On September 21, ACE Robotics (大晓机器人) and PetroChina Shanghai Marketing Company announced a strategic partnership to place embodied intelligent robots into gas station convenience stores for a "full closed-loop trial operation." It is described as the first robot project landed in a gas station convenience store scenario in China, pushing robots from "demonstrations in flagship stores" to "working shifts in real 24-hour outlets." The cooperation follows a four-part framework: scenarios set the problems, platforms coordinate, models drive decisions, and robot bodies execute.
The Scenario: Gas Station Convenience Stores
PetroChina Shanghai operates a citywide network of 24-hour gas station convenience stores with random foot traffic, dense shelving, and unpredictable order peaks. This stresses robots on three fronts simultaneously:
- Product recognition: many SKUs with similar packaging
- Picking and transport: narrow aisles, tall shelves, precise lifting required
- Order fulfillment: sharp demand peaks, low error tolerance, coordination with human cashiers
- Chairman: Wang Xiaogang (co-founder and executive director of SenseTime)
- Chief Scientist: Academician Tao Dacheng (member of the Australian Academy of Science)
- Standardization: highly consistent store layouts and operations, easy to replicate
- Long-tail coverage: highway, suburban, and remote locations that chain convenience stores usually skip
- 24-hour operation: overnight shifts are hard to staff; robots can work continuously
- AgiBot Zhiyuan A3 Ultra: multi-robot collaboration in car dealerships, hotels, and convenience stores (shelf restocking, inventory counting, heavy-item transport)
- Galbot G1 (Galaxy General): entered a FamilyMart in Beijing's Zhongguancun in April, handling grilled sausages, chilled drinks, coffee, and hot food
- JD "Intern" robots: entered JD Mall stores in June for greeting, guidance, and shelving; the JoyAI-RA model achieves 90%+ single-item pick success
- Scale: the pilot did not disclose store or robot counts; real evidence awaits a review of the first 10+ stores
- Cross-store generalization: can picking behavior learned in store A transfer directly to store B?
- After-sales and spares: manual takeover procedures, spare-part networks, and maintenance cycles are hidden costs of real deployments
- Regulation and safety: robots working near fueling areas face stricter electromagnetic-compatibility and explosion-proof requirements than ordinary convenience stores
ACE's hardware for this deployment is a self-developed wheeled robot with an manipulator arm and lift mechanism, combining a mobile base, arm manipulation, and vertical reach covering both low and high shelves. On the software side it runs Kairos 3.0 (开悟世界模型 3.0), iterated from China's first open-source and commercially available world model. The signal chain reads orders and inventory data from the store system, and the model generates execution sequences for recognition, picking, transport, and packing.
About ACE Robotics
The company spun off from SenseTime's robotics division and began operating as ACE Robotics in 2026. Its board includes:
The company's "ACE R&D paradigm" chains three elements: ambient data collection → Kairos 3.0 world model → generalizable embodied modules. Its Embodied Super Brain Module A1 packages visual perception, task planning, and a skill library as a "universal brain" that can run on different robot bodies across security, inspection, service, retail, logistics, manufacturing, and public-sector scenarios.
Why Gas Stations Were Chosen
The announcement cites roughly 110,000 gas stations plus 100,000+ public charging stations nationwide — a dense network of offline energy nodes with:
By handing robots the "night shift + repetitive labor + standardized workflows" bundle, the pilot tests whether they can run a full shift without human takeover.
How This Differs from Earlier Embodied Deployments
Since September, several Chinese vendors have moved robots into retail:
ACE's deployment differs in two ways: the scenario extends from "storefront core areas" to the back warehouse and fulfillment chain of gas station stores, and the mode shifts from "robot demos" to robots joining a closed-loop fulfillment pipeline wired into the order system.