On September 9, at JD Discovery (JDD), JD Logistics presented the latest full lineup of its “Super Brain + Wolf Clan” robot application army. The “Super Brain” large model is responsible for intelligent decision-making and collaborative scheduling in complex supply chain scenarios. The “Wolf Clan” has already deployed 9 products and 11 robots in warehousing, sorting, and delivery, with 5 new products and technologies launched at the event.
Logistics is one of the most promising scenarios for robotics. JD Logistics says it aims to build the world’s largest embodied robot application army and plans to procure 3 million robots, 1 million unmanned vehicles, and 100,000 drones within five years, advancing full-process unmanned logistics. This means robots are no longer just point automation devices; they must enter a continuously operating supply chain network.
The real difficulty lies here. From warehouse entry to final delivery, a parcel passes through six stages: storage, picking, palletizing, sortation, transportation, and delivery. In traditional models, equipment at each stage does not “talk” to the others, making unified scheduling even harder. Enabling dozens of robot types and hundreds of machines with different functions to operate simultaneously and collaborate efficiently tests global decision-making, not merely a single unit’s ability to grasp or walk.
To turn the “Wolf Clan” from isolated fighters into a unified force, JD Logistics uses Super Brain large model 3.0 as the commander. Five supply-chain-native model categories—forecasting, decision-making, multimodal, spatiotemporal, and embodied—work together to first calculate a globally optimal plan, then break decisions into executable instructions for each device, and collect execution feedback in real time. The conference also launched new products including Warehouse Wolf, Low-Temperature Smart Wolf, Mother Wolf Health Edition, Lone Wolf Gen 6, and Flying Wolf L05, covering warehousing, cold chain, retail fulfillment, transportation, and delivery.
The core challenge of this system is essentially the same as scheduling on a computing power platform: if the preceding stage is fast but the following stage cannot keep up, substantial waiting and idling occur. JD Logistics’ approach is to let decision models calculate the optimal rhythm for each stage in real time and dynamically recalculate and synchronize instructions when orders, road conditions, or capacity change. From another angle, StarWar Technology faces a similar problem in GPU computing platforms and compute scheduling—organizing distributed computing resources into a unified resource pool so tasks can be efficiently coordinated by priority and actual load.
According to published data, the system has been deeply applied in more than 1,000 scenarios, improving frontline operational efficiency by nearly 20%. The “Wolf Clan” has been deployed in more than 20 provinces in China and more than 10 countries worldwide. From an industry perspective, the scaling of Physical AI is moving from “devices can work” to “legions can coordinate,” and the unified decision-making and scheduling layer is becoming a high-value battleground.
For the supply chain, the next phase of embodied AI is not about adding a few more prototypes, but about whether systems can operate stably under the real load of hundreds of millions of parcels per day. As hardware gradually converges, the true dividing line will be the efficiency of connection among decision AI, process AI, and physical AI.