IN Brief:
- Geek+ has won a 2026 RBR50 Innovation Award for its Robot Arm Picking Station.
- The system has been deployed at Schneider Electric’s Shanghai warehouse.
- The solution combines Geek+ Brain, zero-shot learning, robotic picking, and AMR integration for unmanned item handling.
Geek+ has won a 2026 RBR50 Innovation Award for its Robot Arm Picking Station, following deployment at Schneider Electric’s Shanghai warehouse.
The award recognises the company’s move into automated item picking, a warehouse function that has remained more difficult to automate than storage, internal transport, or goods-to-person presentation. The system combines Geek+ Brain, the company’s embodied intelligence foundation model, with zero-shot learning technology designed to let robotic arms pick across large SKU ranges without individual item retraining.
At Schneider Electric’s Shanghai facility, the robotic arm system has been integrated with existing autonomous mobile robot infrastructure and warehouse workflows. Geek+ says the station doubled manual picking efficiency, achieved accuracy of at least 99.99%, and reached production readiness within 48 hours.
Although AMRs and automated storage systems can bring shelves, totes, or pallets to a workstation, the final picking movement often remains labour-intensive. That handover creates a productivity ceiling, particularly in high-SKU environments where product size, packaging material, surface texture, fragility, and handling requirements vary heavily from order to order.
Traditional robotic picking has faced three persistent barriers: object recognition, gripping reliability, and SKU retraining. Vision systems can struggle with reflective, deformable, or irregular items, while conventional models may require additional training when product catalogues change. Zero-shot learning is intended to reduce that implementation burden, allowing the system to adapt to changing packaging and product mixes without repeated retraining cycles.
The Schneider Electric deployment is especially relevant because industrial warehouses often combine high accuracy requirements with complex inventory profiles. Electrical, automation, and industrial components can be small, fragile, high-value, or visually similar, increasing the cost of picking errors. A robotic arm that can handle varied SKU catalogues without heavy retraining could extend automation into operations where fixed, highly engineered picking systems have previously been difficult to justify.
Warehouse robotics is moving beyond the first large wave of AMR adoption, which focused on reducing travel time and improving labour productivity by changing how goods move through buildings. The next phase is shifting towards decision quality and physical manipulation, with systems that do not only move inventory, but also identify, verify, handle, and place it with minimal human intervention.
Flexible robotics are also being deployed in constrained urban and brownfield environments, as shown by Geek+ and OMLOG’s Hong Kong fashion fulfilment project, where automation was used to raise productivity inside a space-limited warehouse. The Schneider Electric deployment adds another layer by moving from goods-to-person productivity into item-level picking autonomy.
Robots are now expected to deal with the variability of real inventory, real order profiles, and real operating pressure. Item picking sits at the centre of that test because it affects productivity, accuracy, labour availability, and the practical limits of end-to-end automation in older facilities.
Geek+ intends to expand the Robot Arm Picking Station across further industries and use cases. If the system proves repeatable beyond controlled deployments, it could help close one of the most stubborn gaps in automated fulfilment.
