Idealism Boosts Autonomous Driving Leadership
Advertisements
On January 16, Li Auto officially rolled out the OTA 7.0 version of its vehicle operating system, marking a significant milestone for the companyWhat sets this version apart is the display of the internal version number of AD Max, the intelligent driving system, which is now at V13.0. This upgrade represents a commitment to continuously iterate and maintain the AD Max version number in accordance with product updates and optimization efforts.
One of the most exciting features included in this version is an industry-first AI inference visualization capabilityThis enhancement not only improves the user interface experience but also integrates a full-scene end-to-end functionality for intelligent driving, making Li Auto a trailblazer in the Chinese automotive sector.
As a result, Li Auto has become the first company in China, and the second globally, to apply end-to-end technology across all driving scenarios, including urban, highway, and loop NOA (Navigation on Autopilot). This application signifies a true advancement in how automotive companies employ artificial intelligence to enhance automated driving.
The journey does not stop here; Li Auto has made a pivotal move toward becoming a global leader in artificial intelligenceOver the past decade, the company has established a systematic capability in the field of AI, positioning itself as a core player in AIEV (Artificial Intelligence Electric Vehicles). The integration of AI technology with AIEV is proving to be a competitive battleground that is elevating the automotive industry's electronic and software architectures while significantly improving driver-vehicle interactions.
This revolutionary product not only redefines the way people travel but also elevates intelligent driving to unprecedented heightsAs artificial intelligence matures, the competition in AIEV is predicted to become fierce by 2025, making intelligent driving the strategic battleground for automotive manufacturers
Advertisements
Advanced AI technologies, such as end-to-end systems and large models, are emerging as the mainstream technological focus for automakers, guiding the industry toward higher levels of innovation.
The smart driving market, valued at hundreds of billions, is becoming a battleground for various playersGiven the intense competition between AI and automotive integrations, the smart driving market is thriving faster than everA forecast by iiMedia Research predicts that by 2030, the penetration rate of NOA functionality could reach 80%, leading to a market scale exceeding 300 billion yuan.
In this silent conflict, metrics like the number of users and driving mileage have become critical indicators of a company’s strength in the marketAs of December 31, 2024, Li Auto had approximately 1.109 million users of its intelligent driving system, accumulating a remarkable 2.93 billion kilometers in smart driving mileageThis extensive user base and the wealth of driving data support continuous optimization of Li Auto's technology and product iterations.
Huawei, a competitor heavily invested in AIEV, reported that its QianKun intelligent driving system had reached over 500,000 users by 2024, with a total driving distance of 1.45 billion kilometers, benefitting from a multi-brand ecosystem that includes models like Wenjie and AvitaHowever, when compared to the single-brand focus of Li Auto, Huawei still falls short by a significant margin.
Xpeng Motors, another deep player in the AIEV space, has positioned smart driving as its unique selling pointOn January 14, Xpeng released its 2024 intelligent driving report, reporting a stunning 241.28% year-on-year growth in total driving distance, with 97.3% of its owners employing the intelligent driving functionalitiesHowever, the company did not disclose specific user numbers or mileage, which raises questions about its competitiveness considering it sold fewer than 200,000 units annually.
It is clear that Li Auto, having swiftly achieved an industry record for luxury car sales in the Chinese market, finds itself leading in a close combat with its industry rivals.
Moreover, Li Auto has exhibited a more aggressive stance in the realm of intelligent driving
Advertisements
For instance, the company has introduced an end-to-end + VLM (Visual Language Model) dual-system framework that represents a true One Model structureThis innovative approach integrates the VLM into the vehicle's onboard chip framework, successfully rolled out in the OTA 6.4 version, thereby positioning Li Auto ahead in the smart driving technology space amidst fierce competition.
It's worth mentioning that in 2024, Li Auto made remarkable strides in academic achievements in intelligent driving, publishing papers in top-tier international conferences such as ECCV, CoRL, and AAAINotable works included DiVE (world model) and StreetGaussians detailed at ECCV 2024 and DriveVLM (dual-system) showcased at CoRL 2024. Further, Li Auto's dual-system model received accolades for technological excellence from recognized media outlets.
In summary, 2024 marks a milestone year for Li Auto, establishing itself as a leader in intelligent driving.
Within the top tier of smart driving manufacturers, Li Auto’s rapid ascent can be attributed to a series of technological innovations and an unwavering dedication to user experienceThe evolution from NPN (prior information) architecture to the no-map solution, and ultimately to the integrated end-to-end + VLM dual system, has demonstrated three generations of upgrades in smart driving technologyEach technical leap represents a significant breakthrough for Li Auto in the field of intelligent driving.
The first-generation intelligent driving system utilized the NPN architecture, which focused heavily on rule-based algorithmsBy leveraging a network of sensors to gather data and processed it through a small perceptual model, the system could make informed decisions based on navigation data.
While this architecture provided the Li Auto vehicles with comprehensive NOA capabilities that bridged highway and urban contexts, it suffered from certain latency issuesThe second-generation technology adeptly mitigated these shortcomings.
Li Auto then pivoted away from the NPN framework in its second-generation technology, embracing a more sophisticated no-map solution
Advertisements
This model comprises perceptual and planning components, allowing for integrated perception, localization, and decision-makingIt transitioned planning from a rule-based to a model-based approachThis enhancement granted Li Auto vehicles national driving capabilities, significantly broadening the application of urban NOA across the country.
As a result, Li Auto has carved a path into the upper echelon of intelligent driving technology.
The third-generation intelligent driving technology represents a qualitative leap, utilizing a true One Model architecture—an integrated end-to-end modelThe entire process from sensor input to trajectory output is handled by a single large model, eliminating the need for intermediary rule-based algorithms.
Moreover, the third-generation technology incorporates VLM cognitive capabilities, establishing the dual-system architecture of end-to-end + VLMThis innovation not only enhances the efficacy of Li Auto's intelligent driving solutions but also lays down an industry benchmark, effectively catalyzing the advancement of the overall sector.
Li Auto has also constructed a unique “smart driving validation system,” which rigorously assesses and grades the end-to-end model, ensuring only qualified models are considered for iteration guidanceThis significant initiative accelerates the iterative capabilities of Li Auto’s AI models, providing users with increasingly intelligent and reliable driving experiences.
To ensure continuous technological innovation, Li Auto supports ample training mileage and computing powerIn 2024, the company added an impressive 1.72 billion kilometers of smart driving mileage, while its current training computing capacity reached 8.1 EFLOPSAnnual investments in training computing power have exceeded 1 billion yuan, reflecting Li Auto's commitment to its intelligent driving technology pursuits and maintaining its industry leadership.
Even as intelligent driving technology rapidly progresses, safety remains a cornerstone principle for Li Auto
While primarily employing a pure vision technology route, Li Auto retains LiDAR as a safety mechanism, ensuring driving safety through a complementary approach to vision technologyWith continuous OTA updates, Li Auto has effectively prevented over 4.239 million potential accidents and 1,230 severe incidents, offering users a heightened sense of security during travel.
In addition to technological advancements and safety measures, Li Auto has introduced innovative user-centric features such as "freestyle parking," "upgraded closing doors while parked," and "parking from space to vehicle." These enhancements enrich the user driving experience while amplifying the user's reliance on and affinity for Li Auto’s intelligent drivingConsequently, the penetration of these features within the user base further cements their position as a core competitive strength for the company.
For Li Auto, intelligent driving has emerged as a critical factor for potential customers when considering vehicle purchasesAccording to Li Auto, as of December 31, over 72% of users engaged in NOA test drives, with more than 75% of AD Max models costing over 300,000 yuan delivered, and over 84% of those above 400,000 yuan delivered.
Through a firm commitment to innovation, significant investments, safety adherence, and user-centric measures, Li Auto has successfully entrenched itself among the top tier of intelligent driving manufacturers.
Li Auto is set to navigate the seams of AIEV's futureScheduled from December 25-27, 2024, the company will launch its Li Auto AI Talk, a daily 30-minute session that delves into the underlying logic and developmental pathway of its intelligent driving technology.
Li Auto asserts that achieving "parking space to vehicle" capabilities is central to realizing Level 3 automated drivingBy late November 2024, the company had globally pioneered this capability, resolving the first and last 100 meters of travel challenges, ultimately enhancing user experience
Drawing from profound foresight and continuous investment, Li Auto has been poised with an equipped status of “algorithms, data, and computing power,” successfully launching its end-to-end + VLM intelligent driving system.
In terms of its smart driving roadmap, Li Auto aims to achieve supervised Level 3 autonomous driving by 2025 and intends to launch its first Level 4 defined automated vehicle by 2028. Attaining these ambitious goals hinges on three conditions: a fleet of over 5 million vehicles, mastery of the VLA foundational models, and sufficient capital to attract top-tier talent.
In an age where AI and automotive sectors are deeply intertwined, Li Auto has ascended to the peak of intelligent drivingHowever, the company understands that intelligent driving is merely the tip of the iceberg in the evolution of AIEV.
As AI technology continues to mature and become ubiquitous, AIEV will continue to showcase its unique allure and boundless potential across various fieldsLi Auto is making steady strides to adapt to future competition and the evolving needs of consumers, maintaining a sense of reverence for technology and gratitude toward its users as it drives innovative advancements in intelligent driving, paving the way for a new chapter in smart mobility.
Notably, during the AI Talk, Li Auto articulated its AGI strategic roadmap, segmenting its AI development into three phases: first, the "enhanced my capabilities" phase that aims for Level 3 supervised autonomous driving where AI assists in elevating human work efficacy under human oversight; second, the "become my assistant" phase transitioning to Level 4 autonomous driving and agent development with capabilities such as Level 4 vehicles autonomously fetching children, synthesizing cognitive and spatial intelligence into VLA models—marking AI's iPhone 4 moment; and third, the "become my silicon-based family" phase where AI independently executes tasks without human directives, achieving prominence as vital family members, emblematic of AGI's ultimate product.
With its AGI strategic map laid out, Li Auto is striving toward higher levels of automobile intelligence, underscored by robust R&D investments aimed at ensuring core technology iterations driven by AI, fostering AI accessibility for every household.
Advertisements
Advertisements
Leave A Comment