Youlu Robotics released three property-service robots on 20 November 2026 — the AI130S for outdoor sweeping, the AI60P for patrol and the AI70W for deep floor cleaning — all running a common embodied-AI controller the company calls LPLM, and used the launch to publish a plan for 108 robot models built with manufacturing and channel partners.
The pitch is not a better sweeper. It is a split of the robot value chain: Youlu keeps the brain and the platform, contract manufacturers build the hardware, and property management companies co-develop models for their own sites. In a category where every operator buys a different mix of sweeping, patrol and cleaning machines from different vendors, standardising on one controller and one fleet-management layer is the actual product.
Property robotics in China has been a hardware race with thin margins: dozens of vendors, similar chassis, and buyers who compare on price per square metre cleaned. Youlu's chairman Gu Zulin is proposing the opposite structure — the company supplies the autonomy layer and the fleet platform, manufacturing partners take the mechanical build, and property operators develop products for their own scenarios. Green Source Group, a two-wheeler manufacturer, has already signed on as the first hardware partner.
If it works, the economics shift from selling machines to collecting a per-robot platform fee across a much larger installed base, and the capital burden of tooling moves to partners who already have factories. If it does not, Youlu ends up as a software vendor with no control over build quality, which in a category where a single malfunction in a public space becomes a liability claim is a real risk.
The 18,000-hour benchmark is the more quietly useful contribution. Property operators have no standard way to compare reliability claims across vendors, and MTBF figures published by robot makers are rarely measured under the same conditions. A stated duty cycle and duration gives buyers something to hold contracts against, even if the number itself is aspirational.
No unit pricing, no delivery dates and no per-model autonomy specifications were released. LPLM's capability is described rather than demonstrated, and the 108-model target is a plan, not a pipeline. The partner-built model also raises an unresolved question of who owns the safety certification on a robot that one company designed, another built and a third operates.
For overseas buyers, the relevant fact is that Chinese property robot makers are now competing on platform economics rather than hardware cost. That tends to compress prices faster than feature lists suggest.
Reported by TMTPost, 20 November 2026. Chinese-language sources only; no English-language coverage of this launch was found at the time of writing.