Hydrosoft: Non-Holonomic Hydroelastic Models for Compliant Tactile Manipulation
Miquel Oller, An Dang, Nima Fazeli
AI summary
Problem
Capturing the nonlinear, path-dependent dynamics of compliant tactile sensors remains a challenge, as existing models are either computationally prohibitive or ignore critical effects like static friction and deformation history. This limits precise force regulation and continuous contact control in dexterous manipulation.
Approach
The method extends hydroelastic modeling to incrementally track path-dependent contact forces and Coulomb friction, then applies smoothing to ensure differentiability. These dynamics are embedded in a quasi-dynamic system and solved via gradient-based model predictive control.
Key results
- Accurately models path-dependent force buildup and dynamic contact area variations
- Enables smooth gradient propagation through non-smooth contact transitions via force relaxation
- Successfully executes planar pushing, rotation, rolling, and in-hand reconfiguration in simulation and hardware
- Demonstrates that ignoring path dependence degrades control performance in shear-heavy tasks
Why it matters
Provides a scalable, differentiable framework for real-time control of dexterous robots using compliant tactile feedback, bridging the gap between physical realism and computational efficiency.
Abstract
Tactile sensors have long been valued for their perceptual capabilities, offering rich insights into the otherwise hidden interface between the robot and grasped objects. Yet their inherent compliance—a key driver of force-rich inter- actions—remains underexplored. The central challenge is to capture the complex, nonlinear dynamics introduced by these passive compliant elements. Here, we present a computationally efficient non-holonomic hydroelastic model that accurately models path-dependent contact force distributions and dynamic surface area variations. Our insight is to extend the object’s state space, explicitly incorporating the distributed forces gen- erated by the compliant sensor. Our differentiable formulation not only accounts for path-dependent behavior but also enables gradient-based trajectory optimization, seamlessly integrating with high-resolution tactile feedback. We demonstrate the effectiveness of our approach across a range of simulated and real-world experiments and demonstrate the importance of modeling the path dependence of sensor dynamics.