Dual-Process Atomic Skill Learning: Decoupling Semantic Reasoning and Real-Time Control
2026-07-12 • Robotics
Robotics
AI summaryⓘ
The authors focus on teaching robots to follow complex multi-step instructions in natural language. They identify problems with existing methods that mix high-level task planning and low-level action execution in a way that causes instability during training. To fix this, they create a new approach called DASL that separates slow decision-making about what skill to use from fast control of robot movements, using separate models working asynchronously. Their method uses special techniques like Vector Quantization and latent diffusion to keep skills organized and distinct. Tests show that DASL learns skills better and handles new instructions more effectively than previous methods.
Language-conditioned Imitation LearningHierarchical Imitation LearningVector QuantizationDual-Process TheoryLatent Diffusion ModelDecision TransformerSkill CodebookCompositional GeneralizationAsynchronous Training
Authors
Jun Chen, Erdent Bao, Wenlong Dong, Jierui Liu, Qi Cai, Hao Wan, Shaopeng Li, Weijun Qin, Jing Liang, Huiping Zhuang
Abstract
Language-conditioned Imitation Learning (IL) is essential for enabling robots to perform complex tasks following natural language instructions. However, generalizing to multi-step compositional tasks remains a significant challenge. While hierarchical approaches attempt to address this by decomposing tasks into atomic skills, existing methods often suffer from training instability and codebook collapse due to the tight coupling between high-level skill reasoning and low-level action generation in joint training paradigms. Inspired by the Dual-Process Theory of cognition, we propose Dual-Process Atomic Skill Learning (DASL), a novel asynchronous hierarchical imitation learning framework that decouples slow semantic reasoning from fast, real-time motion control. DASL comprises a Slow-Frequency Policy that predicts interpretable, discrete skills via Vector Quantization, and a High-Frequency Policy that leverages a latent diffusion model and a Decision Transformer to generate precise actions conditioned on these latent skills. By asynchronously coordinating these modules and utilizing diffusion to structure the latent space, our framework mitigates the skill codebook interference problem common in joint training paradigms. Evaluations across simulation benchmarks and experiment demonstrate that DASL significantly outperforms state-of-the-art baselines, excelling in skill acquisition and compositional generalization to unseen instructions. GitHub page: https://github.com/Hatakekaka/DASL