Fundamental Challenges in Deep Learning for Stiff Contact Dynamics
This highlights a fundamental challenge for robotics applications like legged locomotion and manipulation, as current simulation environments may not reflect real-world stiffness, potentially leading to poor model performance.
The paper tackles the problem of deep learning's inability to accurately model stiff frictional contact dynamics, which are crucial for robotics, and finds that increasing stiffness degrades training, generalization, and data-efficiency in simulations.
Frictional contact has been extensively studied as the core underlying behavior of legged locomotion and manipulation, and its nearly-discontinuous nature makes planning and control difficult even when an accurate model of the robot is available. Here, we present empirical evidence that learning an accurate model in the first place can be confounded by contact, as modern deep learning approaches are not designed to capture this non-smoothness. We isolate the effects of contact's non-smoothness by varying the mechanical stiffness of a compliant contact simulator. Even for a simple system, we find that stiffness alone dramatically degrades training processes, generalization, and data-efficiency. Our results raise serious questions about simulated testing environments which do not accurately reflect the stiffness of rigid robotic hardware. Significant additional investigation will be necessary to fully understand and mitigate these effects, and we suggest several avenues for future study.