Itshi shows a full embodied AI model driving an automotive wiring line at WAIC 2026
量子位 · wechat · 2026-07-22
Itshi Robotics used WAIC 2026 to show a 1:1 automotive wiring-harness line where multiple robots collaborate on different stations in parallel.
The company says its new embodied-native foundation model AWE 3.5 is the first to fully connect pretraining and post-training in a single embodied pipeline. Key claims include:
- A OneModel design that can act as both a policy model and an action-conditioned world model
- Training on 1M+ hours of human-centric data
- Explicit support for long-term memory and spatiotemporal understanding in multi-minute closed-loop interactions
- A model-internal simulation loop that lets the system imagine actions like zipper pulling before trying them on real hardware
The article argues that embodied AI is moving from “robot demos” to “robots doing real work,” and quotes Chief Scientist Ding Wenbo saying the field’s next phase will be defined by scalable data, infra, and clear industrial feedback loops.
Ding also says:
- Dexterous hands matter more than expected; grippers solve about 80% of tasks, but the last 20% is hard
- New tasks can often be learned with only hours of data once the base model is strong enough
- The field is entering a Scaling + Scaling Research stage, where industrial competition will increasingly narrow purely academic room for novelty
More from Embodied
- Johns Hopkins Launches Full-Stack Hands-on Robot Learning Class with SO-101 Arm Kits — _krishna_murthy · 2026-09-11
- SyncWorld: In-Context Robot World Model Simulates Unseen Views and Embodiments Zero-Shot — ChongZzZhang · 2026-09-11
- A 3D Pose Dataset for Dogs Released — ducha_aiki · 2026-09-11
- Swaayatt demos autonomous driving at 52 km/h on mountain roads, self-recovers after skid — sanjeevs_iitr · 2026-09-11
- AUAR's MicroFactory brings a deployable robotic wood-panel factory to the construction site — lukas_m_ziegler · 2026-09-11
- Musk: Cybercab certified at 165 Wh/mi, the most efficient production EV ever — elonmusk · 2026-09-11