While the assisted driving industry continues to debate whether Level 3 capabilities are a necessity or a discretionary upgrade, regulatory approval has already begun to draw a line between theory and deployment.
China’s Ministry of Industry and Information Technology (MIIT) has announced the country’s first batch of approved Level 3 conditional automated driving vehicles. In parallel, the Harmony Intelligent Mobility Alliance (HIMA) has started on-road Level 3 system testing in Chongqing and Hefei, using production models including the Maextro S800 and Aito M9 on designated highway sections.
Test footage released by HIMA shows that once Level 3 mode is activated on approved highway segments, the vehicle interface shifts from an “available” state to a clearly marked “activated” and “in navigation” status. Under these conditions, the system assumes responsibility for driving tasks. The driver is no longer required to keep hands on the steering wheel, but must remain alert and ready to take over when prompted.
With regulatory clearance only beginning to take shape and much of the industry proceeding cautiously, HIMA’s decision to move early raises two questions. Why deploy Level 3 systems on public roads now, and what is the Huawei-led alliance attempting to validate through these tests?
Level 3 is no longer about whether the car can drive
For much of the past decade, discussion around Level 3 autonomy was driven by computing power, sensor counts, and ambitious timelines. That framing has shifted. Today, investment in Level 3 is less about being first to market and more about clarifying a long-term technological path.
As industry views diverge, some automakers and suppliers have chosen to slow or bypass Level 3 development, focusing instead on the eventual promise of Level 4 autonomy. That shift does not make Level 3 irrelevant. Instead, it underscores a central distinction: Level 3 cannot be treated as a simple extension of Level 2.
The key difference lies in responsibility. Level 2 systems operate under a “driver in command” model, where the system assists but accountability remains with the human. Level 3 requires the system to assume primary control and legal responsibility under defined conditions. That shift forces companies to build fundamentally different engineering architectures, including redundancy at multiple levels and validation across extensive real-world mileage.
Short demonstrations of hands-free driving in controlled environments are no longer the hurdle. Many companies can already achieve that. According to HIMA, more than 90% of its users engage assisted driving functions daily, suggesting strong user interest in more capable systems.
The challenge lies elsewhere. Stability, fault tolerance, and consistent performance across weather, traffic, and edge cases determine whether Level 3 can move from demonstration to everyday use. Without that foundation, technical capability alone does not translate into user confidence.
At this stage, the question is not whether a vehicle can drive itself under ideal conditions, but whether it can operate predictably enough to earn trust. That threshold can only be tested on real roads, where uncertainty is unavoidable.
HIMA’s response has been pragmatic. Rather than extending debate, it has placed Level 3 vehicles into controlled but real traffic environments, exposing systems to varied lighting, weather, and behavioral patterns. This includes testing human-machine interaction, system handover logic, liability boundaries, and long-tail scenarios that rarely appear in simulations.
The company argues that experience accumulated through real-world testing will shape its next generation of advanced assisted driving systems. The objective is not only to validate existing features, but to build organizational and technical capacity for long-term evolution.
What real-world testing can prove
HIMA’s Level 3 test footage has drawn attention because it makes a qualitative shift visible. Once activated, the system maintains speed in traffic, executes lane changes, and performs overtaking maneuvers autonomously. When vehicles ahead slow abruptly or cut in, the system adjusts proactively, while the driver’s hands remain off the wheel.
This capability is built on Huawei’s ADS 4.0 platform, which is designed to support Level 3 autonomy by emphasizing stability in perception, decision-making, and control.
At the system level, HIMA relies on its WEWA architecture, short for “world engine, world action.” The framework combines a cloud-based simulation engine with onboard behavioral models to construct a virtual world representation. This setup enables the system to train against a wide range of dynamic scenarios that extend beyond what is commonly encountered on public roads.
By simulating extreme or compound situations before they occur, the system reduces reliance on real-time improvisation. When unexpected obstacles appear or lead vehicles disappear suddenly, the architecture is designed to evaluate outcomes and respond within predefined safety envelopes. Continuous simulation and validation are used to reinforce decision consistency.
Real-world testing feeds directly back into this loop. Edge cases encountered during road tests are incorporated into subsequent training cycles, accelerating iteration. Over time, this feedback mechanism compounds system learning.
China’s expressways provide a demanding test environment. High-speed traffic includes heavy trucks, debris, and occasional animals, alongside significant seasonal variation such as fog, rain, snow, and ice. These conditions place sustained stress on perception and control systems.
To address this, HIMA integrates multiple sensor types, including LiDAR (light detection and ranging) for three-dimensional perception, millimeter-wave radar for velocity and object detection, and cross-redundant sensor fusion through its collision avoidance systems. Coverage spans frontal, rear, and lateral scenarios across a wide speed range, with full 360-degree awareness.
According to a Gasgoo report on LiDAR installations during the first eight months of 2025, China’s market has become increasingly concentrated. Huawei, which supplies LiDAR systems used by HIMA, ranked first with 643,800 units installed, accounting for 41.1% of the market. That scale supports system integration and data accumulation, advantages that are difficult to replicate quickly.
Some safety mechanisms developed for Level 3 have already been applied to existing products. For example, driver incapacitation functions originally designed for Level 3 scenarios are now available in production vehicles via ADS 4.0 updates. If a driver becomes unresponsive, the vehicle can slow, move to the emergency lane, and activate hazard lights.
In this way, Level 3 development has begun to influence broader assisted driving safety standards, even before large-scale commercialization.
Where assisted driving is heading
Policy signals suggest expansion will be gradual rather than rapid. MIIT’s latest roadmap for smart and connected vehicles and industry research from Deloitte both forecast growth in China’s Level 3 market, from an estimated RMB 150 billion (USD 21 billion) in 2025 to RMB 800 billion (USD 112 billion) by 2030, implying a compound annual growth rate of 38.7%.
Adoption, however, will depend less on headline capability than on sustained reliability. Trust between users and automated systems is built incrementally, through consistent performance rather than isolated milestones.
HIMA’s approach reflects that assumption. By prioritizing system engineering, data depth, and ecosystem integration, it is treating Level 3 less as a feature upgrade and more as a structural transition.
By late 2025, competition in assisted driving is increasingly defined by execution rather than demonstrations. Level 3 marks the beginning of a phase in which differentiation hinges on reliability, accountability, and integration.
HIMA’s early move into regulated Level 3 testing does not settle the industry’s debates, but it does clarify one thing. As Level 2 systems reach diminishing returns, the path forward depends on whether companies can deliver not just autonomy, but confidence.
The central question remains unresolved but sharply framed: what degree of reliability is required before drivers are willing to relinquish control at highway speeds? HIMA’s answer is to accumulate evidence, one kilometer at a time.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This article was written by 36Kr Brand.