Qian Zhan Wang·2025-09-29 08:37Core Viewpoint - Xiaopeng Motors has decided to abandon LiDAR technology in favor of vision-based systems for autonomous driving, believing this shift will enhance system development and reliability [2][4]. Group 1: Company Strategy - Xiaopeng Motors' autonomous driving director stated that the company is confident in removing LiDAR, as the new AI system is built on short video data from customer driving experiences, which cannot utilize LiDAR data [2]. - The decision to adopt vision technology aligns with Tesla's approach, which emphasizes a pure vision strategy without LiDAR [4]. Group 2: Industry Comparison - Tesla has been a strong proponent of vision-based systems, claiming that multi-sensor data conflicts can reduce safety and that a camera-based system is more cost-effective and reliable [4]. - In contrast, many domestic manufacturers prefer multi-sensor fusion solutions, integrating LiDAR, cameras, and radar to enhance perception capabilities [7]. Group 3: Technology Advantages and Disadvantages - Vision-based systems are cost-effective and benefit from improving camera technology, but they struggle in adverse weather conditions, which can impair performance [5]. - Multi-sensor fusion systems leverage the strengths of various sensors, such as LiDAR's precision and radar's performance in poor weather, but face challenges in data integration and increased complexity [7]. Group 4: Market Trends - The penetration rate of advanced driver-assistance systems (L2 and above) has increased by 15.1% globally over the past three years, with China's rate rising from 0.5% in 2022 to 5.5% in 2024 [9]. - The Chinese market for automotive LiDAR is projected to exceed 3 billion yuan in 2023, with a compound annual growth rate of 124.20% over five years [10]. Group 5: Future Outlook - Both vision and multi-sensor fusion technologies are still in the early stages of development, with each facing unique challenges that need to be addressed for further advancement [12]. - The competition in the autonomous driving perception field is expected to intensify, with the most cost-effective solutions likely to dominate the market [13].