Heterogeneous Tactile Transformer
About
Tactile sensors are inherently heterogeneous: a model trained on one sensor cannot be directly used on another, which limits learning contact-rich manipulation policies from diverse tactile data at scale. To bridge this gap, we propose the Heterogeneous Tactile Transformer (HTT), a framework that learns shared tactile representations across heterogeneous sensors. HTT consists of sensor-specific encoders and a shared transformer trunk, and is pretrained with per-modality masked reconstruction together with cross-modal alignment between paired sensors. Pretraining uses our novel Heterogeneous Paired Tactile (HPT) dataset, containing 1.6M synchronized paired frames across four vision- and array-based tactile sensors. Across distinct tactile perception and real-world manipulation tasks, HTT is shown to learn transferable representations that adapt to new tasks and previously unseen sensors. Dataset, code, and model checkpoints will be released upon publication at https://jxbi1010.github.io/htt-gh-page/.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| 20-class Object Classification | 9DTact tactile sensor | Top-1 Accuracy94.84 | 5 | |
| 20-class Object Classification | GSMini | Top-1 Accuracy91.35 | 5 | |
| Peg Insertion | ManiFeel 44 (simulation) | Success Rate48 | 5 | |
| Slip Detection | Tactile Sensors | 9DTact Score53.09 | 5 | |
| 20-class Object Classification | Heterogeneous tactile sensors Overall | Top-1 Accuracy66.2 | 5 | |
| Bulb Installation | ManiFeel 44 (simulation) | Success Rate77 | 5 | |
| Force Estimation | Tactile Sensors | 9DTact Performance Score0.606 | 5 | |
| 20-class Object Classification | Xela tactile sensor | Top-1 Accuracy56.68 | 3 | |
| 20-class Object Classification | TAC-02 | Top-1 Accuracy26.2 | 3 |