Tenma：基于Diffusion Transformer的鲁棒跨体型机器人操作
Tenma: Robust Cross-Embodiment Robot Manipulation with Diffusion Transformer
AI总结：
提出轻量级扩散Transformer模型Tenma用于跨体型双臂机器人操作，通过跨体型归一化器和联合状态-时间编码器整合多模态数据，在中等数据规模下实现88.95%的分布内平均成功率及强泛化能力。
摘要：
扩展Transformer策略和扩散模型推动了机器人操作的发展，但在轻量级、跨体型学习设置中结合这些技术仍具挑战。我们研究了在异构、多模态机器人数据上训练的扩散Transformer策略中，对稳定性和性能影响最大的设计选择，并提出了用于双臂控制的轻量级扩散Transformer——Tenma。Tenma通过跨体型归一化器将不同的状态/动作空间映射到共享潜在空间，从而整合多视角RGB、本体感知和语言；通过联合状态-时间编码器实现时间对齐的观测学习并提升推理速度；通过针对训练稳定性和学习能力优化的扩散动作解码器生成动作。在基准测试中且计算量匹配的情况下，Tenma在分布内实现了88.95%的平均成功率，并在物体和场景变化下保持强劲性能，大幅超越了基线策略（其最佳分布内平均成功率仅为18.12%）。尽管使用中等规模数据，Tenma仍实现了鲁棒的操作和泛化能力，表明多模态和跨体型学习策略在进一步增强基于Transformer的模仿学习策略能力方面具有巨大潜力。
英文摘要：
Scaling Transformer policies and diffusion models has advanced robotic manipulation, yet combining these techniques in lightweight, cross-embodiment learning settings remains challenging. We study design choices that most affect stability and performance for diffusion-transformer policies trained on heterogeneous, multimodal robot data, and introduce Tenma, a lightweight diffusion-transformer for bi-manual arm control. Tenma integrates multiview RGB, proprioception, and language via a cross-embodiment normalizer that maps disparate state/action spaces into a shared latent space; a Joint State-Time encoder for temporally aligned observation learning with inference speed boosts; and a diffusion action decoder optimized for training stability and learning capacity. Across benchmarks and under matched compute, Tenma achieves an average success rate of 88.95% in-distribution and maintains strong performance under object and scene shifts, substantially exceeding baseline policies whose best in-distribution average is 18.12%. Despite using moderate data scale, Tenma delivers robust manipulation and generalization, indicating the great potential for multimodal and cross-embodiment learning strategies for further augmenting the capacity of transformer-based imitation learning policies.
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。
