EfficientFlow: 为具身AI高效的等价流策略学习
EfficientFlow: Efficient Equivariant Flow Policy Learning for Embodied AI
AI总结：
EfficientFlow通过引入等价性与加速正则化策略，提升具身AI的效率与性能。
摘要：
生成建模最近在视觉-运动策略学习中展示了显著的潜力，能够实现跨多样具身AI任务的灵活和表达性控制。然而，现有的生成策略往往在数据效率方面存在问题，需要大规模的示范数据，并且在采样效率方面也存在问题，在推理过程中导致动作生成缓慢。我们引入EfficientFlow，一种统一的框架，用于高效的具身AI流基于策略学习。为了提高数据效率，我们将在流匹配中引入等价性。我们理论证明，当使用各向同性高斯先验和等价速度预测网络时，所得到的动作分布保持等价性，从而提高泛化能力和大幅减少数据需求。为了加速采样，我们提出了一种新的加速正则化策略。由于直接计算加速度对于边际流轨迹来说是不可行的，我们推导出一种新的替代损失，使得仅使用条件轨迹即可实现稳定和可扩展的训练。在广泛的机器人操作基准测试中，所提出的算法在数据有限的情况下实现了竞争或优越的性能，同时提供了显著更快的推理速度。这些结果突显了EfficientFlow作为高性能具身AI强大而高效范式的潜力。
英文摘要：
Generative modeling has recently shown remarkable promise for visuomotor policy learning, enabling flexible and expressive control across diverse embodied AI tasks. However, existing generative policies often struggle with data inefficiency, requiring large-scale demonstrations, and sampling inefficiency, incurring slow action generation during inference. We introduce EfficientFlow, a unified framework for efficient embodied AI with flow-based policy learning. To enhance data efficiency, we bring equivariance into flow matching. We theoretically prove that when using an isotropic Gaussian prior and an equivariant velocity prediction network, the resulting action distribution remains equivariant, leading to improved generalization and substantially reduced data demands. To accelerate sampling, we propose a novel acceleration regularization strategy. As direct computation of acceleration is intractable for marginal flow trajectories, we derive a novel surrogate loss that enables stable and scalable training using only conditional trajectories. Across a wide range of robotic manipulation benchmarks, the proposed algorithm achieves competitive or superior performance under limited data while offering dramatically faster inference. These results highlight EfficientFlow as a powerful and efficient paradigm for high-performance embodied AI.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
