Leju and Zhaofeng Launch Auto Parts Loading Solution with 15 Humanoid Robots

On October 15, Leju and Zhaofeng Electromechanical jointly launched an embodied AI solution for automotive parts loading at Zhaofeng’s Hangzhou factory, announcing that 15 humanoid robots are now in routine operation. The robots work across different stations on the auto parts production line, handling identification, grasping, transfer and loading tasks.

According to Leju, the 15 robots cover 15 stations and more than 10 types of parts, with each loading cycle completed in under 40 seconds and a success rate of about 98%. The project has been running for over two months, accumulating roughly 1,000 hours of operation and helping produce more than 60,000 parts.

Leju CEO Chang Lin said the deployment of 15 robots in the factory environment is “not a demo,” and the company hopes to expand the validated solution to more production lines. He noted that the current loading pace of about 35 seconds still leaves room for improvement compared with skilled workers. Zhaofeng Chairman Kong Aixiang said humanoid robots would be more ready for industry-wide adoption if they can reach a cycle time of about 15 seconds, a goal he expects could take half a year to a year.

In real deployment, the robots must cope with changing lighting conditions, part shapes and placement positions, and they need to be connected to the factory’s production system. Leju has also included data feedback in the solution: the robots collect data during loading, and records such as failed grasps or dropped parts are used for model training. The two companies donated the real-robot loading dataset to the OpenLET community under the OpenAtom Open Source Foundation.

To improve skill reuse, Leju split the loading skill into more than 30 basic modules and provided a development toolchain and a multi-machine scheduling platform. Zhaofeng plans to explore additional applications such as inspection, quality checks and equipment adjustment.