Tenma: Robust Cross-Embodiment Robot Manipulation with Diffusion Transformer
This work addresses the problem of enabling robust and generalizable robot manipulation across different embodiments for robotics researchers, representing a strong specific gain rather than a foundational breakthrough.
The paper tackled the challenge of scaling Transformer and diffusion models for robust cross-embodiment robot manipulation by introducing Tenma, a lightweight diffusion-transformer that achieved an average success rate of 88.95% in-distribution, substantially exceeding baseline policies at 18.12%.
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.