Controllability-Aware Unsupervised Skill Discovery

Authors: Seohong Park, Kimin Lee, Youngwoon Lee, Pieter Abbeel

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

Reproducibility Variable Result LLM Response
Research Type Experimental Our experimental results in six robotic manipulation and locomotion environments demonstrate that CSD can discover diverse complex skills including object manipulation and locomotion skills with no supervision, significantly outperforming prior unsupervised skill discovery methods.
Researcher Affiliation Collaboration 1University of California, Berkeley 2Google Research. Correspondence to: Seohong Park <seohong@berkeley.edu>.
Pseudocode Yes Algorithm 1 Controllability-aware Skill Discovery (CSD)
Open Source Code Yes Videos and code are available at https://seohong.me/projects/csd/
Open Datasets Yes We test CSD on six environments across three different domains: three Fetch manipulation environments (Fetch Push, Fetch Slide, and Fetck Pick And Place) (Plappert et al., 2018), Kitchen (Gupta et al., 2019), and two Mu Jo Co locomotion environments (Ant and Half Cheetah) (Todorov et al., 2012; Brockman et al., 2016).
Dataset Splits No The paper does not explicitly provide specific training, validation, and test dataset splits with percentages or counts, or refer to standard predefined splits.
Hardware Specification Yes We run our experiments on an internal cluster with NVIDIA Tesla V100 and NVIDIA GeForce RTX 2080 Ti GPUs.
Software Dependencies No The paper mentions optimizers (Adam) and frameworks (SAC) with citations, but does not specify version numbers for general software dependencies like Python, PyTorch, or other libraries. It states implementations are 'on top of the publicly available codebase of MUSIC' and 'based on the publicly available codebase of LSD', implying dependencies, but without specific version numbers for them.
Experiment Setup Yes We present the full list of the hyperparameters used in our experiments in Tables 1 and 2