SimpleTouch improves robot touch-based manipulation without extra training
SimpleTouch: Can Vision-Language-Action Models Master Contact-Rich Manipulation Without Tactile Policy Pretraining?
Robotics
Summary
Robots often need to feel what they touch to perform tasks well, but teaching them to use this sense usually requires extra complex training. The authors created SimpleTouch, a simpler method that lets robots learn to use touch along with vision and language by combining pretrained models in one training step. This approach works better than previous ones on both simulated and real robot tasks, achieving higher success rates with less data. Their work shows that complex, separate touch training is not always necessary to improve robots' contact handling.
What this means in practice
- •For robotic system builders: Integrate pretrained touch and vision models in robots with a single training stage to improve grasping and manipulation tasks involving contact.
- •For industrial automation teams: Deploy simpler training pipelines for robots performing contact-rich assembly without needing costly additional tactile policy pretraining.
Authors
Chen Yang, Linzhe Shi, Changjie Wu, Hang Zhang, Ronghan Chen, Lingjun Zhang, Xu Hu, Mu Xu, Jiansheng Fan, Chen Wang
Abstract
Tactile sensing provides essential contact information for robotic manipulation, yet incorporating it into pretrained vision-language-action (VLA) models remains challenging. A common concern is that simply introducing touch during task-specific fine-tuning may fail to bridge the cross-modal gap, yielding limited gains or even reduced success. Consequently, existing methods often rely on large-scale tactile policy pretraining or separate visuotactile alignment, adding data requirements and training stages. We introduce SimpleTouch, a simple VLA extension that augments $π_{0.5}$ with a tactile expert, to test whether these additional stages are necessary. Leveraging all tokens from a frozen pretrained tactile encoder, the expert learns from action supervision and multi-horizon prediction of future tactile latents. This single-stage training uses only task demonstrations, without additional tactile policy pretraining or separate alignment. With 50 demonstrations per task, SimpleTouch achieves the highest success rate among evaluated methods on all six UniVTAC tasks. Its average success rate reaches 77.5%, compared with 45.2% for FTP-$π_{0.5}$ and 66.7% for FTP-1, corresponding to gains of 32.3 and 10.8 percentage points, respectively. Across four real-world tasks, it averages 71.3%, exceeding FTP-1 by 8.8 percentage points. These results demonstrate that, given pretrained VLA and tactile representations, additional tactile policy pretraining is not a prerequisite for strong performance on these tasks, offering a simpler route to contact-rich manipulation. Project page: https://simpletouch-robot.github.io/