BoBDeLaN: Combining Deep Lagrangian Networks and Stable Filters for Robust Robot Control

Jan_R_Peters · x · 2026-11-21

Lucas Schulze et al. introduce BoBDeLaN (Bounded-Bandwidth Context Adaptation for Deep Lagrangian Robot Control), tackling how to make learned physics-encoded models robust to disturbances in robot control. The approach combines Deep Lagrangian Networks with Recurrent Equilibrium Networks (REN) and stable filters, using bounded-bandwidth context adaptation to handle perturbations — part of the line of work injecting physical structure into learned robot control.

Related event: BoBDeLaN: Bounded-Bandwidth Adaptation Improves Deep Lagrangian Robot Control(2 posts)→

Original post →

More from Embodied

Embodied channel →