch (Paszke et al., 2019) to implement our proposed method. (Note: No specific version number for PyTorch is given in the text).
Experiment Setup Yes Table 2: Hyperparameters of Cu Co. (includes GAE parameter λ 0.95, Learning rate 2.5 10 4, Policy epochs 8, Design steps 10, Number of layers 1, Transformer Embedding dimension 64, etc.)