Xpeng's Autonomous Driving Vision: L4 Simplicity Over L2+ Complexity

Covers infotainment, ADAS and software updates. Daily: Hyundai Ioniq 5. Project car: NB Miata with a homebrew CarPlay retrofit.
In a landscape where the future of driving is being reshaped by technology, Xpeng's director of autonomous driving, Candice Yuan, offers a unique perspective. At the 2025 IAA Mobility event in Munich, Yuan shared insights into why Level 4 (L4) autonomous driving is simpler than the existing Level 2+ systems with human drivers, shedding light on the evolving dynamics of autonomous vehicle technology.
The IAA Mobility event in Munich served as a backdrop for a revealing conversation with Candice Yuan, who oversees Xpeng's autonomous driving technology. Yuan's experience, which spans over a decade in the industry, positions her at the forefront of a transformative era in automotive technology. Prior to joining Xpeng, Yuan worked at Alibaba on unmanned logistic vehicles, a role that emphasized efficiency and safety without the complexities of human passenger comfort. This background now informs her work at Xpeng, where the challenge lies in balancing technological efficiency with the subjective preferences of human drivers. Yuan explained that while L4 technology is developed primarily for unmanned vehicles, L2+ systems must cater to human expectations and comfort, which adds layers of complexity.
Xpeng has positioned itself uniquely in the competitive Chinese electric vehicle (EV) market by emphasizing its advanced driver-assistance systems (ADAS) and autonomous driving capabilities. Unlike its competitors such as Nio, BYD, and Li Auto, which focus on different aspects like premium features or extended range, Xpeng is pioneering smart driving technology. A significant milestone for Xpeng was its collaboration with Volkswagen in 2023, where Volkswagen invested over $700 million in Xpeng to accelerate its EV development cycle. This partnership underscores the global recognition of Xpeng's technological advancements. Yuan emphasized that Xpeng's integration of AI and proprietary AI chips allows for seamless hardware-software synergy, enabling rapid advancements in autonomous technology.
In a candid discussion, Yuan detailed Xpeng's decision to phase out lidar from its vehicles, a move that reflects confidence in its AI-driven vision systems. Xpeng's AI system, based on a large language model, processes vast amounts of road data through short video clips, allowing the system to learn and evolve rapidly. This approach contrasts with lidar's limitations, which cannot be fully integrated into the AI learning model. Yuan explained that by focusing on visual inputs—termed as Vision, Language, and Action (VLA)—Xpeng's system can learn from both positive and negative driving scenarios, enhancing its decision-making processes on the road. This strategic move towards a vision-based system is expected to accelerate the development of Xpeng's autonomous capabilities.
Looking ahead, Xpeng aims to expand its autonomous driving capabilities globally, with plans to roll out its XNGP system, a Tesla FSD alternative, by 2025. The system is designed to function independently of high-definition maps, making it adaptable to diverse environments. However, the expansion into international markets is contingent on regulatory approvals, particularly in regions like Europe where policies for autonomous driving are still evolving. Xpeng's recent approval for remote autonomous parking in Hong Kong marks a significant step toward its global ambitions. Yuan expressed optimism that as regulatory landscapes mature, XNGP can be swiftly adapted and implemented worldwide, showcasing Xpeng's readiness to lead in the autonomous driving space.
The conversation with Yuan also touched on the future trajectory of autonomous driving technology. She noted that achieving Level 4 autonomy is a crucial step toward full autonomy, with Level 5 being a natural progression. Yuan stressed that the transition from L2+ to L4 is the most critical and challenging phase, given the need for sophisticated software capabilities and supportive regulatory frameworks. Xpeng's commitment to advancing this technology is evident in its plans to mass-produce L4 robotaxi vehicles by next year, a testament to its leadership in the field. Yuan's insights highlight the intricate balance between technological innovation and regulatory compliance that defines the future of autonomous vehicles.
As Xpeng continues to push the boundaries of what's possible in autonomous driving, its focus on integrating AI and developing in-house technologies positions it as a formidable player in the global EV market. Yuan's vision for Xpeng underscores the potential of AI-driven autonomous systems to not only enhance driving efficiency but also redefine mobility. With a team of over 1,000 employees dedicated to autonomous driving, Xpeng is poised to lead the charge in this transformative sector. The company's ongoing innovations and strategic partnerships highlight its commitment to shaping the future of transportation, offering a glimpse into a world where human oversight is gradually reduced in favor of smarter, more efficient vehicle control systems.

About Noah Stein
Covers infotainment, ADAS and software updates. Daily: Hyundai Ioniq 5. Project car: NB Miata with a homebrew CarPlay retrofit.