Dialogue with Voyah CTO Wang Junjun: Achieving L3 assisted driving requires sufficient safety redundancy

Author | Zhou Zhiyu

Editor | Zhang Xiaoling

Multiple intelligent driving accidents this year have led to strong regulation of the rapidly advancing "intelligent driving equality."

On April 16, the Ministry of Industry and Information Technology held a meeting to promote the management of intelligent connected vehicle product access and software online upgrades, prohibiting exaggerated promotion of intelligent driving, strengthening regulation of enterprises, and standardizing the disorderly development of the intelligent driving industry.

Under strong regulation, car manufacturers have shifted from showcasing technology to focusing on the essence of intelligent driving solutions—safety.

The latest competitive point is L3-level autonomous driving. On April 16, Voyah released the world's first L3-level intelligent architecture "Tianyuan Smart Framework," which aims to elevate the safety of intelligent driving from "function realization" to "systematic guarantee."

Voyah's CTO Wang Junjun repeatedly emphasized one term at the exchange meeting: safety redundancy. He told Wall Street News that the biggest change in L3-level advanced intelligent driving compared to L2 is that the driving responsibility shifts from the user to the OEM. From a technical perspective, true L3 represents a hundredfold or even thousandfold improvement in safety.

Wang Junjun stated that under this change, safety redundancy is the most basic bottom line. Because if it fails once, it means misfortune for an individual or a family.

In terms of specific solutions, it can be seen that Voyah has implemented multi-dimensional redundancy in the Tianyuan Smart Framework, covering different dimensions such as battery packs, wiring harnesses, and domain controllers.

The Tianyuan Smart Framework is based on Voyah's L3-level intelligent safety driving platform and Kunpeng's L3-level intelligent safety driving system, two core technology clusters.

The "Qingyun Platform" of Voyah's L3 architecture centers around a digital driving control platform, integrating advanced technologies such as axial flux motors, distributed electric drive, steer-by-wire, brake-by-wire, rear-wheel steering, and fully active suspension, to construct comprehensive intelligent capabilities for the chassis system.

Compared to traditional radial flux motors, the axial flux motor in Voyah's solution has a 30% increase in torque density, a 40% reduction in size, and a peak torque of 680N·m per motor.

Wang Junjun explained that as L3 intelligent driving gradually expands, axial flux motors will definitely be widely promoted. Axial flux motors have advantages such as a wide torque range and long high torque retention time, further improving efficiency and performance. When the vehicle is matched with a fully active suspension, it can meet spatial requirements.

The innovation of Voyah's fully active suspension lies in incorporating safety protection into the core design logic. The dual-camera stereo pre-scan system can identify obstacles within ±20mm height at a distance of 40 meters, combined with a system lifting force of 41,000N and a suspension travel of 200mm, allowing for proactive adjustment of the vehicle's posture before a collision.

For example, when detecting the risk of a rear-end collision with a large truck ahead, the suspension will quickly lift the rear of the vehicle to avoid impact on the passenger compartment. This "active defense" mindset upgrades traditional passive safety to a dynamic prediction system.

The full-line control of technologies such as steer-by-wire, brake-by-wire, and rear-wheel steering provides millisecond-level response capabilities for L3-level intelligent driving. For instance, the response time of brake-by-wire has been reduced from the traditional hydraulic system's 150ms to under 100ms, decreasing the braking distance at 100 km/h by 1.8 meters. The dual-redundancy design (dual sensors, dual ECUs, dual power supplies) ensures that 50% of functionality can be maintained even in the event of a single point failure A 700-line high-level lidar paired with 3 laser radars and 1000 TOPS dual-chip computing power redundancy, along with dual stereo pre-aiming cameras, extends the small object recognition distance from 50 meters at L2 to 100 meters; the AEB braking speed limit has also increased from the L2 level of 130 km/h to a maximum of 160 km/h.

Currently, the global intelligent driving industry is at a critical stage of transitioning from L2 to L3. According to data from the Ministry of Industry and Information Technology, the penetration rate of new L2 vehicles in China reached 57.3% last year. Several automakers and institutions expect that the penetration rate of L2 combined driving assistance systems will reach 90% by 2030, and the next few years will be a key period for the gradual advancement of L3 and above systems.

According to Voyah's plan, it will release its first L3-level Tianyuan intelligent architecture model within this year, codenamed "Taishan," which is Voyah's first full-size large six-seat SUV.

Wang Junjun believes that currently, L3 is still achieved in specific scenarios, such as highways. Voyah hopes to broaden the scenarios and working conditions for L3; otherwise, the practical value of L3 is not high, and the next step will be a gradual advancement process.

He revealed that Dongfeng Motor Corporation is involved in the formulation of L3 standards. Currently, L3 will involve adjustments to the "Road Traffic Safety Law," so no formal regulations have been issued yet.

Wang Junjun pointed out that all technological developments have a gradual process. Based on this process, enterprises and national regulations are continuously communicating, including the Road Traffic Safety Law and regulations related to L3.

In this context, automakers still need to focus on forward-looking development in technology, gradually maturing the technology, performance, and rationalizing costs.

As intelligent driving shifts from "functional competition" to "responsibility," safety is no longer just a technical parameter but requires real investment to genuinely uphold user trust with technology. This is an all-encompassing battle of technological routes, industrial chain reconstruction, and business model innovation, which will gradually create the core competitiveness of Chinese automakers in the era of intelligent driving.

The following is a dialogue between Wall Street Journal and Wang Junjun, CTO of Voyah, and Jiang Tao, COO of Voyah Technology Co., Ltd. (edited):

Q: After recent industry intelligent driving accidents, the boundaries of L2+ usage scenarios, OTA, and corporate responsibility have been particularly strict. Is Voyah's push for the L3 architecture related to this?

Wang Junjun: When developing the "Tianyuan Intelligent Architecture," Voyah was not overly influenced by external events; Voyah has always adhered to its own research and development pace. The entire L3-based intelligent driving development began a year ago. Voyah has consistently pursued forward-looking development based on its own rhythm, aiming to provide the latest and safest technology and product experiences to all customers.

All technological developments have a gradual process. Based on this process, enterprises and national regulations are continuously communicating, including the Road Traffic Safety Law and regulations related to L3. These timing issues will gradually be implemented in the future.

Q: You just mentioned safety redundancy. Where is Voyah's safety redundancy mainly reflected? Wang Junjun: Voyah has many safety redundancies, for example, the L3 intelligent driving system has four LiDARs. Even if one or two fail, it can still operate normally. Additionally, the domain control uses a dual SOC architecture, which means that even if one fails, it can still perform related functions to provide customers with the best safety experience.

Question: How can intelligent assisted driving safety be achieved through updated technology and the integration of new software and hardware? What advantages does Tianyuan Zhijia have?

Wang Junjun: In fact, the biggest change with L3 is that the responsibility for driving safety, which was previously on the vehicle owner, will shift to the manufacturer or other parties. From a broader technical perspective, true L3 represents a hundredfold or even thousandfold improvement in safety. However, when a person hands over safety to someone else, it is a significant responsibility to protect that person's safety. We are not simply transitioning from L2 to L3 by adding a few simple sensors.

A hundredfold performance improvement and safety redundancy are the most basic bottom line. Because if there is a single failure, it means misfortune for an individual or a family. You need to have corresponding redundancy plans; we have many sensor redundancies. On Tianyuan Zhijia, we have implemented redundancy for batteries, wiring harnesses, and domain control. These redundancies are designed to enhance safety, even at a relatively high cost. For L2 driving safety, even in emergency situations, we can truly ensure that collisions do not occur, guaranteeing the maximum safety outcome.

Question: Due to some industry accidents, there are significant concerns about AEB. How should consumers correctly understand AEB? Is it L2 or L3?

Wang Junjun: Today, we are discussing how to enhance the safety of AEB based on L2, making AEB very effective, including forward, backward, and lateral protection, covering all scenarios we can think of.

In the current context, we emphasize that it is not about equal rights for intelligent driving, but rather equal rights for safety. Currently, all driving is assisted, requiring the driver's full attention. Just because AEB is available for forward, backward, and lateral protection, it does not mean the driver can sleep while driving. A person still needs to be responsible as part of the process. Therefore, we do not advocate for infinitely expanding the safety boundaries of L3.

For manufacturers, we must achieve the maximum safety redundancy to avoid adverse safety issues for users.

At this point in time, as far as I know, full-domain L3 is still not achievable. What people refer to as L3 is in specific scenarios, such as being able to perform L3 assisted driving on highways. In any case, highways are much simpler than urban roads.

Jiang Tao: Whether L2 or L3, they are both forms of assisted intelligent driving. Intelligent driving requires the driver's hands to remain on the steering wheel. This is the customer's right to know; automakers must inform customers what they can achieve. For L3, there is a requirement for the driver to take over the vehicle within 10 seconds, meaning the driver must always be able to take control within that time frame; it is certainly not acceptable for the driver to completely ignore the vehicle.

Question: Besides the basic architectural redundancy, L3 also involves standards, regulations, and the allocation of responsibilities. What is the current progress of industry standards, and is Voyah participating in the formulation of these standards? Wang Junjun: First, the entire Dongfeng Company is participating in the formulation of the L3 standard.

Second, since L3 will involve adjustments to the Road Traffic Law, there are no formal regulations yet.

Third, the country has some ICV (Intelligent Connected Vehicles) pilot projects that require operational partners and manufacturers to apply together. After the application, pilot projects will be conducted in certain areas, and Dongfeng Voyah is now also involved. We have been tracking the progress of the regulations from national ministries, which require changes to the entire supporting regulatory system. At this point, Dongfeng and Voyah have been involved in the entire regulatory formulation, and we are also applying for ICV (Intelligent Connected Vehicles).

Question: How does Voyah's axial flux motor ensure strong performance while reducing manufacturing costs and maintaining competitiveness in the market?

Wang Junjun: Indeed, axial flux motors are very expensive. Let me share an example: our first-generation Voyah FREE was one of the earliest in the country to adopt air suspension. Initially, air suspension was only found in luxury cars priced over a million. When Voyah FREE was first proposed, we worked with the startup Konghui to overcome numerous challenges, not just technical challenges but also process and cost challenges. In the end, we managed to implement air suspension on a 300,000 yuan new energy SUV and established the ecosystem. Voyah and Konghui collaborated to promote the popularization of air suspension in the industry, and as everyone knows, the cost of air suspension has become very cheap now.

I give this example because we believe that axial flux motors will follow a similar path.

Once L3 intelligent driving is implemented, we believe that axial flux motors will definitely be widely promoted. Axial flux motors have several advantages: they can reduce weight by 50%, decrease volume by nearly 40%, and they have a very broad torque range, with high torque maintained for a long time. Their efficiency is very high, significantly enhancing performance.

These advantages of axial flux motors make them a great solution for matching fully active suspension in future vehicles, especially when space requirements are high. Currently, the main issues are related to production processes and raw materials for components. If they achieve large-scale adoption, costs will significantly decrease.

Question: What core technical challenges did Voyah overcome during the development phase of fully active suspension?

Wang Junjun: The development of fully active suspension involves challenges in technology, processes, and costs. For example, we encountered issues with pumps early on, as fully active suspension mainly relies on the response of pumps and the consistency of main power output. Currently, the technology for pumps has matured. After rationalizing costs in the future, we aim to make fully active suspension a standard feature, providing users with a great experience, which is what we are currently working on.

One obvious advantage of fully active suspension is its stronger perception capability. Voyah previously used monocular cameras like our competitors but found that binocular cameras can see further and in more detail. For example, with a precision of 15 centimeters at a distance of 100 meters, it can provide better responses and safety responses.

From the perspective of chassis control, the industry has always aimed to achieve coordinated control of lateral, longitudinal, and vertical movements. Based on fully active suspension, we have integrated VMC control for lateral, longitudinal, and vertical movements, collecting various information and executing it in coordination to provide users with the best comfort experience In terms of scenarios, Voyah hopes to combine fully active suspension and safety considerations, seeking the best posture for the vehicle to face collisions in various situations, thereby ensuring the safety of passengers and external personnel, and minimizing various risks as much as possible. The above three points are the innovations and forward-looking explorations made by Voyah.

Question: Quality inspection is a very important link in the production process. What new technologies does Voyah have to improve production efficiency?

Jiang Tao: Cars are becoming increasingly complex with more functions. According to traditional quality control methods, it would require running a certain number of kilometers and testing all functions, which now seems impractical.

This requires a change in our quality control methods. Additionally, functions need to be upgraded from software development; each functional unit must undergo automatic testing. It should automatically generate testing codes, conduct automatic testing, report automatically after testing, and automatically inform us how to correct these functions.

The same applies after the vehicle comes off the production line; we can issue commands from the cloud to allow the vehicle to conduct automatic testing. For example, we can simulate vehicle usage, such as turning on the headlights 50 times, moving the seats back and forth a certain number of times, etc., to specifically test the intelligent driving functions. This method improves the current detection rate.

After our vehicles come off the production line, they will automatically test for 60 to 90 minutes. This simulates customer usage in the first month, such as how many times the doors are opened, how many times the windows are raised and lowered, how many times the wipers are used, etc. By simulating such scenarios, we can reduce the failure rate during the initial usage period.

Thirdly, when real problems arise in the market, we can control them through the cloud. Since process data can be uploaded to the cloud, we can anticipate failures before they occur and communicate with customers. Through remote diagnostics and software updates, we can achieve automatic fault clearance. When software issues arise, there is no need for in-store repairs, which is also our future direction. Through these methods, from the development process to manufacturing testing, and then to after-sales automatic diagnostics, this combined approach can enhance our quality