AtomicVLA: Unlocking the Potential of Atomic Skill Learning in Robots

TL;DR

AtomicVLA introduces a unified framework for robotic manipulation that leverages atomic skill abstractions and continual learning, significantly improving long-horizon task performance and scalability in simulation and real-world settings.

Contribution

It proposes AtomicVLA, a novel planning-and-execution framework with a Skill-Guided Mixture-of-Experts for scalable atomic skills and dynamic expert routing for continual learning.

Findings

Outperforms baselines on LIBERO and CALVIN benchmarks.

Achieves 18-21% improvement in real-world long-horizon tasks.

Enables continual learning of new skills with dynamic expert routing.

Abstract

Recent advances in Visual-Language-Action (VLA) models have shown promising potential for robotic manipulation tasks. However, real-world robotic tasks often involve long-horizon, multi-step problem-solving and require generalization for continual skill acquisition, extending beyond single actions or skills. These challenges present significant barriers for existing VLA models, which use monolithic action decoders trained on aggregated data, resulting in poor scalability. To address these challenges, we propose AtomicVLA, a unified planning-and-execution framework that jointly generates task-level plans, atomic skill abstractions, and fine-grained actions. AtomicVLA constructs a scalable atomic skill library through a Skill-Guided Mixture-of-Experts (SG-MoE), where each expert specializes in mastering generic yet precise atomic skills. Furthermore, we introduce a flexible routing encoder that automatically assigns dedicated atomic experts to new skills, enabling continual learning. We validate our approach through extensive experiments. In simulation, AtomicVLA outperforms $\pi_{0}$ by 2.4\% on LIBERO, 10\% on LIBERO-LONG, and outperforms $\pi_{0}$ and $\pi_{0.5}$ by 0.22 and 0.25 in average task length on CALVIN. Additionally, our AtomicVLA consistently surpasses baselines by 18.3\% and 21\% in real-world long-horizon tasks and continual learning. These results highlight the effectiveness of atomic skill abstraction and dynamic expert composition for long-horizon and lifelong robotic tasks. The project page is \href{https://zhanglk9.github.io/atomicvla-web/}{here}.