Two-month-old startup's humanoid robot drives a go-kart at full speed on a real track
量子位 · wechat · 2026-08-23
A viral video shows a bipedal humanoid robot driving a go-kart at racing speed — steering, cornering and avoiding obstacles in one take — from a two-month-old startup called 共生知行 (Gongsheng Zhixing). Founder Ding Pengxiang, born in 1996, leads an all-PhD-student team born after 2000 that has won two Best Paper awards in embodied AI and built China's first embodied base model with 2K+ GitHub stars.
The team bets on a purely end-to-end approach, more radical than the layered brain/cerebellum design used by Figure: the model directly outputs joint targets, combining task behavior modeling with motion prior distillation (DriftDistill) so a single model handles both task intelligence and body stability, with a 1B MotionExpert and a ForceExpert for safe contact being distilled in. For data, their TrajBooster extracts end-effector trajectories from robot arms and lets humanoids track them in simulation to cheaply scale whole-body mobile manipulation data.
Ding is sober about timelines: 100K hours of data still yields lab demos only, and commercially meaningful scale likely requires millions of hours; the hardest open problems are how to combine perception and control and what data modality to use.
More from Embodied
- SEAR: first large-scale controlled study of self-evolving robot agents, 7 LLMs, 330 tasks — xiye_nlp · 2026-10-16
- Podcast: subsea robotics startup Ulysses on operating 1,500m underwater and ocean carbon — blakemenezes · 2026-10-16
- Sony unveils $5,900 gold-plated Walkman with AI music enhancement — Polymarket · 2026-10-16
- ROBIB Meetup #3 returns to Ghent on Nov 19 with Physical AI talks, demos and panel — MarwaEldiwiny · 2026-10-16
- VLX-Go: Multi-Stage Vision-Language Training Boosts Robot Person-Following Among Distractors — omlab · 2026-10-16
- Ledger Atom Debuts as Agent-Approval Security Device; Critic Says a $100 iPhone 8 Beats It — adamamcbride · 2026-10-16