In the context of rapid development of intelligent driving technology worldwide, L4 algorithm companies are gradually becoming the core force in the autonomous driving industry. Especially in the field of urban NOA (Navigated Open Autonomy), these companies, leveraging their leading edge in AI technology, data processing, and software architecture, are actively helping traditional automakers accelerate the mass production and implementation of autonomous driving technology.

The rise of L4 algorithm companies and their entry into new fields

1.1 From Robotaxi to L2+: Technological accumulation and transformation of L4 algorithm companies

Initially, L4 algorithm companies focused mainly on the Robotaxi sector, developing and operating autonomous taxi services. Through extensive road testing, they accumulated rich technical experience and vast amounts of driving data. This large-scale data collection and algorithm optimization have given L4 algorithm companies a significant lead over traditional automakers in the accuracy and stability of autonomous driving technology.

For example, L4 algorithm companies such as Momenta, Pony.ai, and DeepRoute.ai accumulated high-quality road test data in their early Robotaxi projects, covering complex urban road conditions and various extreme scenarios. With this rich experience, they gradually expanded into the L2+ (Advanced Driver Assistance System) field. L2+ technology is a transitional technology between traditional driving and full autonomous driving, aimed at enhancing the driving assistance capabilities of vehicles to achieve semi-autonomous driving in complex road conditions.

The core of this technological transformation lies in L4 algorithm companies leveraging their technological accumulation in the Robotaxi sector to adapt their advanced L4-level software systems to L2+ hardware by appropriately lowering hardware computing requirements. This approach not only enhances the performance of L2+ systems but also makes them more competitive in the market.

1.2 Compatibility advantages of data-driven and software architecture

L4 algorithm companies exhibit significant advantages in data-driven and software architecture compatibility, which are crucial factors in their rapid establishment in the L2+ market. Early investments in AI technology and data processing by L4 algorithm companies enabled them to establish robust data closed-loop systems. These systems collect and process vast amounts of driving data through extensive real-road testing, covering various driving scenarios and extreme road conditions.

Taking Momenta as an example, the company adopts a unified technical architecture and data processing solution across its L2+ and L4 businesses, enabling technology sharing and data interoperability between the two. This synergy enables Momenta to optimize its algorithms more effectively and improve the performance of its autonomous driving systems through continuous technological iterations.

Furthermore, L4 algorithm companies exhibit high compatibility in their software architecture design. Through modular design, these companies can flexibly apply L4-level autonomous driving technology to L2+ systems. For instance, LightSail Autonomous Vehicles continuously optimizes its software architecture to adapt L4-level technology to lower-configured L2+ hardware, significantly reducing costs and enhancing market competitiveness.

This data-driven and software architecture compatibility not only enhances the technological advantages of L4 algorithm companies but also enables them to provide more cost-effective and efficient solutions in collaboration with traditional automakers. This advantage is particularly evident in the mass production and implementation of urban NOA.

Practical cases of L4 algorithm companies assisting traditional automakers in urban NOA implementation

2.1 In-depth collaboration between Momenta and traditional automakers

Momenta, one of the earlier L4 algorithm companies to enter the L2+ business, has made significant progress in the urban NOA field through collaborations with several traditional automakers. As early as 2019, Momenta proposed the "L2+L4 dual-track" product strategy and gradually implemented it.

Momenta's "L2+L4 dual-track" product strategy

In collaboration with SAIC Motor, Momenta successfully implemented the mapless NOA function for IM Motor vehicles on May 25, 2024, in cities like Shenzhen, Guangzhou, Suzhou, and Shanghai. This function enables autonomous driving in complex urban environments without relying on high-precision maps. Leveraging its powerful data-driven algorithms and efficient software architecture, Momenta's function offers high accuracy, stability, and rapid adaptation to different urban road conditions.

Moreover, Momenta is collaborating with automakers like GAC Motor and BYD to gradually apply its end-to-end large models to mass-produced vehicles. This partnership accelerates the development and implementation of intelligent driving features for traditional automakers, significantly enhancing their competitiveness in the intelligent driving market.

2.2 Pony.ai's technological layout and market expansion

Pony.ai, another influential L4 algorithm company, boasts a strong technical background with its core technical team originating from renowned tech companies like Google and Baidu. Through extensive testing in North America and China, Pony.ai has accumulated rich autonomous driving data and gradually applied these technological achievements to its L2+ business.

In August 2023, Pony.ai successfully launched the Geestone 01 model equipped with its high-computing platform. This model realizes mass production and implementation of urban NOA through Pony.ai's L4-level algorithms. The platform's robust data processing capabilities and efficient algorithm optimization enable stable and safe autonomous driving in complex urban environments.

Pony.ai is not only actively expanding its domestic market but also exploring overseas opportunities. Through collaborations with renowned automakers worldwide, Pony.ai is gradually extending its technological advantages to broader markets. This global layout enhances Pony.ai's international competitiveness and provides broader development prospects.

2.3 Collaboration case between WeRide and Bosch

As a significant player in the L4 algorithm field, WeRide successfully applied its L4-level autonomous driving technology to the passenger car market through collaboration with Bosch. In March 2024, the high-level intelligent driving solution jointly developed by WeRide and Bosch was mass-produced in the EXEED Starlight ES model. This solution combines Bosch's hardware advantages with WeRide's algorithm technology, enabling stable NOA functionality in highway and urban driving conditions.

WeRide's technological layout extends beyond the passenger car market to autonomous shuttles and cargo vehicles. Through diversified business layouts, WeRide continuously expands its technological applications and collaborates deeply with traditional automakers to promote large-scale implementation of intelligent driving technology.

Innovation and challenges in the technical path of urban NOA

3.1 Evolutionary path of urban NOA technology

Urban NOA technology has undergone multiple stages of development, evolving from early reliance on high-precision maps for path planning to the current "high-precision map-free" approach. Autonomous driving technology is advancing towards greater autonomy and intelligence.

Traditional urban NOA systems relied on high-precision maps providing precise road information like lane lines and traffic signs to assist in accurate path planning. However, with technological advancements and the maturity of data-driven algorithms, more L4 algorithm companies are exploring the "high-precision map-free" approach.

Occupancy perception technology, represented by Tesla, is already replacing traditional high-precision maps. This technology divides the three-dimensional space into voxels and utilizes occupancy networks to perceive and predict object movements in the environment. Compared to traditional BEV (Bird's Eye View) perception solutions, Occupancy perception technology offers more flexibility in dynamic scene processing, effectively addressing complex urban road conditions.

The image shows a double-decker bus starting: Blue represents moving voxels, and red represents static voxels

This technological evolution significantly enhances the generalization ability of autonomous driving systems and reduces maintenance costs for high-precision maps. By introducing end-to-end deep learning large models, L4 algorithm companies can achieve more flexible path planning and environmental perception, enabling highly autonomous driving without high-precision maps.

3.2 Challenges in technological innovation and coping strategies

Despite significant progress in urban NOA technology development by L4 algorithm companies, numerous challenges persist. Firstly, technological complexity and reliability are concerns. Autonomous driving technology requires processing vast sensor data and making real-time decisions, posing stringent requirements on algorithm accuracy and response speed. Ensuring system stability and safety in complex urban environments is a pressing issue for L4 algorithm companies.

Secondly, cost control is essential. The hardware and software costs of autonomous driving systems are high, especially during initial development stages. To reduce overall costs, L4 algorithm companies must consider hardware selection, algorithm optimization, and data processing comprehensively. Continuously optimizing software architecture to be compatible with lower-cost hardware configurations is a common industry strategy.

Additionally, L4 algorithm companies must navigate regulatory and standard uncertainties. Autonomous driving technology progresses faster than regulatory frameworks, leading to varying standards and requirements across different countries and regions. Achieving global technological standardization and compliance is a challenge faced by L4 algorithm companies expanding their markets.

Collaboration modes and prospects between traditional automakers and L4 algorithm companies

4.1 Diversified exploration of collaboration modes

Collaboration modes between traditional automakers and L4 algorithm companies are diversifying, extending beyond technology introduction and application to encompass the entire chain from R&D to production, testing, and mass production. For example, Momenta's collaborations with SAIC Motor and BYD involve not only providing advanced autonomous driving algorithms but also participating in system integration and debugging. This deep collaboration transforms L4 algorithm companies from mere technology providers to key drivers of traditional automakers' intelligent driving transformation.

Pony.ai's collaborations with Geestone and GAC Motor apply its high-computing platform to mass-produced vehicles. This approach enables rapid technology validation in real-world driving environments and continuous optimization to enhance system performance and reliability.

WeRide's collaboration with Bosch further showcases the potential of cross-border cooperation in autonomous driving. As a leading automotive component supplier, Bosch boasts robust hardware R&D and manufacturing capabilities. Meanwhile, WeRide excels in autonomous driving algorithms. Their strong partnership led to the successful mass production of L4-level autonomous driving systems in the EXEED Starlight ES model, enhancing both parties' competitiveness in intelligent driving and providing valuable insights for other automaker-algorithm company collaborations.

4.2 Collaboration prospects and future outlook

With the continuous development of intelligent driving technology, collaboration prospects between traditional automakers and L4 algorithm companies will broaden. As autonomous driving technology matures and market demand increases, traditional automakers will increasingly rely on L4 algorithm companies' advantages in AI algorithms, data processing, and system integration. In turn, L4 algorithm companies will expand their technological applications through collaborations with automakers, facilitating large-scale production and commercialization.

This process will foster closer and deeper collaborations. Traditional automakers will transform from hardware manufacturers to intelligent driving solution providers. L4 algorithm companies, through automaker collaborations, will elevate their positions in the autonomous driving industry chain and drive overall technological advancements.

Future trends of urban NOA driven by L4 algorithm companies

5.1 Continuous technological innovation

In the coming years, L4 algorithm companies will continue to drive multiple innovations in urban NOA technology, including more efficient perception algorithms, smarter path planning technologies, and more stable system architectures. By incorporating advanced AI technologies, L4 algorithm companies will further enhance autonomous driving systems' intelligence, enabling autonomous driving in more complex environments.

Furthermore, with the further development of end-to-end deep learning models, autonomous driving systems will gradually detach from high-precision map reliance, achieving more flexible and autonomous driving decisions. Companies like Tesla and Huawei have made significant progress in this field, and more L4 algorithm companies will join the ranks, promoting technology popularization.

5.2 Industry chain collaboration and standardization

As L4 algorithm companies deepen collaborations with traditional automakers, the autonomous driving industry chain will exhibit stronger synergies. This synergy extends beyond technological development and product application to supply chain integration and optimization. Through establishing unified technical standards and data interfaces, L4 algorithm companies and traditional automakers will jointly promote intelligent driving technology standardization.

This standardization will enhance autonomous driving systems' compatibility and scalability, facilitating faster adaptation to diverse market demands. Additionally, standardization will reduce R&D and production costs across the entire industry chain, accelerating autonomous driving technology popularization.

5.3 Commercialization and market expansion

With technological maturity and cost reduction, L4 algorithm companies will accelerate urban NOA commercialization. Through close collaborations with traditional automakers, these companies will gradually achieve large-scale mass production of intelligent driving systems and enhance consumer acceptance through marketing and user education.

In the future, as autonomous driving technology becomes prevalent, L4 algorithm companies will explore more business models, such as Robotaxi and autonomous cargo vehicles. By continually expanding market applications, L4 algorithm companies will vigorously support further intelligent driving technology advancements.

Conclusion

Leveraging their leading edge in AI technology and data processing, L4 algorithm companies are actively promoting technological breakthroughs and implementations in the urban NOA field for traditional automakers. Through deep collaborations, these companies not only enhance their market competitiveness but also inject new vitality into the entire intelligent driving industry. In the future, with continuous technological innovation and further industry chain integration, L4 algorithm companies will occupy a more prominent position in the global intelligent driving market, driving the universal adoption of urban NOA technology.

References:

Western Securities: L4 algorithm company joins to help traditional auto plants implement city NOA.pdf [Intelligent Driving Frontier] WeChat official account background reply: C-0528, to obtain: L4 algorithm company joins to help traditional auto plants implement city NOA pdf download method.