Robots That Combine in 3 Seconds: Chinese Firm Sells Modular Machines to 50 Countries

Benmo Technology, founded in 2020, has developed a line of direct-drive wheeled robots that can operate independently or combine into larger configurations. Its D1 model, described as the world's first fully modular embodied intelligent robot, can split or merge within three seconds, switching between two-wheeled, four-wheeled, electric-foot and flat-foot forms depending on the task.

Videos posted by the company show two units aligning and joining, then standing up and moving as one. The design theoretically allows unlimited units to connect into longer chains. While humanoid robots often grab attention with performances, Benmo's machines are already handling real jobs such as carrying water buckets, delivering documents and patrolling warehouses.

The company was founded by Zhang Di, who studied mechanical engineering at Beijing Institute of Technology and later robotics at the Hong Kong University of Science and Technology under Li Zexiang, known as the "godfather of DJI." Benmo specializes in integrated direct-drive joints without reducers, a choice that reduces noise and extends lifespan to 50,000 to 100,000 hours.

The firm operates two business lines: power modules and complete robots. Its direct-drive power modules shipped about 8.5 million units in 2025, accounting for 61.1 percent of China's consumer robotics direct-drive module market. Its wheeled robots have been exported to more than 50 countries and regions, making it the world's top shipper of wheeled robots.

The company argues that while much of the industry channels investment into "brains" - AI models and planning algorithms - a robot still needs a reliable "body" to reach where work happens. Obstacles such as steps, gravel and heavy loads remain physical constraints that no model alone can solve. Benmo says its approach is to perfect the body and lower-level control, paving the way for advanced AI to operate in the physical world.

The D1's modularity also serves practical needs. For light tasks like map scanning, two units can split and run in parallel. For climbing stairs, they merge for stability. Developers can use one machine for both two-wheeled and four-wheeled testing, cutting equipment costs in half because both modes share a single algorithm framework.

While humanoid robots are still largely in the demo phase, Benmo's wheeled machines are already in use across parks and campuses, hauling 80-kilogram loads, delivering files in office buildings and splitting into smaller units for inspection work. The company says that as the industry waits years for large-scale embodied AI models to mature, advancing the physical chassis is essential to bringing autonomous machines closer to full deployment.