ContactNets: Learning Discontinuous Contact Dynamics with Smooth, Implicit Representations
This addresses the challenge of accurately modeling discontinuous contact dynamics in robotics, which is crucial for simulation, control, and planning, representing a novel method for a known bottleneck rather than an incremental improvement.
The paper tackled the problem of unrealistic model predictions for robot dynamics during discontinuous contact events like impacts and stiction by introducing ContactNets, a method that learns smooth, implicit representations of contact dynamics. The result was the ability to predict realistic impact, non-penetration, and stiction with training on only 60 seconds of real-world data.
Common methods for learning robot dynamics assume motion is continuous, causing unrealistic model predictions for systems undergoing discontinuous impact and stiction behavior. In this work, we resolve this conflict with a smooth, implicit encoding of the structure inherent to contact-induced discontinuities. Our method, ContactNets, learns parameterizations of inter-body signed distance and contact-frame Jacobians, a representation that is compatible with many simulation, control, and planning environments for robotics. We furthermore circumvent the need to differentiate through stiff or non-smooth dynamics with a novel loss function inspired by the principles of complementarity and maximum dissipation. Our method can predict realistic impact, non-penetration, and stiction when trained on 60 seconds of real-world data.