[2506.03863] STAR: Learning Diverse Robot Skill Abstractions through Rotation-Augmented Vector Quantization

Computer Science > Robotics arXiv:2506.03863 (cs) [Submitted on 4 Jun 2025 (v1), last revised 7 Apr 2026 (this version, v3)] Title:STAR: Learning Diverse Robot Skill Abstractions through Rotation-Augmented Vector Quantization Authors:Hao Li, Qi Lv, Rui Shao, Xiang Deng, Yinchuan Li, Jianye Hao, Liqiang Nie View a PDF of the paper titled STAR: Learning Diverse Robot Skill Abstractions through Rotation-Augmented Vector Quantization, by Hao Li and 6 other authors View PDF HTML (experimental) Abstract:Transforming complex actions into discrete skill abstractions has demonstrated strong potential for robotic manipulation. Existing approaches mainly leverage latent variable models, e.g., VQ-VAE, to learn skill abstractions through learned vectors (codebooks), while they suffer from codebook collapse and modeling the causal relationship between learned skills. To address these limitations, we present \textbf{S}kill \textbf{T}raining with \textbf{A}ugmented \textbf{R}otation (\textbf{STAR}), a framework that advances both skill learning and composition to complete complex behaviors. Specifically, to prevent codebook collapse, we devise rotation-augmented residual skill quantization (RaRSQ). It encodes relative angles between encoder outputs into the gradient flow by rotation-based gradient mechanism. Points within the same skill code are forced to be either pushed apart or pulled closer together depending on gradient directions. Further, to capture the causal relationship between ...