Phy-Tac: Toward Human-Like Grasping via Physics-Conditioned Tactile Goals

2025/11/03 by Shipeng Lyu, Sheng Lu, Lyu, Shipeng +13
Engineering · #Advanced Sensor and Energy Harvesting Materials #FOS: Computer and information sciences #Muscle activation and electromyography studies #Robot Manipulation and Learning #Robotics (cs.RO)

paper · pdf · doi:10.48550/arxiv.2511.01520

openalex publication_date 2025/11/03 · openalex created_date 2025/11/06 · openalex updated_date 2026/07/28

Abstract

Humans naturally grasp objects with minimal level required force for stability, whereas robots often rely on rigid, over-squeezing control. To narrow this gap, we propose a human-inspired physics-conditioned tactile method (Phy-Tac) for force-optimal stable grasping (FOSG) that unifies pose selection, tactile prediction, and force regulation. A physics-based pose selector first identifies feasible contact regions with optimal force distribution based on surface geometry. Then, a physics-conditioned latent diffusion model (Phy-LDM) predicts the tactile imprint under FOSG target. Last, a latent-space LQR controller drives the gripper toward this tactile imprint with minimal actuation, preventing unnecessary compression. Trained on a physics-conditioned tactile dataset covering diverse objects and contact conditions, the proposed Phy-LDM achieves superior tactile prediction accuracy, while the Phy-Tac outperforms fixed-force and GraspNet-based baselines in grasp stability and force efficiency. Experiments on classical robotic platforms demonstrate force-efficient and adaptive manipulation that bridges the gap between robotic and human grasping.

Citations

• Learning Gentle Grasping Using Vision, Sound, and Touch
• Towards Generalization of Tactile Image Generation: Reference-Free Evaluation in a Leakage-Free Setting
• Vision-based Tactile Image Generation via Contact Condition-guided Diffusion Model
• Touch2Touch: Cross-Modal Tactile Generation for Object Manipulation
• TextToucher: Fine-Grained Text-to-Touch Generation
• PhyGrasp: Generalizing Robotic Grasping with Physics-informed Large Multimodal Models
• A Touch, Vision, and Language Dataset for Multimodal Alignment
• DexDiffuser: Generating Dexterous Grasps with Diffusion Models
• Binding Touch to Everything: Learning Unified Multimodal Tactile Representations
• PixArt-α: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis
• Generating Visual Scenes from Touch
• 9DTact: A Compact Vision-Based Tactile Sensor for Accurate 3D Shape Reconstruction and Generalizable 6D Force Estimation
• Planning Visual-Tactile Precision Grasps via Complementary Use of Vision and Touch
• Touch and Go: Learning from Human-Collected Vision and Touch
• DexGraspNet: A Large-Scale Robotic Dexterous Grasp Dataset for General Objects Based on Simulation
• PoseIt: A Visual-Tactile Dataset of Holding Poses for Grasp Stability Analysis
• Deep Learning Approaches to Grasp Synthesis: A Review
• High-Resolution Image Synthesis with Latent Diffusion Models
• Goal-Driven Robotic Pushing Using Tactile and Proprioceptive Feedback
• Denoising Diffusion Implicit Models
• Denoising Diffusion Probabilistic Models
• Connecting Touch and Vision via Cross-Modal Prediction
• 6-DOF GraspNet: Variational Grasp Generation for Object Manipulation
• Robust Learning of Tactile Force Estimation through Robot Interaction
• Dense Tactile Force Distribution Estimation using GelSlim and inverse FEM
• Conditional Generative Adversarial Nets

Related

• Tactile-Conditioned Diffusion Policy for Force-Aware Robotic Manipulation
• Feel the Force: Contact-Driven Learning from Humans
• Embedding high-resolution touch across robotic hands enables adaptive human-like grasping
• Tracing Energy Flow: Learning Tactile-based Grasping Force Control to Prevent Slippage in Dynamic Object Interaction
• MAT: Multi-Fingered Adaptive Tactile Grasping via Deep Reinforcement Learning