Huawei’s Li Wenguang Details Challenges and Evolution Path of Advanced Autonomous Driving

At the plenary session of the 13th Annual Conference on Intelligent Connected Vehicles, Li Wenguang, President of Huawei's ADS (Autonomous Driving Solution) product line, delivered a keynote speech, stating that intelligent connected vehicles are currently at a critical transition phase—from "human-machine co-driving" toward high-level autonomous driving. He noted that leveraging over 11.1 billion kilometers of accumulated intelligent driving data, Huawei’s latest ADS version has reduced severe accident rates to once per 7.2 million kilometers—4.37 times safer than human driving. Furthermore, since its commercial launch over a year ago, Huawei’s fully driverless Valet Parking (VPD) solution in closed campuses has demonstrated safety performance 30 times better than human-driven parking. Li emphasized that despite strong performance during the human-machine co-driving phase, this does not signify that Level 4 (L4) autonomous driving technology has truly matured. High-level autonomous driving imposes extremely stringent requirements on systems, including achieving million-kilometer-level stability without human supervision and ensuring deterministic scheduling capabilities in the operating system that “never fail.” Regarding technical approaches, Li expressed support for multi-sensor fusion solutions while raising doubts about pure vision-based approaches. He argued that even with image sensors upgraded to 12-bit resolution, dark-colored objects at night can still be obscured by background noise. Combined with challenging scenarios such as backlighting, glare, or rain-induced lens contamination, pure vision perception may fail to accurately detect obstacles or reliably mitigate risks. On the integrated cabin-and-driving architecture, Li voiced skepticism, pointing out significant differences between cockpit and autonomous driving systems in terms of algorithms, operating systems, and safety standards. Forcing integration, he warned, would not only fail to reduce costs but could also lower chip yield rates, as current hardware and software ecosystems are not yet ready to support such convergence. Based on this analysis, Huawei forecasts a steady industry progression: pilot deployments of Level 3 (L3) autonomous driving will begin in 2024, gradually scaling toward commercialization; L4 autonomous driving in low-speed scenarios will be prioritized for deployment, with widespread adoption of multi-scenario high-level autonomous driving expected by 2028.