Researchers Develop Tactile-Only Blind Grasping System for Dexterous Robotic Hands
Researchers have created a framework enabling multi-fingered robotic hands to grasp objects without visual input, relying solely on tactile (touch) sensing. The approach combines a calibrated simulation environment, improved tactile signal processing, and reinforcement learning to train grasping policies. This advancement could expand robotic manipulation capabilities in scenarios where vision is unavailable or unreliable.
A team of researchers has developed a Real2Sim2Real framework for blind dexterous grasping—enabling robotic hands to grasp objects using only tactile feedback without visual input. The system addresses the challenge of the tactile sim-to-real gap through three main components: a Real2Sim calibration pipeline that creates a contact-calibrated digital twin simulator, a layout-aware tactile encoder that improves signal expressiveness through self-supervised learning, and a diffusion-policy approach that aggregates successful grasp trajectories from object-specific reinforcement-learning experts. When deployed on a physical LEAP Hand with distributed tactile sensors, the system achieved a 27% real-world grasp success rate across 20 objects (10 seen, 10 unseen) without requiring real-world demonstrations or visual input. Simulation studies confirmed that the tactile pretraining and contact-event calibration improved performance consistency between simulation and hardware.
What's missing
The study does not discuss potential failure modes or safety considerations for blind grasping in real-world applications, nor does it compare performance against alternative approaches (e.g., vision-based methods or hybrid tactile-visual systems). The 27% success rate's practical implications for industrial or assistive robotics deployment are not addressed.
What different sources said
- arXiv cs.AICenter
Blind Dexterous Grasping via Real2Sim2Real Tactile Policy Learning
Related
Genetic Drift, Not Selection, Drives Rapid Feather Color Evolution in Island Bird Radiation
A new study of an island bird radiation found that rapid evolution of feather coloration is driven primarily by genetic drift in small populations rather than sexual or ecological selection. The research integrated whole-genome data with detailed plumage measurements across complete species sampling to test whether signaling trait evolution correlates with speciation rates. The findings suggest that neutral demographic processes play a central role in generating phenotypic diversity during island radiations, challenging assumptions about the mechanisms driving rapid evolution.
New AI Model Improves Prediction of Therapeutic Peptide Function from Protein Sequences
Researchers developed a lightweight CNN classifier that predicts whether peptide sequences have therapeutic properties, trained on a database of 54,655 peptides across 48 functional categories. The model uses a novel negative sampling strategy to reduce false positive rates from over 60% in previous approaches to 2.1%. This advancement could accelerate drug discovery by enabling faster computational screening of peptide candidates before expensive experimental testing.
Study Shows Different Metabolic Stress Models Produce Distinct Effects on Human Neuronal Networks
Researchers tested three common in vitro metabolic stress models on human-derived neuronal networks and found each produced different patterns of neuronal activity and cell damage. The models tested were hypoxia alone, oxygen-glucose deprivation (OGD), and hypoxia combined with glutamate exposure. The findings suggest that choice of experimental model significantly affects results and that combining electrophysiological and structural analyses is important for accurately assessing metabolic stress in stroke research.