AXIS: A Crowdsourced Web Simulation Data Engine for Scalable Robot Manipulation
机器之心 · wechat · 2026-08-02
AxisRobotics and academic partners introduced AXIS, a community-driven, continuously growing data engine and benchmark for robot manipulation. Users can teleoperate robot arms via a web browser to generate demonstration data.
Core Design & Data Processing
- Scale: The current fixed snapshot includes 207 tasks and over 50k trajectories; the live engine has amassed over 1.51 million trajectories to date.
- Pipeline: To handle noise from crowdsourced data, AXIS applies a standardized pipeline (filtering, smoothing) and performs augmentation in IsaacSim by randomizing camera views, lighting, and physics parameters.
Key Findings
- The team pre-trained the π0.5 base model on varying scales of AXIS data and evaluated it on the LIBERO-Plus benchmark.
- Results show steady performance gains as data increases, with the full dataset improving overall success rates by 4.9 percentage points over the baseline.
- Under an equal trajectory budget, the AXIS matched group outperformed RoboCasa365 by 31.3 percentage points, highlighting the value of task coverage and diversity.
More from Embodied
- Peking University's SimForcing distills simulation motion priors into robot world models — PekingUniversity · 2026-10-16
- NVIDIA and Mercedes-Benz Expand Partnership on Next-Gen Automated Driving — nvidia · 2026-10-16
- ~25,000 Humanoid Robots Shipped in H1 2026, Over 95% from Chinese Vendors — ingliguori · 2026-10-16
- openpilot brings stoplight-stopping autonomy to a 2019 Corolla — k7agar · 2026-10-16
- Hironobu Fujiyoshi publishes deck: from AI agents to Agentic Robotics — HirokatuKataoka · 2026-10-16
- CoDA co-evolves simulation and robot policies via shared failure memory, lifting OOD success from 5.3% to 27.6% — PLAN-Lab · 2026-10-16