BYD said its assisted driving systems are now deployed across more than 60 vehicle models and nearly 3 million vehicles globally, with severe accident rates reduced to roughly one-sixth of human driving levels based on airbag-triggered incidents per 10 million kilometres, according to Autohome.
The figures were presented by Yang Dongsheng, Senior Vice President of BYD Group and head of the company’s Automotive New Technology Research Institute, during the 13th Intelligent Connected Vehicle Technology Annual Conference in Shanghai on May 21.
Yang said BYD accelerated large-scale deployment of intelligent driving systems in early 2025 and now offers Level 2 assisted driving capabilities across almost its entire passenger vehicle lineup.
According to BYD, navigation-assisted driving activation rates have exceeded 50%, while parking-assistance usage has reached 86% among users.
The company also stated that parking-related scratches and minor collision incidents have fallen to around one-fiftieth of human-driving levels when using BYD’s parking assistance systems.
The automaker referenced its “God’s Eye” intelligent driving platform and associated parking-assistance guarantee programme introduced in July 2025.
The presentation comes ahead of BYD’s intelligent-driving strategy event scheduled for May 28, where the company is expected to announce broader deployment plans for updated God’s Eye systems across additional vehicle models.
Yang attributed part of BYD’s intelligent driving integration capability to the company’s Xuanji Architecture, which combines electronic architecture and electrification systems into a unified vehicle platform.
According to the presentation, BYD uses cloud-based world models and reinforcement learning systems for long-tail driving scenario simulation, supported by approximately 190 million kilometres of daily driving data generation.
Yang said the company currently performs algorithm updates every three days.
BYD also described the use of physical AI models within vehicles for predictive driving behaviour analysis and defensive-response calculations.
The company said its systems combine visual occupancy network detection with lidar-based occupancy detection to identify suspended and hollow obstacles during parking operations.
The presentation also highlighted extreme-condition testing programmes covering high-speed tyre blowouts, rain, snow and low-adhesion road conditions.
