Xpeng's Vision-Based Autonomous Driving Challenges Tesla's Approach






Xpeng, a Chinese electric vehicle manufacturer, is increasingly embracing a vision-centric strategy for autonomous driving, mirroring Tesla's controversial method. This marks a significant pivot for Xpeng, which was the first automaker globally to incorporate lidar into its vehicles back in 2020. However, the company has since reassessed its approach, now favoring a camera and AI-driven system.
Candice Yuan, a leading figure in Xpeng's Autonomous Driving Center, revealed at the IAA Mobility 2025 show that lidar data proved incompatible with their AI architecture. She explained that Xpeng's AI system, named VLA (Vision, Language, Action), is predominantly trained using short video segments from customer vehicles. Unlike lidar-based systems employed by companies such as Waymo and Zoox, which emphasize lidar's role in enhancing environmental perception, especially under challenging conditions, Xpeng believes its vision-only model, XNGP, can theoretically operate across various scenarios in China. This perspective aligns with Tesla's long-standing argument that camera-based systems are more cost-effective and scalable than lidar, despite ongoing debates about the reliability of camera-only autonomy in highly complex urban settings.
While Xpeng and other Chinese manufacturers like Ji Yue are advancing vision-based autonomy, the ultimate goal of true driverless cars remains elusive. Despite promises from companies like Tesla about impending fully autonomous capabilities, these systems often still necessitate human oversight, particularly in intricate urban environments. In contrast, Waymo and Zoox, with their heavy reliance on lidar, currently offer truly driverless rides, suggesting that a comprehensive sensor suite may be critical for achieving full autonomy. This ongoing technological divergence underscores the dynamic and competitive landscape of autonomous driving development, where different manufacturers are betting on distinct pathways to achieve the future of self-driving transportation.