Preface: Autonomous driving only truly commences at the L3 level, albeit still a rudimentary conditionally autonomous system requiring driver vigilance. For automakers, achieving this level of autonomy holds immense significance. This year, numerous domestic automakers have announced plans to incorporate L3 autonomous driving architecture into their new vehicles or commercialize L3 technology within the year.

The second half of 2025 will witness a shift in the autonomous driving landscape, with the focus turning to conditionally autonomous driving—the L3 stage. This will comprehensively evaluate the technical prowess and safety capabilities of all participants' next-generation assisted driving models. L3 necessitates an upgrade in system safety from [Omnidirectional Active Safety] to [Comprehensive Safety], meaning the system must meet L3-level safety redundancy standards in both hardware and software, including but not limited to re-engagement within 10 seconds when the situation exceeds its capabilities.

Author | Fang Wensan Image Source | Network

Huawei Unveils ADS 4: Pioneering Domestic L3 Autonomous Driving Solution for Highways

On the eve of the 2025 Shanghai Auto Show, Huawei hosted the [Sharing Intelligence - Huawei Qiankun Intelligent Technology Conference] and officially launched innovative solutions such as Qiankun Intelligent Driving ADS 4, HarmonyOS Cabin HarmonySpace 5, and Qiankun Vehicle Control XMC.

Huawei's Qiankun Intelligent Driving ADS 4 will be available in four versions:

①ADS SE Basic Version: Aimed at the entry-level market, it offers fundamental active safety, parking assistance, highway NCA, and urban LCC+ functionalities.

ADS SE is anticipated to enter the market at around RMB 150,000, facilitating the widespread adoption of intelligent driving technology.

②ADS Pro Enhanced Version: Designed for mid-to-high-end models, it incorporates an in-cabin laser vision sensor, bolstering active safety and urban LCC+ functionalities.

ADS Pro targets the RMB 150,000 to 250,000 market segment, offering enhanced capabilities but not yet supporting highway autonomous driving.

③ADS Max Premium Version: Tailored for the high-end market, it features high-performance ultra-long-range LiDAR, optional distributed millimeter-wave radar, and high-precision solid-state LiDAR, supporting advanced features like full-dimensional active safety, highway NCA, urban NCA, and valet parking from space to space.

ADS Max is likely to target the market above RMB 200,000, delivering robust assistant driving functionalities in both urban and highway environments.

④ADS Ultra Flagship Version: Boasts full highway L3 capabilities.

ADS Ultra is expected to be exclusive to high-end luxury models, supporting highway L3 with the potential for future upgrades to urban L3.

Huawei revealed that its R&D investment in the automotive BU will surpass RMB 10 billion in 2024, with the R&D team expanding to 8,000 personnel and shipments of intelligent components exceeding 23 million units.

To date, 22 Huawei cooperative models have been launched, encompassing all power forms and vehicle types, laying a solid foundation for the large-scale commercialization of ADS 4.

From its 2019 entry into the field to the 2025 commercialization of highway L3, Huawei has annually invested billions in R&D funds, cultivating full-stack capabilities spanning [intelligent driving + cabin + chassis].

Driven by this trend, it is estimated that by 2025, the number of vehicles equipped with urban intelligent assistant driving systems will reach 3.94 million, rising to 8.25 million in 2026 and anticipated to surge to 14.96 million by 2027.

At the conference, senior executives from 11 automakers, including Thalys, Chery, BAIC, Avita, Deep Blue, Lantu, Mengshi, Audi, JAC, SAIC, and GAC, attended to express their support for Huawei's technology.

Adopting the World Engine + World Action Model Architecture

Qiankun Intelligent Driving ADS 4 employs the World Engine + World Action Model Architecture (WEWA Architecture), achieving a 50% reduction in end-to-end latency, a 20% increase in traffic efficiency, and a 30% decrease in hard braking rates.

Furthermore, Huawei has launched advanced modules such as high-precision solid-state LiDAR, in-cabin laser vision sensors, and distributed millimeter-wave radar.

The first architectural upgrade of Huawei Qiankun Intelligent Driving 4.0 introduces the World Model, adopts a novel technical architecture and cloud engine, and utilizes AI technology to generate high-quality, high-density scenarios for superior training of intelligent driving models.

In the realm of motion control, Huawei's XMC Digital Chassis Engine pioneers a six-in-one full-domain fusion architecture, enabling integrated control of systems including the body, motor, suspension, steering, and braking.

Based on integrated control, the vehicle can accurately predict road conditions and intelligently adjust the suspension, achieving unified control in the vehicle's X, Y, and Z directions.

The vehicle end is equipped with the World Action Model, which integrates full-modal perception (including cameras, LiDAR, millimeter-wave radar, etc.) and MoE multi-expert capabilities, achieving dual breakthroughs in trajectory generation and scene intention understanding.

This results in a 50% reduction in end-to-end latency, a 20% increase in traffic efficiency, and a 30% decrease in hard braking rates.

Huawei has mass-produced high-precision solid-state LiDAR with dimensions of 45×50×44mm for the first time, achieving a minimum accuracy of 3cm, capable of accurately identifying irregular obstacles such as water pipes, fire hydrants, and sunken steps.

Combined with in-cabin laser vision Limera and distributed millimeter-wave radar, it constructs the CAS 4.0 full-dimensional collision avoidance system, offering comprehensive coverage across all targets, speeds, directions, weather conditions, and scenarios.

In motion control, Huawei has introduced new technological breakthroughs, presenting the 6-in-1 full-domain fusion architecture for the first time, integrating the control of components like the body motor, suspension, steering, and braking into a central system, achieving the industry's highest degree of chassis engine integration.

Leveraging integration advantages, Huawei ADS4 provides a vehicle state prediction network capable of accurately forecasting ramp and road information.

The synergy of Huawei's WEWA system and Huawei's XMC Digital Chassis Engine enables functionalities such as tire blowout stability control assistance, anti-skid stability control assistance, spatio-temporal intelligent suspension adjustment, and omnidirectional comfort control.

The Qiankun Intelligent Driving World Engine (WE) located in the cloud leverages AI to generate a difficult case diffusion model, providing high-quality, high-density difficult case scenarios that are 1000 times denser than the real world, enabling AI training for AI.

The parking space-to-parking space experience has been further enhanced (P2P 2.0), supporting ETC passage at highway toll stations and valet parking functionalities for seamless transitions between parking spaces (VPD 2.0).

It is estimated that by June 2025, 500,000 parking lots nationwide will support parking space-to-parking space functionality.

The valet parking VPD function will introduce new capabilities such as cross-floor travel in underground parking lots and valet charging. By June 2025, it is anticipated that 100,000 parking lots nationwide will support the valet parking VPD function.

Conclusion: The market landscape presents a trend of dual dynamics—rise and fall.

The rise pertains to the expansion of the overall assistant driving market.

Data indicates that by the end of 2024, the market penetration rate of new energy vehicles surpassed 57%, and the market penetration rate of intelligent and connected vehicles also exceeded 20%.

The fall is evident in the decreasing cost of equipping assistant driving systems.

At the beginning of last year, intelligent assistant driving features were predominantly found in models priced above RMB 250,000, but this year, this feature has extended to models priced at RMB 150,000, imposing heightened demands on the market coverage capabilities of participants.

Another notable shift is the end of the era reliant solely on R&D.

The automotive industry epitomizes economies of scale, where costs can only be effectively shared through large-scale production.

The same principle applies to the assistant driving industry, where only large-scale mass production can propel the industry into a positive cycle of funds and data, thereby optimizing technical resources.

Some references: Dianchetong: "[True Intelligent Driving] is Here! Huawei Unveils ADS 4, Leaving Competitors Uneasy", Dongchehui: "Huawei Launches the First L3 Autonomous Driving Solution for Highways, Offering Enhanced Safety and Seamless Experience", Zhijiapai: "Huawei Aggressively Launches Highway L3 Intelligent Driving System ADS 4", Jiguo: "Huawei ADS 4 Makes a Big Splash! Highway L3 Autonomous Driving Arrives with Rainstorm Penetration and Self-Parking in Underground Level 5... 11 Automakers Rush to Adopt", Diancheshiyan: "Huawei Launches Qiankun Intelligent Driving ADS 4, Introducing Highway L3 Commercial Solutions", Dianzifayou: "Pioneering Commercial Solution Supporting Highway L3! Huawei Launches Qiankun Intelligent Driving ADS 4, Setting a New Benchmark for Intelligent Driving", Qichezhixin: "The Hand That Opens the Door to L3 Can Only Be Huawei Qiankun Intelligent Driving ADS 4", Guanwangcaijing: "Huawei Unveils Highway L3 Commercial Solutions"