Learning Transferable Friction Models and LuGre Identification via Physics Informed Neural Networks

2025-04-16

AI Summary

Friction modeling mismatch in robotic simulation constitutes a critical bottleneck underlying the sim-to-real performance gap. This paper proposes a physics-informed neural network (PINN)-based framework for transferable friction estimation, enabling, for the first time, end-to-end differentiable identification of LuGre dynamic friction model parameters. By synergistically integrating LuGre’s physical priors with few-shot, noise-robust data-driven learning, the method accurately reconstructs nonlinear, underactuated system dynamics using only minimal noisy sensor measurements. Unlike heuristic friction models in conventional simulators (e.g., MuJoCo, PyBullet), our approach ensures both physical interpretability and strong generalization—supporting cross-system parameter transfer without retraining. Experimental validation on real hardware demonstrates substantial reduction in sim-to-real discrepancies across diverse robotic platforms, thereby bridging the fidelity gap between simulation and physical deployment.

Technology Category

• Intelligent Robots: State Estimation
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior
• Computer Vision: Low Level & Physics-based Vision

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Web Mining and Content Analysis: Web data generation and simulation
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Accurately modeling friction in robotics remains a core challenge, as robotics simulators like Mujoco and PyBullet use simplified friction models or heuristics to balance computational efficiency with accuracy, where these simplifications and approximations can lead to substantial differences between simulated and physical performance. In this paper, we present a physics-informed friction estimation framework that enables the integration of well-established friction models with learnable components-requiring only minimal, generic measurement data. Our approach enforces physical consistency yet retains the flexibility to adapt to real-world complexities. We demonstrate, on an underactuated and nonlinear system, that the learned friction models, trained solely on small and noisy datasets, accurately simulate dynamic friction properties and reduce the sim-to-real gap. Crucially, we show that our approach enables the learned models to be transferable to systems they are not trained on. This ability to generalize across multiple systems streamlines friction modeling for complex, underactuated tasks, offering a scalable and interpretable path toward bridging the sim-to-real gap in robotics and control.

Problem

• Accurately modeling friction in robotics simulations
• Reducing sim-to-real gap with learnable friction models
• Enabling transferable friction models across untrained systems

Innovation

• Physics-informed neural networks for friction modeling
• Transferable models trained on small noisy datasets
• Combines physical consistency with real-world adaptability