BoBDeLaN: Bounded-Bandwidth Adaptation Improves Deep Lagrangian Robot Control

Researchers at TU Darmstadt introduced BoBDeLaN, which combines deep Lagrangian networks with stability filters and bounded-bandwidth context adaptation to improve robustness of learned robot dynamics models against disturbances such as unknown payloads.

2026-11-21 ~ 2026-11-21 · 2 related posts