Daxiao Robotics and Nanyang Technological University S-Lab Release World Model Mundo

Related topics top start Related topics top end Futures promotion begin Futures promotion end Miaopai begin Miaopai end Video player start Video player end Market chart begin Market chart end

Recently, Xuhui sci-tech innovation company Daxiao Robotics, together with Nanyang Technological University's S-Lab, publicly released a new world model, Mundo. Simply put, this new technology requires only a small number of photos or short videos to build a 3D virtual environment that supports both free roaming and physical collision interaction, providing a simulation training ground for robots and helping them better adapt to various scenarios in real life.

In the past, many 3D reconstruction technologies could only achieve a result that "looks passable": the view could be switched, but objects would not realistically collide, be pushed or pulled, or fall. Mundo fills this gap. It can write physical parameters such as object mass and friction into the virtual scene. Users can modify the scene through code, for example by pulling open a drawer, pushing tables and chairs, or throwing objects. The physics engine will realistically calculate object motion and collision results, and then render the image in combination with reference imagery, completing the entire process from scene reconstruction and editing to interactive simulation. With just a few photos, Mundo can reconstruct a 3D office space, freely generate new viewpoint images, and support robot simulation training.

In actual use, only 1 to 12 real photographs are needed for Mundo to restore indoor spaces such as offices and home rooms. The model can freely roam along a set camera route and fill in corners that were not captured in the photos; for larger spaces, it can also generate long-shot exploration footage lasting tens of seconds or even minutes. Even a single painting or generated image can serve as a starting point to expand into an entire explorable 3D space.

More importantly, this virtual world is "operable." In the virtual environment, chairs being pushed, objects falling to the ground, and drawers opening and closing all produce motion feedback consistent with real-world logic. Robots can practice navigation and movement, grasping and placing objects in this replicated environment, and the system can also synchronously output first-person and third-person views of the robot, fully preserving motion trajectories and interaction data. R&D personnel can repeatedly debug robot strategies and complete a large amount of training in the virtual environment, reducing the cost of trial and error in the real world. The technical demonstration also achieved a "bullet time" effect, allowing the action to be paused at the moment it occurs and interaction details to be reviewed from multiple angles. A few photos are enough to reconstruct a full home space, and AI freely generates new roaming viewpoints, providing robots with an interactive home simulation training ground.

Mundo is an important supplement to Daxiao Robotics' "Kairos" world model system, focusing on strengthening environment modeling and simulation training capabilities and paving the way for robots to enter real scenarios such as hotels and commercial spaces. The company has attempted to use this technology to reconstruct hotel floor environments, simulate tasks such as opening doors and picking up and placing objects, and conduct robot simulation training.

As an embodied intelligence company rooted in Xuhui, Daxiao Robotics has its headquarters in Xuhui District. Relying on the regional artificial intelligence industry ecosystem, it carries out technology R&D and industry-university-research cooperation, continues to tackle world model technology, is committed to solving pain points in the robotics industry such as insufficient data and weak generalization ability, and promotes the commercial deployment of service robots. With the public release of the Mundo results, the related project page and technical blog have gone online, providing a new reference for industry research. article_adlist[

Here is a news blind box for you

Come and open it and take a look ]article_adlistarticle_adlist[

Reporter: Wang Yongjuan

Editor: Ning Pingying

Please indicate the source when reprinting: Shanghai Xuhui Official Account ]article_adlistarticle_adlist[

]article_adlist news_keyword_pub,stock,sz300024&sh600628 iframe below the main text iframe below the main text General statement General statement QR code at the end start QR code at the end start

QR code at the end start QR code at the end start Editor name and work code Editor name and work code