Huawei Cloud and Fulongma Jointly Develop Intelligent Sweeping Robot, Launch Pre-sales, and Open-Source Embodied World Model CloudRobo-WM1.2
Huawei Cloud announced on November 16 that pre-sales have begun for a new-generation intelligent sweeping robot jointly developed with sanitation equipment company Fulongma. Huawei Cloud provides the CloudRobo development toolchain, perception and planning models, the Agentic runtime platform, and Ascend Cloud computing power. On the same day, Huawei Cloud released and open-sourced the Ascend-native embodied world model CloudRobo-WM1.2. The model integrates multimodal information such as RGB video, semantic features, depth, and point clouds, adopts the DualAct architecture, and can directly generate the actions required for robot execution. In September this year, Fulongma had already signed a three-year sanitation autonomous driving technology development contract with Huawei Cloud worth 200 million yuan.
Why read this
AI score 60
Huawei Cloud and Fulongma have jointly launched a sweeping robot and open-sourced a world model, allowing readers to understand its embodied intelligence technology foundation and implementation path.
Topics
Body · AI translation
Huawei is extending its artificial intelligence footprint from the digital world into the physical world.
On November 16, Huawei Cloud announced that a new-generation intelligent sweeping robot jointly developed with sanitation equipment company Fulongma has begun pre-sales. Huawei Cloud will provide the product with the CloudRobo development toolchain, perception and planning models, the Agentic runtime platform, and Ascend Cloud computing power. The robot also incorporates a world model, enabling it not only to recognize its environment but also to simulate and predict future states in complex dynamic physical spaces.
On the same day, Huawei Cloud officially released and open-sourced the Ascend-native embodied world model CloudRobo-WM1.2, further extending into the robotics industry chain.
In September this year, Fulongma disclosed in an announcement that the company had signed a sanitation autonomous driving technology development contract with Huawei Cloud worth 200 million yuan for a term of three years, covering research and development for 3-ton and 6-ton vehicle models as well as sweeping and washing scenarios, with Huawei Cloud exclusively developing and deploying the sanitation autonomous driving system.
Fulongma is not Huawei's only partner in the robotics field. Previously, Huawei Cloud has cooperated with companies including Yijiahe, Topstar, and Youibot, covering scenarios such as power inspection and industrial handling. Huawei has chosen not to directly manufacture robot bodies, but instead to provide models, computing power, and development tools through the CloudRobo platform, seeking to become a shared technology foundation for different robot manufacturers.
The CloudRobo-WM1.2 released this time focuses on enabling robots to move beyond recognizing the current environment toward predicting changes in the physical world. Huawei Cloud said the model integrates multimodal information such as RGB video, semantic features, depth, and point clouds, and adopts the world model's native DualAct architecture to combine world understanding with action generation, enabling the model to directly generate the actions required for robot execution. In addition, through multi-granularity supervised pre-training and cross-embodiment motion modeling, the model aims to reduce reliance on fully annotated data and improve transferability across robots of different forms.
In the inference stage, CloudRobo-WM1.2 offers four composable inference modes to adapt to different computing power conditions and response speed requirements.
However, embodied intelligence still faces many technical obstacles before large-scale commercial application. After robots enter real environments, they need to cope with object occlusion, viewpoint changes, sudden obstacles, and differences between different hardware bodies. At the same time, real-time inference of world models places high demands on computing power and response speed, and how to balance decision accuracy, operational efficiency, and deployment cost remains a problem the industry needs to solve.
The sanitation scenario is no exception. Sweeping robots must not only recognize roads and garbage but also cope with pedestrians, vehicle interference, weather changes, and complex road conditions, while meeting continuous operation requirements. Fulongma also noted in its previous announcement that unmanned sanitation products are still in the demonstration operation and small-scale commercialization stage, and large-scale application still requires continued progress.