Roboticist argues full joint space is the wrong abstraction for high-level agents

KyleMorgenstein · x · 2026-11-20

In a discussion with Chen Tessler and Chris Paxton, KyleMorgenstein argues that full joint space is the wrong planning abstraction for higher-level agents: human motion data offers rich priors, but naive joint tracking may not be the right interface. Supporting evidence: many works need 10 future command frames for smooth joint tracking on hardware — a smell of poor formalism, he argues.

Related event: Robotics researcher questions joint-space as planning abstraction for high-level agents(2 posts)→

Original post →

More from Embodied

Embodied channel →