Amap's ABot-Earth 0.7 Generates a Roamable 3D City in 10 Minutes from One Satellite Image
量子位 · wechat · 2026-09-11
Amap unveiled the fully AI-powered 2026 Street Ranking and detailed ABot-Earth 0.7, billed as the world's first 3D-native urban world model: one satellite image plus 10 minutes on a consumer GPU yields a roamable 3D city, 1000x faster than traditional reconstruction, already covering 300+ cities in 190+ countries.
- Native 3DGS generation with a compress-to-latent-space-then-decode scheme; training data rebuilt from satellite/aerial/street imagery via in-house ABot-3DGS and rigorous cleaning
- LOD decoding plus sliding-window blending cut stitching artifacts by 97%; single inferences span 10+ km²; landmarks use dedicated 3D models auto-calibrated by agents
- One causal Transformer architecture spans renderer (ABot-World/Earth), simulator (traffic/crowd forecasting), and planner (ABot-Navigation/Manipulation embodied models), with 20+ SOTA results
- CEO Guo Ning frames spatial intelligence as four layers — 2D topology, 3D space, time, real-world flows — with ABot-Earth covering the 3D layer so far
More from Embodied
- UMO: unified in-context framework turns text2motion model into multi-task motion foundation — Michael_J_Black · 2026-09-11
- NTU spin-off Ropedia raises $22M to train robots from wearable first-person data — liuziwei7 · 2026-09-11
- Why robotics folks stay quiet on AGI: dishes and laundry first, hype later — MannyKayy · 2026-09-11
- Maxinsights: The Data Supplier Behind Silicon Valley's Hottest Robots, With 2M+ Hours of Egocentric Data — 新智元 · 2026-09-11
- Inside China Opto-electronics Expo: The XR Supply Chain, From Waveguides to MicroLED — Scobleizer · 2026-09-11
- Tesla shows Cybercab in Japan, its steering wheel-free robotaxi built from scratch — yunta_tsai · 2026-09-11