MyoDex: A Generalizable Prior for Dexterous Manipulation

Authors: Vittorio Caggiano, Sudeep Dasari, Vikash Kumar

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

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate Myo Dex s effectiveness in fewshot generalization as well as positive transfer to a large repertoire of unseen dexterous manipulation tasks. Myo Dex can solve approximately 3x more tasks and it can accelerate the achievement of solutions by about 4x in comparison to a distillation baseline.
Researcher Affiliation Collaboration 1FAIR, Meta AI 2CMU. Correspondence to: Vittorio Caggiano <caggiano@gmail.com>, Sudeep Dasari <sdasari@andrew.cmu.edu>, Vikash Kumar <vikashplus@gmail.com>.
Pseudocode No The paper describes the reinforcement learning setup but does not include any explicit pseudocode or algorithm blocks.
Open Source Code No The paper mentions a project webpage (https://sites.google.com/view/myodex) but does not contain an explicit statement about releasing source code or a direct link to a code repository within its text.
Open Datasets Yes Every task setup (see Figure 3) consists of a tabletop environment, an object from the Contact DB dataset (Brahmbhatt et al.), and the Myo Hand.
Dataset Splits No The paper describes using a subset of 14 Myo DM tasks for pre-training and 43 'out-of-domain' tasks for fine-tuning/evaluation, but it does not specify explicit train/validation/test dataset splits with percentages or sample counts for data within each task.
Hardware Specification No The paper does not provide specific hardware details such as exact GPU/CPU models, processor types, or memory amounts used for running its experiments.
Software Dependencies No The paper mentions software like MuJoCo, Stable-Baselines, and scikit-learn, but does not provide specific version numbers for any of them.
Experiment Setup Yes Table A.1: Parameters adopted for the reinforcement learning models. Samples for Iterations 4096 Discount Factor (γ) 0.95 GAE-λ 0.95 VF Coefficient (c1) 0.5 Entropy Bonus (c2) 0.001 Clip Parameter (ϵ) 0.2 Batch Size 256 Epochs 5 Network Size pi = [256, 128], vf = [256, 128]