Talk with Zhang Hanwen of ZhiXing Tech: Charging Robots and L4 Autonomous Driving Arrive Simultaneously, 2028 is a Mass Production Milestone

From September 16 to 17, the first Aggregated Intelligence Industry Development Conference (2025) was held in Wuhan. Leading enterprises and research institution heads from multiple sectors, including intelligent robots, low-altitude flight, and the automotive supply chain, gathered at the event to discuss the development of the aggregated intelligence industry.

Nowadays, many manufacturers have become cross-boundary players in both the smart car and intelligent robot industries. Zhixing Technology is one of the earliest tier-one suppliers to make such布局 (layout/investment).

On the morning of September 17, Zhang Hanwen, General Manager of the Ecosystem for the Robotics Division at Zhixing Technology, delivered a speech titled "The Reuse of the Knowledge Chain from Intelligent Driving to Embodied AI" at the conference. He detailed Zhixing Technology's layout in the field of intelligent automobiles and the latest progress in its intelligent robotics this year.

After the speech, Zhang Hanwen accepted interviews from several industry media outlets, including 42nd Wave. He answered questions one by one regarding Zhixing Technology's ongoing layouts in areas such as intelligent charging robots and the reuse of technologies between smart cars and robots.

Zhang Hanwen pointed out that Zhixing Technology entered the robotics industry based on the high adaptability between intelligent driving and embodied AI in terms of technology chains and value chains. In the area of assisted driving controllers, Zhixing Technology expects to mass-produce over 650,000 vehicle-mounted pre-controllers this year, covering mainstream domestic and international chip platforms such as Horizon, Texas Instruments, and Renesas. The brain-body collaborative architecture relied upon by these controllers aligns highly with the control system requirements for robots, providing the possibility for technology reuse.

At the algorithm level, Zhixing Technology already possesses extensive experience in mass deployment—not only being the first company globally to implement BEV perception parking functions on Texas Instruments' 8 TOPS computing platform, but also the sole supplier to deploy integrated driving and parking perception capabilities on Renesas' 28 TOPS platform. This engineering experience in achieving high-performance algorithms on low-computing-power platforms has laid a solid foundation for the transplantation and optimization of algorithms in the robotics field.

To achieve technical validation and commercial closed-loop operations, ZhiXing Technology acquired Xiaogongjiang, a robotics company specializing in integrated joint R&D, earlier this year, and established the Aimoxing Robotics Business Unit based on this acquisition.

In June this year, ZhiXing Technology collaborated with Digua Robotics to successfully develop China's first mid-to-high compute robotics controller. The prototype has recently rolled off the production line and been delivered to leading robotics customers for real-world testing, providing the industry with a high-performance, low-cost, and independently controllable domestic solution.

In terms of algorithms, the ZhiXing Technology team has made significant progress by successfully deploying the world-class open-source model ACT Policy onto HuggingFace's LeRobot robotic arm, ensuring stable operation on edge-side Horizon chip platforms. This marks the first global instance of such model deployment on non-NVIDIA edge computing platforms.

Leveraging its technical accumulation, ZhiXing Technology has launched an automated robot charging solution tailored for the autonomous driving ecosystem. At the 2025 Munich Motor Show in Germany, the company showcased a six-axis collaborative robotic arm equipped with its proprietary monocular visual pose detection algorithm, achieving a charging plug insertion and extraction efficiency of 19.5 seconds, significantly outperforming other industry solutions.

Regarding the timeline for the widespread adoption of automatic charging robots, Zhang Hanwen believes they will arrive simultaneously with L4-level autonomous driving, thereby adding greater value to automatic charging; furthermore, information from multiple leading OEMs indicates that 2028 will be a key node for mass production.

Zhang Hanwen emphasized that intelligent driving and robotics share highly consistent underlying technical logic, both relying on technology chains such as BEV, Transformer architectures, and OCC. While similarities exist in sensor configurations, algorithmic architectures, and execution logic, the main differences lie in the granularity of visual perception and interaction scenarios: intelligent driving focuses on long-distance, road-environment perception, whereas robotics prioritizes fine-grained operations in close-range, enclosed scenarios.

This technological homology provides Flexiv with a unique advantage: it can reuse the perception, planning, and control capabilities accumulated in the field of autonomous driving while optimizing for robot scenarios, forming an innovative solution through cross-domain technology integration.

Zhang Hanwen cited Sequoia Capital's view that technological development is growing exponentially. Since the launch of ChatGPT in 2022, large model-based intelligent applications (Agents/APPs) have reached a revenue scale of $3 billion in just three years, far exceeding the growth speed of traditional SaaS industries.

He believes that, just as mobile phones and the internet gave rise to the mobile application ecosystem, the combination of robots and AI will birth a new generation of 'physical infrastructure + intelligent applications' ecosystem. With breakthroughs in domestic controllers and algorithm models, and the continuous integration of autonomous driving and robot application scenarios, a new ecosystem characterized by 'robots + Agents + APPs' is taking shape.