It's not unusual to see robots at the World Artificial Intelligence Conference (WAIC), but GEEKPLUS-W (ASX: 02590) stands alone in demonstrating both multi-SKU generalization capabilities and end-to-end operational efficiency in a genuine work setting.
At the WAIC exhibition, the GEEKPLUS-W booth replicated an actual operational environment. The humanoid robot Gino 1 was shown not only moving boxes and performing picking tasks but also coordinating in a team with mobile robots to ensure a stable picking workflow. This presentation offered one of the few embodied AI holistic solutions that can be directly deployed in the physical AI era.
Humanoid Robot Gino 1: Prioritizing Efficiency and Immediate Deployment
Making its public debut, the versatile humanoid robot Gino 1 was the star attraction of the day's exhibition. Its appeal, however, stemmed not from its appearance but from its performance. On-site observations confirmed that Gino 1 could accurately grasp items from a tote, lift and transport the tote itself, and even form a coordinated fleet with mobile robots to execute the entire picking process from start to finish. This represents one of the very few embodied AI holistic solutions ready for immediate "on-the-job" application.
Efficiency was the most immediately noticeable feature. Unlike traditional robots that typically follow a sequential "grasp, then turn, then move, then lift" pattern, Gino 1 achieves whole-body parallel control with coordinated limb movement. It can move and turn while its arms simultaneously adjust the height of an item and align it with the target bin. According to technical staff at the booth, this whole-body parallel control reduces the time for a single task by approximately 30%. During continuous, uninterrupted demonstrations, Gino 1's movements remained fluid and precise, without a single instance of stalling or dropping an item, showcasing impressive stability.
Crucially, Gino 1 does not operate in isolation. Through the company's AI-powered intelligent dispatching system, it can collaborate in a coordinated group with mobile robots and dedicated workstations to complete the entire process from picking to delivery without human intervention. What is delivered to the user is not the localized efficiency of a single machine, but the systemic efficiency of an entire workflow. This represents the true essence of intelligent productivity.
Mastering the Pick: Versatile and Adaptive Handling
Mastering the "pick" is the first major hurdle for a robot operating in the real world. Items in a warehouse are endlessly varied, and the state of goods within a tote is different every time—stacked, obscured, flush against the wall, tilted, or even suddenly moved. Each reach presents a novel physical scenario. At its core, "picking" is an open-world operational intelligence challenge, simultaneously testing vision, physics, strategy, and error tolerance. GEEKPLUS-W has evidently mastered this problem.
The cross-category generalization ability of Gino 1 was witnessed firsthand. When faced with bags of potato chips inflated to different levels, it dynamically adjusted its grip position and force. It accurately identified the main body of a transparently packaged loaf of bread, distinguishing it from the bag's edge without being confused by the clear film. For a fuzzy-edged plush toy, it could still locate a stable gripping point. For a small, smooth rubber bowl, it maintained a reliable grip through force-control feedback—neither too loose nor too tight. From rigid boxes to soft bags, from transparent packaging to deformable objects, Gino 1 appeared to have "seen" enough of the world to grasp anything it encountered, accurately every time.
Even more astonishing to onlookers was its dynamic adaptability. The booth featured various extreme interference tests. When a staff member suddenly threw a different type of item into the tote, Gino 1 almost instantly reassessed the scene, avoided the non-target object, and continued its current task. When someone repeatedly moved the target item, the robotic arm persistently tracked its position and adjusted its trajectory in real-time until achieving a stable grasp. The most classic moment occurred when a tiny item was pressed tightly into a corner against the tote wall, a position unreachable by a conventional path. Gino 1 independently switched strategies—first using its end effector to "nudge" the item and create a little space from the edge, then successfully completing the grasp. Solving its own corner cases is something no pre-programmed robot can do; it is supported by an intelligent system capable of reasoning about physical consequences and autonomously adjusting its actions.
Embodied Intelligence That Works and Evolves: Seeing is Believing
Faced with real items, real interference, and the pressure of real, continuous operation, Gino 1 ran autonomously based on its own perception, reasoning, and decision-making. This ability to accumulate experience from real-world tasks and grow stronger with use is precisely what the company emphasizes as "self-evolution"—where every success and failure feeds back into the model, making the system increasingly intelligent. While competitors may still be conceptualizing the future, GEEKPLUS-W has placed that future on the exhibition floor and set it to work doing "real jobs."