RSG: Fast Learning Adaptive Skills for Quadruped Robots by Skill Graph

开发能够快速适应未知野外环境的机器人智能系统是追求自主机器人的重要挑战之一。尽管在腿式机器人的步态稳定性和技能学习方面取得了一些令人印象深刻的进展，但它们的快速适应能力仍然不如自然界中的动物。动物天生具备生存所需的大量技能，并且可以通过将基本技能与有限的经验组合来快速获得新技能。受此启发，我们提出了一种名为机器人技能图（RSG）的新框架，用于组织机器人的大量基础技能并灵活地重复使用它们以实现快速适应。RSG的结构类似于知识图谱（KG），由大量动态行为技能组成，而不是KG中的静态知识，并能够发现存在于机器人学习环境和获得技能之间的隐含关系，为了解机器人技能学习中存在的微妙模式提供了一个起点。广泛的实验结果表明，RSG能够对新任务和环境进行合理的技能推理，并使四足机器人能够快速适应新场景并学习新技能。

The paper proposes a framework, named Robot Skill Graph (RSG), for organizing massive fundamental skills of robots and reusing them for fast adaptation. The problem the paper aims to solve is the challenge of developing robotic intelligent systems that can adapt quickly to unseen wild situations.

The key idea of the paper is to create a structure similar to the Knowledge Graph (KG) but composed of massive dynamic behavioral skills instead of static knowledge. The RSG enables discovering implicit relations that exist between the learning context and acquired skills of robots, serving as a starting point for understanding subtle patterns existing in robots' skill learning.

The paper's experiments demonstrate that RSG can provide rational skill inference upon new tasks and environments and enable quadruped robots to adapt to new scenarios and learn new skills rapidly. The paper also highlights the importance of organizing fundamental skills for fast adaptation and draws inspiration from animals' ability to quickly acquire new skills. The authors provide an open-source implementation of the RSG framework for future research.

Some related research in this field includes 'Learning to Learn for Global Optimization of Black Box Functions' by Li et al., 'Learning to Learn without Forgetting by Maximizing Transfer and Minimizing Interference' by Nguyen et al., and 'Meta-Learning with Memory-Augmented Neural Networks' by Santoro et al.