Simultaneous Learning of Contact and Continuous Dynamics

TL;DR

This paper introduces a novel method for robots to learn contact and continuous dynamics of new objects from motion data, improving data efficiency and handling complex effects like joint friction and stiff contact forces.

Contribution

It presents a system identification approach that infers contact forces and dynamics simultaneously, outperforming existing differentiable simulation methods on real and simulated datasets.

Findings

Outperforms differentiable simulation and end-to-end methods

More data-efficient in learning complex contact dynamics

Effectively models joint friction and stiff contact forces

Abstract

Robotic manipulation can greatly benefit from the data efficiency, robustness, and predictability of model-based methods if robots can quickly generate models of novel objects they encounter. This is especially difficult when effects like complex joint friction lack clear first-principles models and are usually ignored by physics simulators. Further, numerically-stiff contact dynamics can make common model-building approaches struggle. We propose a method to simultaneously learn contact and continuous dynamics of a novel, possibly multi-link object by observing its motion through contact-rich trajectories. We formulate a system identification process with a loss that infers unmeasured contact forces, penalizing their violation of physical constraints and laws of motion given current model parameters. Our loss is unlike prediction-based losses used in differentiable simulation. Using a new dataset of real articulated object trajectories and an existing cube toss dataset, our method outperforms differentiable simulation and end-to-end alternatives with more data efficiency. See our project page for code, datasets, and media: https://sites.google.com/view/continuous-contact-nets/home