Why BPTT works now: 50 Hz PD position control keeps SuperMimic gradients bounded
daibond_alpha · x · 2026-11-10
Part 3 of the SuperMimic thread explains why the field abandoned BPTT (DreamerV2/V3 moved to REINFORCE-style gradients, robotics converged on PPO) and why it works now: the control interface changed. 50 Hz PD position control pulls errors back at every step, keeping BPTT gradients bounded, whereas direct torque control makes them explode — the authors point at MuJoCo.
Related event: SuperMimic Trains SOTA Humanoid Tracker in 3 Hours on a Single GPU(5 posts)→
More from Embodied
- LingBot-Touch lifts VLA success rate from 45.6% to 72.5%, pretrained on 4,983 hours of tactile data — CyberRobooo · 2026-11-10
- How LingBot-Touch works: one shared Haptic Transformer unifies magnetic, piezoresistive and visuotactile signals — CyberRobooo · 2026-11-10
- Robbyant releases LingBot-Touch, a tactile model giving robot hands human-like touch sensing — CyberRobooo · 2026-11-10
- Claude Fable 5.2 scores 64.9% on RoboDojo robot control, dominating rivals — burny_tech · 2026-11-10
- Dyna Robotics shows RL vs IK head-to-head on wheeled robot, unveils workflow agent Dyna-2.1 — chris_j_paxton · 2026-11-10
- AI researcher Chris Paxton comes away positive from a Tesla Cybercab ride — chris_j_paxton · 2026-11-10