Automakers kick off the L3 sprint race

Author | Chai Xuchen

Editor | Zhou Zhiyu

The era of intelligent driving powered by systems is approaching.

Recently, the Ministry of Industry and Information Technology announced the first batch of L3 conditional autonomous driving function vehicle access permits, which include the Deep Blue SL and the Arcfox Alpha S (L3 version).

On the same day that the access models were announced, Huawei's vehicle business unit stated that the Arcfox Alpha S uses the Qiankun Intelligent Driving ADS system, equipped with 34 high-performance sensors, including three LiDARs. This means that with Huawei's support, BAIC has successfully entered the L3 intelligent driving track.

Subsequently, XPeng and Li Auto also officially announced that they have obtained L3 road testing licenses, and will carry out regular L3 road tests in Guangzhou and Beijing respectively; Huawei's HarmonyOS and BYD have jointly launched L3 conditional autonomous driving internal testing with relevant departments in Shenzhen. Brands like XPeng and ZEEKR have also announced that their commercially available models are already hardware-ready for L3 autonomous driving.

Clearly, car manufacturers are gearing up for the high-level intelligent driving track. However, through the noise of publicity, what we see is not an instantaneous explosion of technology, but a more cautious, restrained, and lengthy marathon start.

According to the announcement from the Ministry of Industry and Information Technology, the Deep Blue SL03 and Arcfox Alpha S will conduct L3 testing in designated sections in Beijing and Chongqing, respectively, with certain restrictions.

Among them, the access scenario for the Deep Blue SL03 is currently limited to single lanes on highways and urban expressways in traffic congestion environments on designated road sections, while the access scenario for the Arcfox Alpha S is currently restricted to single lanes on designated highways and urban expressways. The testing routes mainly focus on relatively closed sections such as airport expressways and urban expressways without pedestrian mixing.

In addition to the usage scenario restrictions, the speed of the two L3 autonomous driving access models is also limited. The maximum speed for the Deep Blue SL03 in its usage scenario is 50 km/h; for the Arcfox Alpha S, it is 80 km/h.

Insiders have pointed out that due to safety considerations for the first access, the Changan Deep Blue SL03's intelligent driving lane change has not been included in this access matter and will be added to the subsequent access scope. It can be said that the current L3 still belongs to conditional autonomous driving.

According to the Q&A on the pilot work for the access and road use of intelligent connected vehicles released by the Ministry of Industry and Information Technology last June, the organization and implementation of the pilot is divided into five stages: pilot application, product access pilot, road use pilot, pilot suspension and exit, and evaluation adjustment.

At that time, a total of nine car manufacturers entered the joint pilot for the access and road use of intelligent connected vehicles, but only completed the selection of the pilot application stage, which does not mean that intelligent connected vehicles with autonomous driving functions have obtained access permits or are allowed to be on the road. The two L3 access models have now entered the road use pilot stage.

It is understood that the application needs to specify the pilot models, usage scenarios, speed ranges, and other restrictive conditions. If the Ministry of Industry and Information Technology decides to grant access, it needs to set restrictive measures such as access validity period and implementation area At the same time, with the upgrading and adjustment of policy requirements, some new changes will emerge, which requires car manufacturers to follow the latest standards or requirements.

In this regard, Sun Hang, Chief Engineer of the China Automotive Standardization Research Institute, stated that this release particularly emphasizes conditional access, allowing the activation of autonomous driving functions to be limited to typical, low-risk scenarios, aiming to start with small steps and safely and orderly promote the stable implementation of autonomous driving technology.

It seems that the testing access for L3 is merely a ticket; the test is just a simulation. The real era of autonomous driving is still beyond this license. Why are regulatory authorities and car manufacturers so cautious in advancing L3? The core issue lies not in technology, but in the attribution of responsibility.

In the era of L2 level assisted driving, the legal logic is very clear: the driver is the responsible party. No matter how intelligent the system is, the hands must hold the steering wheel, and the eyes must look at the road; if an accident occurs, it is the person's fault. However, L3 is a "watershed." Once the L3 function is activated, the control of the vehicle is legally transferred to the system.

Another factor of "conservatism" is the complexity of domestic roads. Currently, the officially announced L3 tests are mostly conducted on structured roads (overpasses, highways). This is because the traffic flow here is relatively orderly, with fewer uncontrollable factors. The truly complex urban ground roads for L3 still have a long way to go before implementation.

To meet the safety redundancy requirements of L3, car manufacturers have to "pile on" hardware. This includes perception redundancy, computing power redundancy, and reliance on high-precision maps while reducing map dependency. However, the high cost of hardware means that the first batch of L3 models will be concentrated in the high-end market. The inability to lower costs means that L3 is unlikely to become a mass consumer product in the short term.

Fu Bingfeng, Executive Vice President and Secretary-General of the China Association of Automobile Manufacturers, believes that intelligent connected vehicles, as complex systems across fields, rely not only on technological breakthroughs but also on commercial closed-loop verification. Multiple departments should jointly build a product access + road access mechanism, formulate rules in parallel, and conduct joint reviews to delineate a clear safety baseline for the industry and leave enough space for innovation.

It can be said that this pilot of L3 is merely the starting gun, with the first batch of participants rushing out of the starting line. Ahead of them are more complex ethical regulations, a vast amount of long-tail scenario data, and the inquiry into the ultimate proposition of "absolute safety."

Undoubtedly, high-level intelligent driving is rapidly moving towards the market. Qu Guochun stated that in the future, the technical evaluation process requirements will be improved through pilot projects, a full-chain safety management model will be constructed, and the safe application of intelligent driving technology will be promoted.

Looking back, the explosion of L3 will not occur now, but at the moment when road test data accumulates to a sufficient scale, legal responsibilities are clearly defined, and costs drop to a sufficiently affordable level. Before that, the industry needs a bit more patience and a greater respect for life