Low-level controller design shapes whether blind RL robots discover contact behaviors
_krishna_murthy · x · 2026-11-01
krishnamurthy highlights an under-emphasized detail in blind/proprioceptive-only policy learning: the underlying controller matters greatly. Well sys-ID'd impedance controllers can produce rich signals for torque estimation from feedback histories, and blind RL experts implicitly measure force/torque info akin to commanded-vs-realized poses in teleop stacks. Worth exploring whether F/T and tactile sensing unlock qualitatively distinct behaviors, robustness and speed.
Related event: Blind Robot Policies Outperform Expectations, Sparking Controller Debate(4 posts)→
More from Embodied
- mink 1.4.0 released with new Elastic Constraints feature for robot control — kevin_zakka · 2026-11-01
- Naval predicts 4-passenger drones will eventually end commercial air travel, with a long interregnum ahead — naval · 2026-11-01
- Meta Quest game scans your real room and turns it into a haunted house — Scobleizer · 2026-11-01
- RoboAuto: Robot Program Search With Multi-Level Physical Understanding — DJiafei · 2026-11-01
- Developer dresses robot 'Bob' as a ghost to hand out Halloween candy — chrismatthieu · 2026-11-01
- GRASP authors compare with SuperMimic: extracting more optimization signal from sim experience — hankyang94 · 2026-11-01